Brief takeaways:

• This isn’t an AI bubble bursting. It’s a continued hedge fund deleveraging event that we are nearing the end of. The S&P 500 fell only 0.5% over the last week.

• Economically we remain fine still. Profits continue to grow, margins continue to grow, jobless claims isn’t up as much as the media wants us to believe, demand still remains through the roof.

• Listen to the companies and not the media or research notes. Lumentum is sold out through to 2029, ASML sees shortages to 2028, and Marvell is targeting a 10x in revenue by 2031. That’s not what a demand problem looks like.

• There’s no bubbles in valuations either. Earnings are growing faster than prices and multiples are shrinking (NVDA is at ~15x Fwd PE). We’re only 2-3 years into a decade long buildout.

• Brazil’s iShares MSCI ETF (ETF) sold off 21% post a Lula Da Silva surprise win. Markets were clearly all in on the other side and overshot. NU fell ~20% despite its first $1B profit quarter. That’s an overreaction though I still recommend caution.

November 1st Web-App Release

On November 1st I’m releasing my web-app to all subscribers (paid will have far more use).

Here are a few snips and teasers from the site:

1. Model your own stocks or just follow my models.

2. Create your own watchlist.

3. Track investment themes and the AI supply chain.

4. Track all my daily notes and my stance on individual companies.

5. Track my valuations.

Subscribe now

The panic was loud again this week but the price action told a completely different story.

Last week I titled my weekly article “More FUD. Less Logic”. I said the AI bubble didn’t pop…but hedge fund deleveraging caused the sell off. This last week just gave us more evidence of that.

By lack of logic last week here’s what I mean:

Tokenized stocks got hammered over the weekend…Hence retail investors called for a “Black Monday”…Then a small neocloud’s IPO in Australia fell apart…And then the FT and the rest of the media ran a story on whether the returns on AI will pay for itself.

But look at what actually happened:

• The SPY fell 0.5% for the week.

• The QQQ fell less than 1% last week and is now up 0.5% over the 5 trading days (at the time of writing this on Monday morning).

• iShares Russel 2000 (IWM) / Small caps (the most rate sensitive parts of the market) are up 2.3% over the last week.

That’s the “collapse” narrative the media are trying to paint. This is why I continue to believe we’re in the very late innings of this AI infrastructure sell off.

Here’s some more evidence backing up why we’re in the late innings:

1. Thursday was the day we saw the QQQ fall ~3%, yet over 60% of the SPY names on that day finished in the green. What does that say?

   1. It says that the economy is not being sold…but that a small group of crowded names are being sold by the same people at the same time.

2. And then listen to those at the centre of the AI buildout (rather than FT journalists for example):

   1. SpaceX (SPCX) said that compute demand is substantially greater than supply and they could rent out compute several times over.

   2. Nebius (NBIS) has raised prices twice in three months. In May it lifted on-demand GPU prices by an average of ~29%. Then in October it raised them again by another 17% all because demand for compute is outstripping supply. As a reminder, companies in a bubble cut prices to fill capacity. The opposite is happening.

3. And finally we have GPU rental prices sitting far ahead of standard 6-year depreciation schedules.

To put it simply…until we have one company in this cycle tell us demand is slowing, I don’t believe the risks outweigh the rewards at all. There is potential hedge fund deleveraging may set us back another 5-10% but the upside of the AI trade is still tenfold the downside risk.

We remain in year 2-3 of a 10+ year buildout and people are calling for “bubble.”

Is Hedge Fund Deleveraging Finished?

This is a very tough question to answer… especially compared to the “are we in a bubble?” question which is easy to answer.

Let’s walk through the logic:

1. Hedge fund net selling of AI stocks has now fully reversed a lot of gains since May / June 2025. This is suggestive of a huge amount of risk coming off the table all at once. Goldman Sachs reported that this is the second largest net selling of global AI stocks in the past year and hedge fund net exposure has not been this low (aside from Liberation Day) in the last five years. At these kinds of numbers, there’s simply not much left to sell. Once hedge funds have cut their exposure, the forced selling pressure dries up and price movements revert back to being set by what they should be set by…fundamentals.

2. November is approaching. This sounds trivial but it matters a lot as funds are judged month to month by investors. Each month is a “reset” so once a new month starts, a hedge fund can start adding risk again.

3. Selling last week happened across the entire board…not just in the equity markets. This is a very important distinction because it’s suggestive that hedge funds were almost selling whatever they could…rather than just the AI trade, or just equities. When selling pressure starts hitting much more than just equities…this tends to mean we’re in the late innings rather than the early innings of the deleveraging.

None of this is to say we’re done with the selling pressure. Saying that here would be naive…but I feel much more confident suggesting we’re in the later innings rather than the early innings here.

Macro Update

• S&P now trades at 18.7x forward earnings which is below its five year average of 19.8x. Meanwhile Q3 earnings are growing +29% with the vast majority of companies beating earnings so far (and likely many more to come). Earnings continue to outrun valuations. Not the other way around.

• The 10-year yield is now sub 5% which is still heightened but on the way down from +5.2% we saw at the start of October. As I’ve said the entire sell off, an investment boom raises the neutral rate. High yields alone aren’t a reason to be bearish on AI, but they will keep pressure on housing and small caps.

• AI capital is only just starting to hit the credit markets. Free cash flow has been mostly been used until now by the hyperscalers and it’s only now we’re seeing financing. Again, this is more evidence that this is extremely early on in the buildout.

• Junk spreads have hit an all-time tights which is a sign worth watching for the bears. Just one sign amidst many but alone it is not a bearish indicator.

Brazil: An Overreaction But Worth Being Cautious

Brazil had its own deleveraging event today but just with a political trigger.

Lula won Sunday’s runoff 51.8% to 48.2%. Markets had been all-in Bolsonaro after he took the first round of votes and prediction markets had a Bolsonaro win at ~80% on the day of the vote.

So when the results came in the Brazilian Real fell ~10% and EWZ fell ~21%.

For context, between the first round of results and Friday, EWZ rallied only 7% so today’s move didn’t just give that back… it went well past it.

That’s the same pattern as the AI trade we’ve seen. One crowded bet with everyone on the same side and nobody to sell to when it goes wrong. This makes it more of a positioning reset that can be taken advantage of rather than a sudden collapse in the Brazilian economy.

Some points worth noting on Brazil right now:

1. Lula will be constrained and anything remotely radical will be very hard to push through.

2. There is likelihood of political unrest. No major protests have been reported yet but observers warn any refusal post Bolsonaro’s request for annulment could trigger them. This is why the near term could continue to get worse for Brazil but this is something to monitor.

But finally on a more positive note… let me leave you with this chart below for returns under Lula since 2002.

It may not be all that bad for the stock market.

Commodities

We remain bullish on silver, copper, solar, and scandium.

No change in view this week. The structural case hasn’t changed:

• Silver is still in its fifth straight year of structural deficit with ~80% mined as a by-product so supply can’t respond quick enough. With AI’s push for efficiency (cost per token), silvers conductivity is extremely valuable.

• Copper is needed for every GW of data centre capacity, every grid upgrade and every on-site power plant. And with the grid struggling to keep up, the buildout needs more of all three, so copper demand only rises from here.

• Solar still remains a key play of mine on the power bottleneck and solar being one of the fastest and cheapest sources to build and permit.

Crypto

Bitcoin has continued to hold firm again whilst the AI trade struggled.

Over the last 6 months, BTC realized volatility is 30% vs +45% for many AI names. This is yet more evidence that the “Bitcoin is too risky” party is just wrong.

Perhaps some of the biggest news in the crypto world over the last week was Chamath (a prior BTC bull) suggesting “Bitcoin goes to zero” based on quantum computers breaking Bitcoin.

This is truly a question nobody knows the answer to right now but the worry is likely +15-20 years away at least and Bitcoin can upgrade to quantum-resistant technology as could everything else.

To put it into more perspective though, I think the near term worry is not with BTC. It’s with the general banking world.

Over the last 2-3 weeks we’ve seen increased headlines regarding cyber attacks and I think this is only the early stages of this. This alone should push banks towards cryptography and towards the blockchain which ends up being net bullish for crypto.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering at the time of writing.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Last week we touched on:

• Bloom Energy (BE) - Midsized position

• Coinbase | COIN - Large position

• Marvell (MRVL) - Small position

NuBank | NU (Medium Risk)

NU is a big position of mine. I was up ~20% on the trade but after today’s bloodbath in Brazil I’m back to my entry point…but most importantly with no current intention to sell.

A Bolsonaro win would have been a nice bullish catalyst for NuBank. But a Lula win is not necessarily bearish for NU.

Look at what NU has managed to do quarter after quarter over the last few years under Lula. Just last quarter we saw:

• Net income of $1.1B (the first $1B+ quarter in its history), up 49% year-over-year FX-neutral.

• Revenue up 39%.

• 140M+ customers.

• Conditional approval for a US national bank charter expected in 2027.

Based on current growth, I think analyst forecasts of $1.13 in EPS in FY27 seems completely feasible (~27% growth on FY26) which means at $12.36 we’re trading at ~10.9x FY27 EPS (or 0.4x PEG). That’s too cheap in my opinion even given the macro risks.

On the other hand, the bear case is fair…but it’s no different to the bear case we had 12 months ago. A weaker real hurts the dollar value of earnings and higher rates raise funding costs. It’ll be more important for us to watch delinquencies over the coming quarters but again this news is nothing new.

Personally, I’m already exposed to Brazil via a large position in NU so I won’t be adding anymore here but I remain very happy to be holding for the time being.

Read more

• Model Hub (follow my models, edit my models, or make your own models from scratch).

• Follow my portfolio in real time + my watchlist + daily notes.

• Track all stocks across +25 themes. Track rotation, relative strength, and tech moves vs QQQ.

• A technical and fundamental screener.

• & much more.

Paid subs will get editable access to this CBRS model on the site starting November 1st.

Prices remain at just $24 and will remain like that until launch. If you sign up before prices rise, you’ll lock in that price for the rest of the time you stay a subscriber.

I’m biased of course…but the offerings on this site are top-tier for the price. It’ll only get better too.

Subscribe now

Cerebras (CBRS) 2031 Model

Our model of Cerebras (CBRS) yields a risk-adjusted base case value of ~$485 per share in 2031 which is a ~34% compounded annual return from the $107.80 share price on 22nd October 2026.

The risked bull and bear cases are ~$700 and ~$165 per share respectively (539% upside down to 51% upside).

Because visibility beyond 2026 is genuinely limited, we apply a forecast risking factor of 0.77 to every case which is a 5% haircut for each of the five forecast years. That essentially means each year ahead we become 5% less confident in our numbers. Compounded over five years (0.95 to the power of 5), that takes ~23% off our 2031 values. We think this matters in today’s market: AI hardware, token pricing and customer commitments are changing so quickly that a five-year model is useful only if it is honest about how much could change along the way.

Before risking, the cases are ~$210, ~$630 and ~$900 per share.

Each case is a complete, internally consistent scenario rather than a single variable flexed up or down.

Overview of Our Model

Revenue & EBITDA growth

Cerebras is turning from a company that sells AI computers into one that rents out ultra-fast inference capacity. We believe that shift will define the majority of its value over the next five years. As such, in our base case this cloud inference business accounts for ~97% of Cerebras’s EBITDA by 2031.

That transition is already under way. In the second quarter of 2026, revenue from Cerebras’s cloud and services business rose 281% year on year to $126.0M on a reported (GAAP) basis while hardware sales fell to $54.1M from $70.3M.

Management raised its full year 2026 guidance for core revenue to $880-890M and plans to more than triple revenue in 2027.

In our base case, revenue grows from ~$890M in 2026 to ~$23.7bn in 2031 as installed inference capacity rises from ~60MW to ~3.2GW.

Hardware keeps growing to ~$1.2bn but becomes a small/immaterial part of the business. Much of the near term revenue is already contracted with $25.4bn of remaining performance obligations (signed contracts not yet delivered and billed) at 30 June 2026 which roughly double our cumulative 2026-28 base-case revenue of ~$11.7bn.

Main Assumptions

Our model is built on 13 assumptions that vary by scenario covering inference capacity and pricing, OpenAI’s renewal, hardware sales, margins, operating costs, capital intensity, dilution and the exit multiple.

Our base case is deliberately conservative at $19.57 of diluted EPS in 2030. Three main assumptions drive that:

1. Other customers pay less than OpenAI: The OpenAI deal works out at roughly $27M of revenue for every megawatt of computing capacity over the life of the contract. We assume Cerebras earns below that from other customers.

2. Cerebras’s price premium shrinks: AI companies are charged by the token. Today Cerebras charges about $0.75 per million output tokens against $0.17 at the cheapest GPU-based cloud providers. That means there’s a ~4.4x premium that customers pay because Cerebras models are far faster. We assume rivals catch up on speed or Cerebras cuts prices over time that that premium narrows to ~2.75x by 2031.

3. Margins improve but not to the maximum: Cerebras’s cloud gross margin rises from ~37% today to ~70% by 2031.

Model Simplifications

We value Cerebras on 2031 EBITDA multiplied by an exit multiple, plus net cash, divided by fully diluted shares. We then multiply each case by a risking factor of 0.77 (a 5% haircut for each of the five years we forecast) to reflect how hard it is to see out to 2031. We do not model interest income, the tax benefit of past losses or stock-based compensation expense. Dilution is captured through the share count instead. Each of these omissions should be small relative to the uncertainty in installed capacity.

We kept the model deliberately simple with two main segments: cloud inference, driven by installed megawatts and revenue per megawatt, and hardware system sales. Training revenue, sovereign deals and smaller customers sit inside these two lines rather than being modelled separately.

Bear, base and bull cases - 2031

Our bear and bull cases are deliberate and fair moves around our base case rather than best and worst cases scenarios. For what it’s worth, my bull case (un-risked and assuming a lot goes right) is far above the current bull case here.

The bear case assumes OpenAI does not renew after 2028 and that pricing and margins come under pressure. The bull case assumes OpenAI takes more of its 1.25GW option and pricing holds up better.

Key Assumptions

OpenAI

OpenAI is the foundation of our forecast.

In January 2026 it agreed to buy 750MW of Cerebras inference capacity through 2028 which was worth ~$10B. By the April 2026 S-1 the commitment had doubled to more than $20B (or $27M of revenue per MW), with an option on a further 1.25GW through 2030.

Our base case assumes OpenAI renews and that Cerebras wins steady business elsewhere reaching ~3.2GW of installed capacity by 2031. In our bear case OpenAI does not renew after 2028 meaning 750MW leaves the base in 2029 and only 30% of it is re-leased. That single switch costs ~$125 per share against the un-risked base case.

The speed advantage

Cerebras’s bull case rests on speed.

AI models write their answers one token at a time and Cerebras’s chips generate those tokens far faster than GPUs currently do.

On OpenAI’s gpt-oss-120b which around 15 providers serve, Cerebras runs at ~1,730 tokens per second which is about 6x the median provider. On Moonshot’s much larger Kimi K2.6 it ran at 981 tokens per second which 6.7x the next best GPU cloud.

That is why customers such as OpenAI, Cognition and Figma use Cerebras even though it lists its tokens at ~4.4x the price of low-cost GPU clouds.

In our model, speed ultimately shows up as price. We have to assume rivals narrow the gap so Cerebras’s premium falls to ~2.75x (from 4.4x) by 2031 in our base case (2.5x bear, 3.0x bull). As rivals get faster, customers will pay less extra for Cerebras’s speed so each token it sells earns less. Industry wide price cuts matter less though because we assume each new generation of Cerebras systems produces more tokens per megawatt to offset them.

Revenue per megawatt therefore falls from $11.45M a year in 2027 to ~$7.9M in 2031. Total revenue still rises sharply from ~$3.1B to ~$23.7B because of capacity growing from under 400MW at the end of 2027 to ~3.2GW by 2031.

Pricing per megawatt

The OpenAI contract implies ~$27M of revenue per megawatt over its life. Our base case assumes $11.45M per MW/year in 2027 falling ~9% a year to ~$7.9M by 2031 as Cerebras’s price premium narrows and as commodity token price cuts are offset by new system generations producing more tokens per megawatt.

The bear case assumes $11M with the premium narrowing to 2.5x by 2031. The bull case has $12M with the premium holding at 3.0x. Dynamic pricing for latency sensitive peak demand could add upside but we exclude it for now to stay a bit more conservative.

The path to high margins

Margins are thin today at Cerebras and that’s a key risk.

Core gross margin was 41% in Q2 2026 and GAAP gross margin 14% with management guiding to a 17-19% core operating loss for 2026.

We expect margins to rise as installed capacity fills up for three reasons.

First, Cerebras is currently renting data centre space and capacity from third parties while its own facilities are built and paying someone else’s mark-up as its own data centres come online.

Second, most of a data centre’s costs such as power contracts, cooling, staff and the equipment itself are fixed so each extra paying customer on the same capacity adds revenue at little extra cost.

Third, each new generation of systems produces more tokens for the same power and space.

In our base case, cloud gross margin before depreciation climbs from 37% in 2026 to 70% in 2031 and EBIT margin reaches ~47%. That takes diluted EPS from a loss in 2026 to $19.57 in 2030 and $26.44 in 2031.

For what it’s worth, this is far below forecasted numbers from UBS of $24.35 in EPS in 2030.

Who pays for the data centres

Capital intensity is the biggest open question in our model and one of the biggest in the AI trade right now.

Cerebras had $8.6B of cash after raising $6.4B in its May 2026 IPO plus an $850M credit facility. Our base case assumes partners and customers fund most of the build out leaving Cerebras to spend ~$5M per incremental megawatt.

If Cerebras has to fund chips and facilities itself…capex could be 2-3x higher and it would likely need more equity or debt. That is why our bear case assumes 355M fully diluted shares in 2031 against 345M in the base case.

Our bull case also carries a slightly higher share count 350M but for a different reason. OpenAI’s warrants for up to 10% of the company vest only if it buys 2GW of capacity which happens in a strong outcome.

That dilution is a cost of success and the extra earnings in the bull case would more than offset it.

Exit Multiples

Our base case comes down to three operating numbers for 2031: ~3.2GW of installed capacity, ~$7.9M of revenue per megawatt and a ~61% EBITDA margin. Together they produce ~$14.4bn of EBITDA which we value at ~13.5x.

If we then add ~$22.5bn of net cash and divide by 345M diluted shares we get ~$630 per share before risking and ~$485 after.

Why 13.5x: We apply the multiple to 2031 when Cerebras is no longer a start-up but is still growing revenue ~23% a year at a ~47% EBIT margin.

At 13.5x EBITDA, our base case equates to ~24x 2031 earnings, ~17x EBIT and a ~4% free-cash-flow yield. We think that is a fair price for that profile rather than an overly stretched one but tweak that how you will.

EBITDA can flatter a capital heavy business but because we assume partners fund most of the build out… ~60% of our 2031 EBITDA turns into free cash flow.

Why 10x and 14x: Our bear case uses 10x which is the kind of multiple investors pay for mature computing infrastructure. That would be in line with what Cerebras would look like if the OpenAI deal weakened and its speed premium faded.

Our bull case uses only 14x which is conservatively barely above the base case. We would rather put the upside into what the business achieves than into a higher multiple: the bull case is ~45% above the base because of more capacity and better pricing, not because the market necessarily pays more for it. I think this could be tweaked higher though but a lot of this will be dependent on the macro down the road.

Opportunities not included in the model

OpenAI’s 1.25GW option

OpenAI holds an option on a further 1.25GW through 2030 on top of its 750MW commitment. Our base case does not depend on it being exercised. If it were, OpenAI alone would account for 2GW, close to two-thirds of our base-case capacity and its warrants would vest.

Hyperscaler distribution

Cerebras plans to bring disaggregated inference to Amazon Bedrock in the first quarter of 2027 and expects a joint offering with AMD for up to 5x throughput gains. Selling through AWS could open Cerebras to enterprise customers that would never sign a direct contract.

Sovereign and on-premise AI

Governments and regulated industries increasingly want AI compute within their own borders. Cerebras built its early business on G42 in the UAE and could repeat that model elsewhere. Our hardware line grows at 20% a year in the base case which leaves room for a large sovereign contract to surprise to the upside.

Key risks

• Customer concentration: OpenAI’s $20bn+ contract is roughly 80% of Cerebras’s $25.4bn backlog. A delay, a cut or a move to rival chips hits revenue directly which is why our bear case models a non-renewal after 2028. This makes CBRS quite a concentrated bet on OpenAI’s continuous growth for the time being.

• Rivals close the speed gap: Nvidia, with its CUDA software ecosystem and its Groq licence, AMD, Broadcom’s custom chips and Google’s TPUs are all targeting fast inference. If they catch up, Cerebras’s price premium falls faster than we assume.

• Funding the build-out: We assume partners fund most data-centre capacity. If Cerebras has to fund it, capex could be 2-3x higher and require more equity or debt.

• Near-term margins: Cerebras is renting third-party capacity while its own data centres come online which compresses margins: GAAP gross margin was 14% in Q2 2026.

• Dilution: Fully diluted shares of ~314M compare with 237.6M outstanding, and the 2026 equity plan reserves a further 42.7M shares with annual increases from 2027.

• Share overhang: The IPO lock-up ends two trading days after Cerebras reports third-quarter results, releasing the last tranche of pre-IPO shares. Insiders have already sold including a filing by the CTO to sell over $150M of stock in August. This is a near-term supply issue rather than a threat to 2031 value in our opinion but there appears to be a lot more room to the downside at the moment.

• Volatility and valuation: The shares are down ~72% from their post-IPO high of $386 in May 2026 and our valuation depends on an exit multiple five years out which has a lot of uncertainties.

  ---

What I’m Personally Doing

Read more

Brief takeaways:

• This isn’t an AI bubble bursting. It’s a hedge fund deleveraging event. Crowded AI positions were forced to all sell at once. The S&P 500 is still less than 3% from all-time highs.

• Listen to the companies and not the media or research notes. Lumentum is sold out through to 2029, ASML sees shortages to 2028, and Marvell is targetting a 10x in revenue by 2031. That’s not what a demand problem looks like.

• There’s no bubbles in valuations either. Earnings are growing faster than prices and multiples are shrinking (NVDA is at ~15x Fwd PE). We’re only 2-3 years into a decade long buildout.

Another week. More FUD. Less Logic.

The loudest (and laziest) call in the market and on X right now is that the AI bubble has burst.

It lacks any proper evidence beyond a red week and completely ignores what companies are actually telling us. It also conveniently forgets that the S&P 500 is less than 3% away from all-time highs.

Last week was one of the noisiest weeks in the market I’ve ever experienced. If you read the headlines you’d assume that the AI trade is finished just 2 years into the buildout. For comparative purposes…other infrastructure buildouts in history rarely last less than 7 years. The fibre buildout lasted ~8 years and the railway buildout lasted ~15-20 years.

Yet, the FUD was real. CoreWeave (CRWV) sold off. Barclays told clients the best of the AI trade may be behind us. Gartner said AI data centres will generate less than half the revenue needed to justify the spend by 2030, and Bain said the industry will come up $800B short.

Let’s look at what actually is happening and try show put some logic to all of this:

1. The S&P 500 (SPY) closed Friday less than 3% away from all-time highs (+12% this year).

2. The Nasdaq (QQQ) is up 15% YTD.

3. Hyperscalers (which are the companies supposedly as per media headlines set to stop paying for compute)…are now within a few percentage points away from highs.

4. Lumentum (LITE) is sold out through to 2029. The CEO has also said that “to a large degree, commitments are non-cancellable.” and that no customer has slowed, delayed, or asked to renegotiate an order since interest rates rose.

5. ASML (ASML) expects shortages through to 2028 with AI chip demand extending into 2030.

6. Marvell (MRVL) is targeting a 10x in revenue to $70-90B by 2031.

This is absolutely no sign of a bubble bursting. This is narrative from companies doing the actual selling and this is who you should be listening to.

On November 1st I’m releasing my web-app to all subscribers (free & paid though paid will have far more use).

Here are a few snips and teasers from the site:

1. Model your own stocks or just follow my models.

2. Create your own watchlist.

3. Track investment themes and the AI supply chain.

4. Track all my daily notes and my stance on individual companies.

5. Track my valuations.

Subscribe now

So what did happen last week?

Hedge funds were massive net sellers last week. They were forced to sell the same crowded positions at the same time. That’s very different from an “AI bubble”.

The main issue came because everyone is in the same trade (AI infrastructure and/or AI winners). So when everyone is in the same trade…there are no buyers when the trade turns against them.

And plus when hedge fund positions start falling…we see the following:

• Very limited ways to hedge trades at the moment. The “AI trade” is semis, optics, energy, software. So a pure index put isn’t a good hedge at the moment. Hedging via Brazil or alike is probably the best bet…hence why I’ve done that.

• When losses all show up at once because the entire AI trade is down…margin terms tighten which forces more selling.

Put those together and you essentially get a big deleveraging event. The purest AI names fell the hardest whilst the index, Europe, Brazil, and some other markets all held firm. It’s evidence that some of the biggest hedge funds in the world, despite seemingly being diversified by ticker count…actually weren’t diversified at all outside of the AI trade.

The unknown is how long this deleveraging actually lasts which is impossible to tell. Whilst we’re in this near-term deleveraging…fundamentals don’t matter. But over the course of a 3-5 year period fundamentals do matter and that should remain every retail investors focus.

The next catalyst / reminder for this demand is over the next 10 days through to November where we’ll see all hyperscalers (META, AMZN, MSFT, GOOGL) report.

Are We In A Bubble?

No.

Earnings are rising way faster than prices.

Multiples are shrinking.

Example: NVDA now trades ~15x NTM PE (it’s lowest PE since 2015). Some estimates put us ~13x NTM PE as well.

An AI bubble would see the opposite happening (rising multiples and falling earnings growth).

I touched on this in the introduction section as well but it’s important to realize exactly what you’re reading here.

We are in year 2-3 of the AI buildout.

Previous infrastructure cycles ran for 7-15 years.

Actual spend is ~$1 trillion so far today. We have forecasts for ~$10 trillion.

The Risk:

The main risk I see to the above is an oversupply of data centres (not demand for compute). That sounds a bit counterintuitive so let me explain:

Data centres are getting built faster than most can comprehend. Industry estimates puts that close to $1T of locked-in, long-term leases signed in 2026 alone. The issue is the chips (in particular memory) to fill them can’t keep up at the moment.

This creates a risk that by the end of 2028 or mid 2029 we have ~10 GW of fully built data centres with no chips to put in them. That will lead to a slow down in new orders which could be a clear risk the market reacts to.

The positive to take from that is another reinforcement that this isn’t a demand problem. Demand for compute still far exceeds supply. It’s just a mismatch within the supply chain itself and understanding the dynamics of this for your investments is key.

The pricing power will remain with those players who are constrained (memory and optics). Less pricing power will be with those data centre developers and/or some infrastructure suppliers.

This is why I think a “bubble” is a 2030-2031 story and not a story today. We’ll start to see the first cracks show up when supply overtakes demand in a small part of the supply chain. That could be 2028.

From there, we’ll likely see developers cutting prices to fill space or lease growth slowing (maybe in 2029/2030)…and from there there can be legitimate, and more evidence based discussions around an AI bubble.

Until then, it’s just all noise, FUD, and engagement farming on X.

Macro Update

• The S&P 500 remains close to record highs with earnings growth ~29%. That earnings growth is not the issue. The issue is everything beneath this:

• Credit: July’s selloff was purely equity but this selloff saw credit crack as well. Once credit cracks that leads to more deleveraging.

• The 10-year remains ~5.1%. I remain of the opinion that this is only a small part of the story and understanding why yields move is much more important. High yields alone do not automatically mean you should go bearish on the market. An investment boom simply raises the “neutral” rate. If businesses expect high returns from AI, they’re willing to borrow at higher rates and the equilibrium shifts up.

  • Barclays have argued this case many times. They’ve argued that AI-driven productivity could force the markets to reprice long-run Fed expectations and pus the 30-Year Treasury to +6%.

  • However, high yields do create a more challenging environment for the much more rate sensitive plays like housing and small caps.

Commodities

We remain bullish on gold & silver miners, copper, solar, and scandium.

I’m going to focus on silver today because it’s been hit the most (along with solar) but structurally we still remain bullish.

Here’s my silver bull case in a few bullet points:

• AI Buildout Demand: Silver has the highest electrical and thermal conductivity of any metal. As AI moves from training to inference, and agents drive constant token demand, the focus moves to cost per token which means efficiency needs to be stellar. In other words, any electricity or heat lost across the system will come under more and more scrutiny. Silver helps reduce electrical and heat loss across the system. It’s also a very small fraction of the entire build out meaning a rise in silver prices won’t affect buyers materially.

• Other Demand: Silver was added to the 2025 US critical minerals list likely because of its role in drones.

• Supply can’t respond: We’re currently in the fifth straight year of a structural deficit that we can’t get out of. About 80% of silver is mined as a byproduct of lead, zinc, and copper so supply can’t rise easily.

Crypto

Bitcoin’s price action last week told us a lot. Whilst the entire AI trade struggled, BTC held firm.

That’s a strong counter for believers that BTC is “too volatile” or “too risky”.

It’s also a positive sign for higher beta equities as BTC has historically been a proxy for higher beta equities.

The main opportunity still remains tokenization and regulation which keep moving faster than people think. In fact, Morgan Stanley last week put out a tokenized asset forecasted market size of $3 - $10 trillion by 2030.

We’ve also recently seen Stripe purchase OpenRouter which is a sign for what’s likely to come.

Think about why a payments company would be buying a token routing company. This is the new economy we’re entering.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Last week we touched on:

• Securitize | SECZ - No position

• Coinbase | COIN - Large position

• Cerebras Systems | CBRS - No position

Bloom Energy | BE (Medium Risk)

Power is the biggest constraint on AI over the next 2-3 years and if you believe that, then Bloom Energy (BE) is the cleanest play on the solution.

The reality is that the US grid is running out of capacity. From 2028 onwards, reports suggest that over half of new US data centres will have to generate their own power on site. Here are the key advantages to understand:

• Permitting: Fuel cells produce a fraction of the emissions of turbines or engines. This is often the difference between a fast standard permit and years of major source review. For example, Oracle’s Project Jupiter hit permitting delays and then they redesigned around Bloom. Nebius did the same.

• Labour: Fuel cells don’t need much labour like pipe welders or boilermakers (extremely scarce in the current buildout).

• Market Share: Bloom has almost the entirety of all fuel cell orders.

• Speed to Power: The economics makes Bloom the best route. Today, it’s all about speed to power. A power plant for 1GW data centre costs ~$5B and that can generate returns ~$100B in inference revenue.

For the next 2-3 years (minimum), BE is the solution to the bottleneck.

Read more

I still remain cautiously bullish on AI infrastructure and crypto (if your time frame is longer than a few months) though I appreciate there’s a lot of moving parts that makes buyers a little apprehensive.

We remain in this challenging position between oil rising (+2% today), midterm politics, 10 year yields, Korean markets, AI infrastructure demand, AI infrastructure economics and bubble fears (very overblown), and much more. Each of these is a legitimate input. And together it’s creating one of the noisiest short-term environments many investors have experienced.

It’s led to a market that seems like it’s lacking buyers (yesterday saw the lowest volume since 2003) to drive any material bullish price action…hence why we’re just in this period of constant rotation in and out of sectors, and constant chop.

I would have expected we’d be seeing more institutional buying here or more bullish reiterations. Instead, the signals remain mixed.

Yesterday alone, we had BlackRock downgrade their take on US equities and Goldman turn bullish on stocks.

Are We In A Bubble?

Short answer: No.

Longer answer: Calling a rising market a bubble is the easiest / laziest call in investing. It requires no analysis and eventually every bear will be claim to be right about something…even after missing out on hundreds of percent gains.

In the dot-com bubble we had companies like CSCO trade at 100x earnings and companies with $0 in revenue worth billions. Today’s market is completely the opposite - we have earnings growing faster than prices so multiples are compressing and PEG’s at 30 year lows. It’s a market centered around macro fear, fear of future financing, and future payback vs how good the technology is…rather than a “bubble.”

There’s a good reason for this compression though. AI makes it cheaper to build competitors so moats are much weaker than they were a decade ago meaning a company’s sustainable profits may be much less.

The other sign of a bubble is when we build far more supply than there is demand. If that were the case we’d see idle capacity and suppliers cutting prices to find buyers. If you look at almost any quality AI infrastructure name (LITE, AAOI, MU, SNDK, NVDA, NBIS etc) the complete opposite is happening.

This is not to say we won’t eventually reach a bubble. I expect that’s a 2030-2031 time frame personally. Until that happens, the best thing you can do is to stop predicting doom and focus on companies with insatiable demand for the next 3-5 years.

I share my exact portfolio with all numbers to subscribers. That’s just one of the many parts of the service I offer.

And in the next two weeks I’ll be releasing my new web-app which is a game changer.

Here’s a couple of snips from it:

Subscribe now

Prices will be kept the same from now until 2 weeks after release. If you subscribe now you’ll forever hold that price.

Macro Update:

• The SPY still trades at a 30 year PEG low.

  • We’re seeing growing profits and compressing multiples…not indicative of a “bubble” in any sense.

• 12th October 2026 was the lowest SPY volume recorded since 2003.

• Housing has broken below 18-month support and mortgage rates just hit 7.5%.

• Only 18% of utilities and 12% of autos trade above their 200-D moving average whilst 73% of tech sits above the 200-D moving average.

  • You can make the argument both ways here. You can make the SPY market breadth argument at all time highs or you can realize that we’re in a new digital led economy (I’m a firm believer that the latter is true here).

  

Commodities

We remain bullish on gold & silver miners, copper, solar, and scandium.

Today’s focus is on solar which is where we’re seeing the most bullish signs at the moment as TAN approaches a key support level:

Solar makes up ~51% of planned 2026 capacity additions with battery storage ~28% (total of 67.7 GW vs 6.3 GW for natural gas fired capacity).

India has been the clearest example of this trend so far. In India, solar is now some of the cheapest electricity in the world with contracts being locked in for 25 years. On the other hand, coal is getting more expensive every year as fuel costs rise with inflation so the gap between the two continues to widen.

This cheap power is turning into an industrial advantage in India. For reference, data centers there cost ~30-50% less to build in the US.

All data is leading to solar over the next few years but the more difficult question becomes how to invest in it. We’ll be doing an article on this in the next two weeks.

Currently, I remain long SHLS and TAN.

Subscribe now

Crypto

Bitcoin (BTC) continues to hold above the $82,500 support level despite the rising yields, rate hike talks, and every other macro fear.

Unless structure turns bearish again, there’s no reason not to be bullish.

This is a positive sign for higher beta equities as BTC has historically been a proxy for higher beta equities.

The opportunity still remains tokenization and regulation which keep moving faster than people think. In fact, Morgan Stanley just today put out a tokenized asset forecasted market size of $3 - $10 trillion by 2030.

We’ve also recently seen Stripe purchase OpenRouter which is a sign for what’s likely to come. Think about why a payments company would be buying a token routing company. This is the new economy.

Micro Update

• Lumentum (LITE) CEO said that the company is completely sold out through to 2029.

• Marvell (MRVL) laid out their forecasts to hit $70-90B in revenue by 2031 which shines light on the entire optics trade (COHR, LITE, AAOI, CRDO, FN, etc)

  

• Applied Optoelectronics (AAOI) completed its $600M ATM selling 5.7M shares at an average cost of $105.36.

• Coinbase (COIN) reports their Q3 earnings on October 29th with options pricing in ~11% move either way so far.

• UBS released their Cerebras (CBRS) model with $23B in revenue in 2030 at 46% operating margins (more on this in the “What’s Mispriced?” section.

• Truist estimated AI Cloud ARR will rise from $310B (2025) to $2.1T in 2030 which is a 47% CAGR.

• Rosenblatt estimated that Amazon (AMZN) AWS revenue will hit $334B in 2028. That’s just below my $341B estimate but in line with my estimates to hit $1T in ARR by 2035.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Last week we touched on:

• NuBank | NU - Top 3 Position

• Recursion Pharmaceuticals | RXRX - No position

• USA Rare Earths | USAR - Small position

Securitize | SECZ (High Risk)

SECZ is a pure play on tokenization with $5B of tokenized AUM and the potential to capture a lot of the economics as long as the incumbents don’t try and compete (so far there’s more signs of incumbents partnering with SECZ rather than competing with them).

SECZ owns a large part of the stack:

• Transfer agent: Keeps the official record of who owns each token.

• Broker dealer: Lets SECZ buy, sell, and distribute tokenized securities for clients.

• Administrator: Handles the back office work.

Valuing SECZ is a little more difficult as we’ve yet to see any longer term forecasts or multiple earnings calls. But at ~$2B MC already off just $71M in FY26 revenue, we’re already seeing the stock trade at a premium.

That premium is because of how embedded SECZ has already become and the potential monetization routes this opens up. The current administration focused business has been a great way to get the business into the right areas of the ecosystem, but that alone wouldn’t be what can revalue SECZ from a $2B company to a +$10B company.

The opportunities are therefore:

• Through the Computershare deal, SECZ can start taking a small slice of all US listed equities that become tokenized. Currently, SECZ mostly earns fees from tokenized funds so the greenspace ahead is quite substantial.

• The SEC’s innovation exemption allows tokenized stocks to trade around the clock 24/7 so SECZ could earn trading fees in markets that never close. With more hours, global access, and instant settlement against stablecoins…we’ll likely see far more trading volume and therefore more fees.

• Through its partnership with Cantor, SECZ can earn fees through tokenized IPOs and follow-on offerings.

Read more

It’s not a Deep Dive into specific companies or a detailed review of a specific theme. But it’s a concise look at the opportunities in the market right now, the data backing up why the opportunities are there, and much more.

This report focuses on what I’m calling the Big 5 at the moment:

• AI x Biology

• Financial Revolution

• Commodities

• Photonics

• Energy

Note: These themes change and Q1 27 Investment Opportunities will unlikely be the same as these. I also own more themes in my portfolio than these five here, but these are the ones I am most excited about right now.

A Look At The Macro

We’re in the early innings of a technological revolution

I’ve spoken about this week in week out in my What Markets Are Saying & What’s Mispriced? series. You can see recent editions here.

We are entering a period where everything is converging. Here’s a couple of examples of what I mean by that:

1. The space theme & compute

2. AI & crypto

3. AI & Materials Science & Biology

People are fully aware of the above, and yet markets and investors still get spooked by illogical headlines, doomsday articles targeting clicks, and the AI bubble and 30Y yields.

In response:

AI Bubble: We currently have an SPY PEG at 30 year lows. We have CapEx spend as a percentage of GDP at levels not seen since 1998 but importantly the earnings are growing at a completely different rate to what we experienced with other productivity booms like the railroad, the internet, and electrification.

We also have:

• Oil on the verge of coming down.

• Real GDP growth likely to head above 3%. Anthropic’s bull case indicated a world where we see 15.4% GDP in 2030. Almost all of that divergence is from 2027 onwards so we’re just on the verge now in Q4 2026.

30-Year Yields: Yields are elevated but doomer headlines on X fail to discuss why yields are rising. Yields can rise because of strong growth and that’s exactly what is happening. People also underestimate how much sensitivity these hyperscalers actually have to yields. Corporate interest is at lows relative to GDP.

The Big 5 Themes

AI x Biology

In our view, AI could have the most impact on biology/medicine/drug discovery out of all industries.

Relative to today, the addressable market for biology/drug discovery seems almost infinite because of the flywheel that we’ll see.

More data → Better drug discovery models → Better therapeutics, diagnostics, and longevity medicines → Longer life expectancies → More diseases discovered → More data

This was always the flywheel but AI compressing this cycle majorly. The reasons for this compression is twofold:

1. Genome sequencing: Sequencing has made biological data cheap to generate but without AI it was always slow to reach. Interpretation that took years of manual analysis now happens in days.

2. Drug development timeline: AI is set to cut drug development timelines by ~40% or more. The important part to understand is that failed drug data is now just as valuable as successful drugs. This wasn’t the case before AI.

Phase 1 of this AI x Biology theme was the evidence that AI could integrate into biology. That’s why companies like Twist (TWST) which made it fast to write DNA, and 10x Genomics (TXG) which let scientists read genomics at extraordinary resolution, became early beneficiaries.

Both were plays on AI entering the biology world, not necessarily on AI delivering results.

Phase 2 will be all about the output:

• Which gene shall we edit?

• Which protein shall we express?

• Which nucleotide shall we change?

And there are only a few companies with machines that can do this which in our opinion are CRISPR Therapeutics (CRSP), Beam Therapeutics (BEAM), and Prime Medicine (PRME).

• CRSP cuts DNA at a specific site.

• BEAM edits individual bases.

• PRME rewrites sequences.

Other Ways to Invest:

Diagnostics: Detecting disease earlier and with more accuracy because of AI (combined with multiomic signals)

• Guardant Health |GH: GH is building a proprietary data layer for AI capturing epigenomic data at scale. It also has software “apps” on that data that can flag things like tissue biopsy misses, like a lung tumor, or the origin of an occult cancer.

• Tempus | TEM: TEM is the data layer behind AI drug discovery with its data licensing segment growing 36% last quarter. It’s winning $100M+ multi-year deals from pharma such as AZN, GSK, MRK, and BNTX.

• Veracyte | VCYT: VCYT sells genomic tests that tell doctors how aggressive a cancer is and the best treatment. Each test adds to GRID (VCYT’s database) of genomic and slide data that researchers use to find new biomarkers that AI tools can learn from.

Treatment:

• Intuitive Surgical |ISRG: ISRG is turning its surgical robots into an AI data platform: its Case Insights AI analytics tool is the second layer of a five-layer AI stack that runs surgeon supervised autonomy. Now trades far more reasonably but still far above 1x PEG.

• PROCEPT Biorobotics | PRCT: Another surgical robot but trades far more reasonably than ISRG.

Note: Numbers are forecasts in the above section may not be realized. Not all companies are held within my portfolio. You can see my portfolio at beta.mmmtwealth.com

Subscribe now

The New Website…Coming Soon

Before we get into the rest, I want to show you some snips of my new website which is in the final stages of production and testing.

You can use it how you see fit.

Want to just follow my portfolio, watchlist, entries, and exits? Easy:

Want to follow my daily notes in a simple or detailed way? Easy. Every note I write gets summarized by an agent so you can read it in seconds or minutes…however you prefer.

Want to model your own stocks or view my models? Easy:

And lots more.

October will be an exciting month for MMMT Research. Join us now whilst prices are cheap.

Subscribe now

Financial Revolution

The applications for crypto are a lot more understood in today’s agentic driven world than they were even 2 years ago.

5-10 years ago prices of Bitcoin and Ethereum were based far more on speculation rather than application.

That’s a very different story today. Crypto finally has its use case which is always on, programmable infrastructure that agents can use.

This is the big convergence of AI and crypto.

Agents operate continuously across applications, without sleep, and outside of business hours. If they want to act economically (which they will), they need financial infrastructure that’s always on, that’s programmable, and that’s machine readable. Our existing financial system is the complete opposite (business hours, KYC gates, intermediary approvals, and settlement windows).

A personal finance agent that is negotiating your mortgage rate at 3am, can’t use SWIFT. But it can use USDC on Base. Or an AI treasury manager reallocating idle corporate cash can’t route through a traditional brokerage. But it can interact with an on-chain lending protocol.

This is why crypto, after years of speculation, now has a clear purpose.

Two markets are opening up here: Programmable dollars through stablecoins and programmable financial assets through tokenization.

Stablecoins: This is the foundation everything described above is getting built on. Before stocks can move on-chain, or before agents can transact in milliseconds, a digital dollar needs to work. That’s stablecoins.

Tokenization: A share of stock sitting in a brokerage account currently does nothing. It just sits there idle. Tokenization will change this. That same stock sitting there could serve as collateral, a lending asset, or a hedging input…all on programmable rails. Tokenized US Treasury funds have now reached ~$15 billion (vs ~$31 trillion in outstanding Treasuries which is ~0.05% showing you how large the greenspace is ahead).

The investment thesis here is a little more complex. Big picture, it’s clear stablecoin volume is increasing, and tokenization is happening. But that rise doesn’t necessarily mean every play in the stack accrues value.

The question becomes about what layers monetize it the best. Here’s what we think:

How to Invest:

• Coinbase | COIN: The market still treats it as a crypto trading venue with the stock moving in line with BTC and ETH prices. COIN is building the infrastructure behind the entire payment revolution. The only main unknown to me is exactly how much of the economics move to COIN’s earnings vs how much accrue to the apps built on COIN’s infrastructure. I think the next 3-5 years where we’re still building the infrastructure will be very bullish for COIN though.

• Robinhood | HOOD: HOOD has a major benefit in that it has an active retail base already. Their tokenized stock demand is growing rapidly, albeit from a very small base of ~$30M per year which is immaterial relative to $5.8B in forecasted FY26 revenue. The main upside for HOOD is if they can monetize this retail demand into something a lot more than just chain fees.

• Securitize | SECZ: This is the purest play on tokenization out there positioned as the legal link between token and real security. It takes care of all the backend tokenization compliance and legal risks. So far, incumbents are partnering with SECZ and using their expertise rather than building in house which is a bullish sign. We think a lot of the value accrues here.

  

• Circle | CRCL: CRCL is the leading expression of stablecoins as its the default regulated stablecoin. The risk for me with a company like CRCL is if own bank-issued stablecoins (like SOFI is doing with SOFIUSD) capture the reserve income themselves rather than sending it to CRCL. It’s feasibly something that is a risk to CRCL. On the other hand, CRCL could continue to be the default universal dollar standard and that in itself would value CRCL well above its current $20B valuation.

Commodities

Breaking this down into:

1. Copper

2. Silver

3. Scandium

Copper:

The copper thesis is dead simple:

Copper demand is set to rise 13-14 million tonnes by 2040 whilst supply is only expected to grow by 4 million tonnes.

Mine output fell 1.1% in the first half of 2026 and this isn’t being fixed by new mines that take ~15-20 years to bring online.

It really as simple as that. Copper is a way to expose yourself to the physical AI buildout, give you some exposure to the commodities, and play the long term supply demand imbalance.

Personally invested in Global X Copper Miners ETF (COPX) as my lower risk play and Trilogy Metals (TMQ) as my higher risk, higher reward play.

Silver:

There’s a few reasons to be bullish silver in our opinion:

1. US M2 (money supply) was $1.5 trillion in 1980. Today, it’s $21.9 trillion (a 14x expansion). If you scale the 1980 peak of $50 an ounce for silver by that factor then you’re left with a money-supply adjusted high of ~$700. At $59 per ounce today, with stronger fundamentals than in 1980 (next points), we think silver prices are materially undervalued.

2. Industrial demand for silver rose 4% in 2024 (to record highs) and this was before the AI infrastructure buildout really started. Silver has the highest electrical and thermal conductivity of any metal so it appears in connectors, switches, relays, semiconductor packaging etc. As AI shifts from training to inference more focus moves to efficiency and this is where silver comes in.

3. The World Silver Survey said we’re in a structural deficit of 148.9 million ounces in 2024. Making this deficit up is not easy since ~80% of all silver is mined as a by-product of lead, copper, and zinc. The market is already tight (just like copper) and it’s only getting tighter.

Personally long iShares Silver Trust (SLV), Hecla Mining (HL), and Contango Silver & Gold (CTGO).

Scandium:

Scandium remains the underrated metal in the AI buildout and even with SREMF rising over 1,000% over the last year, I still remain bullish.

The main demand drivers for scandium remains:

• Solid-oxide fuel cells: To understand the demand for SOFC’s, understand that Morgan Stanley is forecasting Bloom Energy (BE) to be 20% of the entire fix for America’s data centre power shortage through to 2028.

• Aerospace & Defense: Scandium gives metals like aluminium a strength to weight ratio ~4x better than steel.

• 5G/6G: Scandium is in all microchips within cell phones which allow for 5G and 6G wireless communication.

Scandium is still priced in line with prices set in China. It’s only now, through SREMF, that the true value of scandium will start to become a lot more transparent.

Photonics

On 06/10/2026 we had the Marvell (MRVL) investor day where we got the first photonics piece of visibility through to 2031. For reference, forecasted revenue for MRVL in FY26 is $8.2B and management just forecasted a $80B midpoint in revenue for FY31 (57.7% CAGR).

We’ve also had:

• Citibank TMT: “Lasers and optical fibers are being positioned as the “next HBM.”

• Goldman Sachs upgraded their forecast for the global optical module market to ~$150 billion by FY28 compared to $68 billion in FY26.

• Lumentum (LITE) said “we’re completely sold out. There is no end in sight.”

• A report out from Jefferies suggested that as AI systems get larger, a smaller share of the cost will go to GPUs and a larger share will go to networking. I.e. in today’s rack networking makes up ~8.6% whilst in more futuristic multi-rack systems networking will climb above 22%.

It’s more evidence that the next 2-5 years will see value accrue to the optics world.

How to Invest:

• Marvell | MRVL: MRVL is the “connective tissue” of every hyperscaler date centre that Jensen Huang famously called “the next trillion dollar company.” Given their $80 billion FY31 forecast, I think there’s a good chance of this.

• Ciena | CIEN: CIEN’s WaveLogic technology encodes data more efficiently so that you effectively can multiply the amount of data/information a cable can carry without physically adding more fibre cables.

• Applied Optoelectronics | AAOI: This is a clean play on the markets laser shortage. AAOI manufactures the laser diodes and transceivers that convert electrical signals into light. Out of almost all stocks I know, it’s demand is the highest by far as they work towards a forecasted $471 million / month in just transceiver revenue by H2 27 (vs $455 million in FY 25 total revenue).

• Credo Technology | CRDO: CRDO is a pure play on the last metre of the data centre - the high-speed connections inside server racks where the optical fibre hands off to the chip. It’s also a play on copper not being displaced by optics as quickly as many are forecasting.

• Himax Technologies | HIMX: HIMX is an overlooked name in the optics world. Its precision optics technology is central to TSMC’s next-gen CPO platform which is a solid position. Management have also guided to $100 million in CPO revenue in the first year of mass production (FY26) with a much more material ramp up in 2027-2028.

• STMicroelectronics | STM: STM is a slightly riskier one in that it’s got a lot of exposure to the automotive world…but it’s also got a lot of potential in the photonics world with a multi-billion dollar contract with AWS to supply optical chips to its data centres.

The Power Bottleneck

The US grid runs at roughly 50% utilisation. Most of the infrastructure is built…it’s just sitting idle and will be for 5+ years.

What’s keeping most of this energy stranded is the absence of storage. Plants run inefficiently because demand is lumpy - they spin up and down rather than running flat out which drives up the cost of every kilowatt-hour they produce.

Battery storage changes that by absorbing surplus energy when demand is low and discharging it when demand peaks allowing plants to run more consistently and spread fixed costs evenly.

But storage alone isn’t the answer. Solar paired with storage can be deployed in the next 12-24 months vs grid connection queues which have 5+ year waitlists. On-site solar is the only route to new power at speed and it’s at a cost ~90% cheaper than 2010.

And then nuclear is the longer-duration answer with an unmatched baseload.

Solar:

We’ve heard for years that solar is the answer to everything…along with nuclear. And so far, that hasn’t materialized like many forecasted until now.

But the physics behind solar still makes complete sense and I think the future relies on a combination of:

1. Behind-the-meter energy (Bloom Energy | BE)

2. Solar with batteries

3. Nuclear (probably further down the line than the above).

The reason BTM and solar work best is purely because of speed. Despite the grid upgrade still being a worthy investment…grid connection queues are now 5-8 years and with turbine companies like GE Vernova (GEV) with orders ~5x their current capacity…the only possible route is to figure out how to get power quickly.

On site solar plus battery energy storage can generate power in ~18-24 months. It’s also by far the cheapest power source (one of the reasons investing in the pure solar plays has been a horrible investment in the last 5-10 years). Prices have fallen ~90% since 2010 and with energy demands set to continue to soar for years to come it seems like one of the few ways to meet the demand at a reasonable cost is via solar.

Along with a lot of different sub-niches of the AI trade, I think the solar inflection point is in 2027-2029 with 2026 being a more difficult inflection point year because of policy changes in China.

If you’re bullish solar, you have to be bullish on battery storage.

How to Invest:

Investing in the solar theme isn’t easy. Some of the purer panel manufacturing plays have struggled as the product is just a commodity where China is setting the price.

We’d look in the picks and shovels plays:

• Nextpower | NXT: NXT is still in its building phase but what it’s building across the solar stack is quite exciting: trackers, inverters, batteries. It recently acquired Prevalon Energy making it a key player now in the battery storage niche as well as a less-commoditized play on solar as well.

• Shoals Technologies | SHLS: SHLS makes the electrical balance of systems that every solar project needs. SHLS also has battery storage with ~$20 million in revenue in the last quarter.

• Fluence Energy | FLNC: This is the largest pure play on battery storage with a $2.9 billion backlog. The underlying thesis is still intact but there’s been some real recent struggles as management cut FY26 guidance by $600 million due to welding issues at their Houston facility. The stock now trades sub 0.5x sales and is likely great value if management can execute. It’s hard to call a bottom here…and FLNC certainly isn’t a stock I’d be interested in owning during a downturn but it’s interesting to look at here if you have a strict place to cut your losses. That’s what I’m doing right now.

  

Nuclear:

Referring you to my recent piece on nuclear here:

Disclaimer

Disclaimer: The Accuracy of Information and Investment Opinion

The content provided on this page by the publisher is not guaranteed to be accurate or comprehensive. All opinions and statements expressed herein are solely those of the author.

Publisher’s Role and Limitations

Make Money, Make Time serves as a publisher of financial information and does not function as an investment advisor. Personalized or tailored investment advice is not offered. The information presented on this website does not cater to individual recipient needs.

Not Investment Advice

I still firmly bullish AI infrastructure and bullish crypto.

Unfortunately, with the choppy market and the constant doomsday narrative around yields I’ve seen many investors switch to a more bearish narrative even despite the QQQ and SPY sitting above all key MA’s with a bullish structure.

My personal portfolio is now just 1.4% away from its high.

I share my exact portfolio with all numbers to subscribers. That’s just one of the many parts of the service I offer. And in the next two weeks I’ll be releasing my new web-app which is a game changer. Here’s a couple of snips from it:

Subscribe now

Right now, there is no sign to be bearish.

As I said last week:

“Investors need to zoom out and realize that this next decade is likely going to be very different from the last three decades…and if that means yields head higher then that could become the reality for a few years.

We’ve been through a period of high yields in the 8-12% range before and the SPY averaged returns ~14% between 1982 and 1995 which is evidence high yields does not necessarily equate to weak market returns.”

The key part to understand isn’t that high yields = bearish stock market returns (like you see scattered all over X for engagement).

Yields rise for different reasons. The popular answer is that yields are climbing because of debt worries or worries about the strength of the dollar. That’s not how I see it at all. If that were happening, we’d see it in credit spreads (reflective of corporate risk) and in the dollar.

Total debt (government + corporate) has not risen relative to the economy. Debt is ~370% of GDP. Debt 20 years ago was ~370% of the economy.

The only difference is that the mix of government vs corporate debt has changed.

Government interest costs are at highs but corporate interest costs are at lows. For the equity market…that matters a lot because companies are far less exposed to higher rates than the headline numbers suggest.

So to make the argument that yields are rising because of debt is a thesis based shallow arguments.

The correct explanation in my opinion is that growth is strong and the bond market is just pricing that in. The Fed is largely following that signal rather than leading it…(as it always does).

A legitimate counterargument I can vouch for against this would be a consistent trends in softer jobs data (weakness in the economy) and golds resilience (to an extent). Gold is holding up nicely in that we remain in a bullish structure (no break of the $363 level on GLD) but relative to other commodities it’s performed slightly weaker so I’m more on the fence about that argument.

Macro Update

• The SPY still trades a 30 year PEG low.

  • We’re seeing growing profits and compressing multiples…not indicative of a “bubble” in any sense.

• Mike Selig (CFTC Chair) officially proposed its first round of rules to regulate the US crypto markets

• Hedge Funds and Asset Managers were net buyers of US tech equities in September. This was the only net bought sector in the entire market.

• Rental prices for older GPUs have risen from where they were at the start of the year…which is more evidence that cheaper intelligence is raising demand rather than decreasing it…all of which is happening in the early innings of AI (98% of households still aren’t paying for AI).

• Flavio Bolsonaro defeated current Brazilian President Lula da Silva in the first round of Brazil’s presidential elections. Final elections are on 25th October. This is a massive net positive for Brazil’s next few years with EWZ rising ~19x from 2002 - 2008 during the last Brazilian bull market. We’re currently sitting at the same level as 2008 highs.

• Morgan Stanley forecasted inference spend to outpace training spend with a CAGR of 42% from 2026 - 2030.

  • Very bullish signs for the obvious NVDA, AMD, and CBRS plays.

Commodities

We remain bullish on gold & silver miners, copper, solar, and scandium over the long term.

Copper | COPX: I’ve seen no new copper related news throughout the week but I just wanted to reiterate this divergence we’re seeing between copper and gold over the last year. This is copper showing some incredible relative strength in a short-term commodity bear market.

As we touched on last week, Deutsche Bank are forecasting a $22,050 per ton copper price by Q2 2027 with their analysts calling it the more acutely scarce copper market since the 1980s.

You don’t have to overcomplicate it.

Own the safe way to play the theme: COPX

And own a higher risk way to play the theme (if you have a higher risk tolerance): TMQ

Silver | SLV: SLV continues to hold its ground in this range between $48 and $65. Until we see a clear break below July’s lows, I remain bullish on silver.

For balance, one piece of data I am tracking quite closely at the moment is solar installations. This is not the main bull case for owning silver stocks for me, but solar is a key demand driver and a reduction here could lead to some weaker sentiment for silver.

Crypto

Bitcoin (BTC) continues to hold up nicely above this $82,500 support level we’re seeing.

As I noted last week, this “risk on” asset is continuing to hold despite the bears arguing about rising yields, rate hikes, and all other macro fears.

Bitcoin remains a leading indicator for higher beta equities and until we see a break of support here, we remain bullish crypto and bullish equities.

Micro Update

• Applied Optoelectronics (AAOI) completed its $600M ATM offering. Perhaps, now this is out the way the focus will revert back to scaling capacity and FY27/FY28 forecasts.

• Goldman Sachs is forecasting hyperscalers (AMZN, GOOGL, META, MSFT, ORCL) will be a combined $510 billion by 2030.

• Micron (MU) reported earnings last week where management reiterated that supply will stay tight until at least 2028.

• Lumentum (LITE) CEO said that they estimate to be under-shipping demand by 70% in 2027.

• Cathie Wood bought $20 million worth of Cerebras (CBRS) last week.

• Rosenblatt estimated that Amazon (AMZN) AWS revenue will hit $334 billion in 2028. That’s just below my $341 billion estimate but is in line with my estimates to hit $1 trillion in ARR by 2035.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Last week I touched on:

1. Coinbase | COIN and my 2031 price target

2. Bloom Energy | BE and my recent buy

3. Contango Silver & Gold | CTGO

NuBank | NU (Lower Risk)

Here’s a more detailed look at the numbers for NU and also why I see NU being a big beneficiary of a potential Bolsonaro presidency:

1. Cost of credit: Cost of credit was $1.7B against $3.7B of net interest income. If Bolsonaro does win, Bloomberg estimates we’ll see 10-year yields cut ~70 basis points. This will ease debt and increase repayment rates which currently has the biggest effect on NU’s earnings.

2. Commodity boom: Brazil will mainly benefit from a commodity boom which should increase cash flow for producers, suppliers, employees, and the wider Brazilian economy. This is only a net positive for the banking industry.

3. Currency conversion: NU has suffered in the past a lot from FX conversions as they report in dollars. If we see the Brazilian Real rise ~3% (as estimated by Bloomberg) then we’ll see an increase in the dollar value of the same earnings.

4. Small business: Lower long-term yields cut borrowing costs for smaller firms. This is an area NU is starting to monetize now so cheaper capital is a net positive for NU.

Current forecasts for NU are $1.13 in EPS in FY27 and $1.46 in FY28. Personally I think this is conservative but I want to work from these numbers at the moment given management haven’t been too vocal on forward guidance…perhaps because of the macro.

$1.46 in EPS in FY28 would be a 31.4% CAGR over 3 years which is less than SOFI at 43.8%…but not materially that different given the difference in multiple.

Right now, most fintechs aren’t trading anywhere near 1x PEG but I suspect we get there over time in a more bullish macro environment. I also think it’s fair to use that as a multiple given I’m remaining fairly conservative on EPS numbers.

$1.46 * 31.4x EPS (3 year CAGR for 1x PEG) gives us a $45.8 stock (or 3x from today).

Read more

And I’m still firmly on the side of seeing a choppy market for the near future as negative macro sentiment pushes the market down whilst an ongoing stream of incredibly bullish micro news continue to hold the market up.

The market looks like it’s creating a bit of a double top here but we still sit firmly above all important moving averages and support zones so until that changes my bias does not shift.

We’re in this period where everyone is focused on 10Y and 30Y yields and pointing towards that being the key reason for a bear market just like 2008.

It’s a fair point but it continues to err too much on the side of history rather than what actually is happening in AI right now.

We’re in the middle of the biggest technological revolution of all time (and likely that we’ll ever see).

Being bearish here is being too careful on a couple of macro worries vs what’s actually happening in the world.

Investors need to zoom out and realize that this next decade is likely going to be very different from the last three decades…and if that means yields head higher then that could become the reality for a few years.

We’ve been through a period of high yields in the 8-12% range before and the SPY averaged returns ~14% between 1982 and 1995 which is evidence high yields does not necessarily equate to weak market returns.

Also to put it into perspective, we saw rising yields in 2022 (granted from ~2%) but there was endless bearish narrative back then regarding the effects on the market.

So to put it into perspective…yes yields are high and yes this will likely add continued pressure to the upside in the near term but basing an entire bear case of rising yields misses the bigger picture entirely.

If yields start to stabilize in the 5-6% range over the next 6-12 months that’s not bearish at all in the middle of a strong economy and the greatest tech revolution of all time.

Macro Update

• Chief Economist at Apollo outlined a world where agents could cause a bank run by sweeping household cash into accounts paying 3-5% instead of the 0.1% national average.

  • I outlined this agent usage in my COIN Deep Dive from last week. Everyone will focus on the clear bear narrative around banks but instead investors should focus on what this means for agent usage and the technology revolution we’re entering. Just more evidence to be bullish crypto and bullish AI infrastructure as agentic usage multiplies.

• Meta’s Muse showed that broader consumer AI adoption is coming fast.

  • It had 642,000 daily active users in the first 12 days post launch vs 231,000 for ChatGPT.

• Goldman Sachs put out forecasts of $1.2 trillion in hyperscaler CapEx in 2027 with $1.4 trillion in 2028. This is in a similar ballpark to NVDA’s near term forecast but Jensen believes we can hit $3-4 trillion by 2030.

• US 30 year yields just hit their highest level in more than 22 years. A noteworthy macro point but one I still believe is just a small part of the wider narrative as I spoke about above.

• The price of tokens is collapsing faster than any other technology in history. This should continue to drive more and more token usage across the supply chain from AMZN, GOOGL, META, and NVDA to the memory players like MU and SKHY.

The big macro news I want to focus on this week is what we’re seeing in the stablecoin and tokenization world as things are moving very quickly over the last two weeks since the CLARITY Act failed. Here’s a very brief overview of what’s happening…a longer piece will be a tokenization specific article (later this week):

1. SEC now allows limited on-chain trading of tokenized US stocks and the SEC chair said that tokenization rules are moving towards a longer-term framework.

2. CFTC outlined plans for 24/7 on-chain tokenized financial markets.

3. Federal Reserve has now proposed rules for stablecoin issuers under the GENIUS Act.

4. SEC staff released new guidance on how securities laws applies to crypto.

This all happened in 7-10 days post the failing of CLARITY Act. It’s a great sign that even though the bill failed…crypto regulation is coming and it’s coming a lot quicker than most think.

Stablecoins are already a $300 billion market with very limited laws and regulations and with very few agents acting economically yet.

We’re still incredibly early and today remains the time to stay long.

Commodities

I remain bullish on gold & silver miners, copper, solar, and scandium.

If the Fed hikes rate it will worsen the interest expense problem and if the Fed cuts it will worsen the inflation problem. That means the Fed are in a sticky position and makes a clear argument that allocating a % of your portfolio towards hard assets is sensible.

Gold & Silver:

We’ve seen continued consolidation below the 200D MA for SLV and GLD and there’s mounting pessimism in the trade as yields rise which makes for a contrarian buy/hold signal for me.

For reference, I’m a current owner of HL which I’m just up 7% on and SLV which I’m down 2% so both sit around my DCA average cost still.

There’s also mounting divergence between silver and copper over the last year as copper continues to lead the market which makes the silver trade much more interesting.

The reason there’s a lot more bullish sentiment around copper is that people understand the direct link between copper and the industrial buildout. With silver it’s less well known but still incredibly correlated to the industrial buildout.

Silver has the highest electrical and thermal conductivity of any metal so as AI shifts from training to inference, the game changes from raw power to efficiency which is why silver becomes an incredibly important part of connectors, switches, relays, packaging, and thermal interfaces.

And secondly, we’re in a period of record buybacks for silver and gold miners. Historically miners have been heavy on dilution whilst also suffering huge losses. We’re now in a completely different period and EPS results for miners over the next 12-18 months will likely be very strong.

Copper:

Copper has held up much better of late but I’m still sitting around my average cost on both TMQ and COPX. I think this holding up relative to other commodities tends to be a bullish signal.

Supply is still massively underappreciated with nations like Chile experiencing output slipping even despite heavy investment during to falling ore grades and rising project costs. This has essentially created an environment where the mining industry simply can’t ramp up production quickly enough for the demand we’re seeing from data centers and grid expansion.

The only answer is higher prices.

It’s a simple supply and demand equation and I think people are beginning to forget just how tight this is getting / will continue to get.

I remain long both COPX and TMQ.

Crypto

Bitcoin and Ethereum continue to perform well and both have broken their bear market structure as the agentic wave continues to accelerate. We’re seeing a lot of momentum in the stablecoin and tokenization regulation world as well which is improving sentiment.

Historically, crypto prices are generally a good indicator for the higher beta equity market so if history repeats this is another reason to stay bullish.

Micro News

• Citi partnered with Coinbase (COIN) to bring stablecoin payments to instituional clients. This is a prime example of “first you fight them, then you join them.”

• SoFi (SOFI) became the first national bank to bring settlement live across Mastercard’s global payments network. It’s migrating its entire $25B card program to blockchain-based settlement.

• MongoDB (MDB) CEO CJ Desai just left his role effective immediately to run Meta (META) Enterprise Platform and said META has a “rare advantage” across models, infrastructure, and agents.

• NuBank (NU) is rumoured to be acquiring UK’s Monzo which put the shares down 8% on trading today on potential of share dilution and a hit to earnings over the near term.

• Truist just estimated that AI cloud ARR could rise from $310 billion (2025) to $2.1 trillion in 2030 which is a CAGR of 47%.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Coinbase | COIN (Low Risk)

COIN is sitting just below a 5% allocation in my portfolio right now but I’m actively looking to make it above ~6% in the near term assuming prices stay around this level which is a level I’m more than happy to buying at.

You can see my COIN Deep Dive here and all the valuation work too. It’s open to all subscribers for the next week.

Bloom Energy | BE (Medium Risk)

I recently made a starter position in BE and have set out levels (as per chart below) where I will look to significantly increase my position size.

Read more

Introduction

Company: Coinbase

Ticker: COIN

Market Cap: $52B

CEO: Brian Armstrong

Headquarters: New York

The market continues to see COIN as a leveraged bet on crypto trading volumes and Bitcoin spot prices. When BTC fell from $127k down to $58k (54% drop), COIN fell 68%.

And vice versa too. When BTC rises, COIN tends to rise more.

Given the revenue mix for COIN in 2021 until last year, this was a fair way to model COIN. But given the lack of diversification in the revenue and PE ratios above 70x…the stock never interested me more than just owning Bitcoin itself.

This is a very different story in 2026 as 88% of net revenue is now not tied to BTC spot prices. And the market is slowly waking up to it which is giving investors an opportunity to enter before we’re past the tipping point.

COIN is now split out between:

1. A regulated exchange (spot BTC, derivatives, equities, prediction markets)

2. On chain settlement stack (Base + x402 + CDP)

3. A stablecoin platform (USDC distribution)

From here on out, COIN’s price will be driven more and more by the on chain settlement stack and stablecoin performance. BTC spot prices will continue to drive the narrative in the near term, but over the course of a 3-5 year window, we’ll start to see a decoupling from the volatility (although even that is lessening) we see in crypto prices.

The Wider Theme

This section is broken down into:

1. Bitcoin and Ethereum and Why you Should be Bullish Crypto

2. Stablecoins

3. Tokenization

4. Agentic Commerce

Bitcoin and Ethereum and Why You Should be Bullish Crypto

As this article goes on, I hope you’ll understand that COIN will not be a stock that will move in line with crypto prices. But the underlying tailwind behind it is of course the adoption of crypto so it’s important we touch on this.

Bitcoin (and Ethereum) has been a technology that only until recently people saw as a store of value rather than a technology that can be widely applied. Only more recently are we seeing crypto actually have clear use cases.

Capital and attention from Bitcoin’s rise over the last 10-15 years has created the buildout in exchanges, blockchains, stablecoins, settlement rails etc and it’s only now when agent count is multiplying month on month that crypto actually has applications and alignment with what is happening with AI.

Like I’ve said before in a very simple sense:

In other words, there’s a growing connection between AI and crypto now. As broader AI adoption transitions into agents acting on behalf of users (rather than an advanced chatbot)…we’ll see demand, utility, and applications increase across the digital asset space too.

As we start to see agentic count creep up slowly…more and more people will start to understand crypto and the true applications. That alone will reprice the asset well before we hit maturity.

Stablecoins

Imagine a world where agents are calling an API 400 times per second. Each call costs ~3/10th of a cent. Visa’s fixed fee is $0.03-$0.04 per payment meaning the payment fee is larger than the thing being bought. On top of that, the payment then takes a day or two to actually arrive (with nothing clearing on a weekend).

An agent calling an API 400 times per second 24/7 using this payment method won’t work here. It’s a crazy mismatch. Yet this is the situation emerging if stablecoins don’t become mainstream.

This is why the next phase in commerce is all about verification, speed, and automation which is exactly the reason stablecoins were created. And if you really start to zoom out and understand what’s been going on, you should have a pretty clear idea that the past 5-10 years has been the slow buildout of the infrastructure layer for the next era in commerce.

The last hurdle that we’re currently going through is mainly regulatory and how stablecoins will be governed. That shows you how early we truly are here to this revolution.

The scale we’re working with now is no longer trivial despite the need to get over regulatory hurdles. Stablecoins now have a $300 billion market cap with transaction volume over $11 trillion (same range as Visa and Mastercard’s annual payment volume) growing at 80% CAGR compared to ~8.5% CAGR for ACH (Automated Clearing House - the network US banks use to move money between accounts).

Relative to $93 trillion for ACH, we’re still just at a fraction of the total movement, but the growth rate tells you which system is built for where commerce is heading.

The important thing to understand as an investor is where the economics actually land. As agent driven transactions mature, the transaction fees will become less and less. When that does happen, the durable value will sit with whoever earns yield on the reserves backing the stablecoin… and not who gets the fee on each transaction.

This is the company COIN is building. It’s not just a platform where these transactions happen. Through its commercial relationship with CRCL it also captures a substantial share of that reserve yield itself.

Tokenization

When you understand tokenization and the benefits of tokenization, I think it’s hard not to think the risk to reward lies on the buyers side.

Whether that’s through COIN, or through HOOD or even through pure plays like SECZ is up to you.

The core thing to understand on tokenization is that assets held (which in the US is ~$195.9 trillion) becomes a lot more mobile. By mobile I mean that assets become easier to value, easier to convert, and easier to use as collateral…essentially assets that actually start working vs assets that are just sitting there idle.

Idleness is the problem tokenization solves.

Idle assets are locked behind slow settlement, fixed market hours, manual administration, and intermediaries who gatekeep who’s even allowed to hold it. Tokenization gets rid of each of these bottlenecks and turns idle capital into working capital.

This leads to:

1. Settlement in seconds

2. Markets that never close

3. Self-executing assets

4. Idle assets become productive collateral

5. Exposure for everyone to the capital markets

Right now we have ~$14.8B in tokenized Treasuries and only just a recent launch of tokenized equities. All US government debt is currently $31 trillion so $14.8B represents only 0.047% so the greenspace ahead is substantial.

A modern brokerage app today earns most of its revenue from activity around the asset…not from the asset itself. A share sitting in a brokerage account doesn’t really do anything unless it’s being bought or sold. It’s deadweight on a balance sheet. Tokenization changes this. If even a small part of a custody balance can be reused as collateral, lending inventory, or a hedge of some sort, the economics of those “idle assets” becomes a lot more interesting for both the platform (COIN) and the holder.

But you don’t need the entirety of this $195.9 trillion to become tokenized for this number to matter. Even a small percentage becoming more mobile is still measured in the trillions of dollars of newly productive capital.

For COIN, it’s betting on exposure to this entire theme through the infrastructure layer (custody, exchange, on/off ramps) that has to exist for any of this $195.9 trillion to actually move.

McKinsey puts tokenized real world assets (ex-stablecoins) at ~$2-4T by 2030. Citi’s base case sits at $5.5 trillion for tokenized securities specifically with a wider range of $2.7-$8.2 trillion. Some forecasts from BCG and others go far higher when you then include tokenized deposits and stablecoins in the definition too.

We get into the numbers of what this means for COIN in the valuation section.

Agentic Commerce

This all comes together with another big tailwind which is agentic commerce.

Just last week, Meta’s consumer agent - Muse - was released and it hit #1 on the app store.

The key difference between Muse and the rest of these frontier models so far is that its the first genuine reach event for consumer agents. Muse launched with 100 million tokens per week which has essentially allowed the general population (those without technological expertise) to put an autonomous agent to work for them on real tasks.

→ Using an agent to find flight credits to book flights.

→ Using an agent to cancel unused subscriptions.

→ Using an agent to shop around for better insurance deals.

→ Using an agent to move your money into a higher yield savings account.

It’s going to be a world where each agent is essentially just performing a load of smaller transactions beneath it. All this requires transacting with services the agent may have never touched before.

If you then extrapolate that to hundreds of millions of Muse users…you get an explosion in the amount of machine to machine transactions that no existing payment rail is built for.

Morgan Stanly & Bain are forecasting the agentic commerce space to be ~$300-500 billion by 2030.

Again, we get into what this means numerically in the valuation section.

Subscribe now

Where We Are Today

Everything above is the exact gap crypto fills.

Not necessarily the payments between parties that already have a billing relationship…but the agents discovering a new API or data course mid-task with no prior account, no API, and nothing set up prior.

This is exactly where COIN’s Base sits. Base already carries ~98% of x402 machine payment volume transactions.

The story today though is that agentic payments are slim. The data is showing no material increase in x402 payments because agents aren’t massively economically useful right now. Agents are assisting with telling humans what to buy and where to buy it…but the next stage of giving agents access to a credit card is not happening.

The reason the numbers are so small today is because there’s millions of tiny pings but no meaningful spend yet. I.e. we’ve seen 5.7 million one-cent payments on Polygon ($57,000 total).

But this does tell you what’s happening.

The infrastructure is being built well ahead of the traffic it was designed for. Things like wallets, settlement rails, and payment standards are live and working, but the population of agents genuinely transacting with purpose hasn’t shown up yet.

Muse is the clearest example of this directionally but even it (the most consumer facing example of an agent) still asks for human approval before spending money. If human approval is still needed, and connecting credit cards is therefore still the route, then agentic x402 payments simply aren’t a material thing yet.

Until agents can transact with stablecoins and until the regulatory environment and the infrastructure continues to get built out this remains a thesis.

But just be aware what is happening despite x402 payments being thin:

• Visa (V) has a framework called Visa Intelligent Commerce (VIC) built with Cloudflare (NET) which lets merchants cryptographically verify an agent is legitimate before treating a payment as trustworthy. This launched back in June 2026 and shows us that the big players aren’t sitting on the sidelines waiting to see if agentic commerce becomes a thing. They’re actively building now.

• AWS shipped native x402 support into CloudFront and AWS in July 2026. This means payment-gating for agents is now built directly into core AWS infrastructure and not something developers need to add on themselves.

• Cloudflare (NET) has also partnered directly with x402 by verifying payments at its network edge (i.e. closer to the agent) rather than having to route back to an origin NET server. It’s also built a “Monetization Gateway” that lets websites, datasets, and tools charge agents directly through the protocol.

• Composition is shifting towards genuine commercial use with payments of $1 or more growing from 49% to 95% of total transaction value.

The Coinbase Opportunity

Some firms out there touch one piece of the above. I.e. Robinhood (HOOD) and Securitize (SECZ) are great plays on tokenization. Circle (CRCL) is a pure play on USDC / Stablecoin adoption.

But nobody owns the entire stack aside from COIN…and that’s ultimately the thesis.

You have:

1. Stablecoins: COIN holds the commercial relationship with Circle that gives it a direct claim on USDC reserve economics. This is a key layer of the stack that will survive fee compression as agentic payment volume scales.

   1. Renewed the same CRCL deal through to 2029. USDC is the #1 regulated stablecoin in the world by transaction volume but sits behind Tether if you blend regulated and unregulated stablecoins by market cap and AUM.

   2. COIN is also becoming a multi-stablecoin platform. It already supports PYUSD (PayPal) and USDT (Tether). The aim for COIN is to have “good economic arrangements” with every major stablecoin and not necessarily just USDC.

   3. Base (COIN’s own chain) did $32 trillion in stablecoin transfer volume over the last 12 months already.

2. Tokenization: It sits at the custody, exchange, and on/off-ramp layer that any tokenized asset has to pass through to actually move. Tokenization doesn’t work if cash and securities can’t settle on the same ledger. COIN owns that infrastructure that makes that settlement possible.

   1. COIN has already launched tokenized equities on Base for non-US customers. US customer access is still pending the SEC’s “innovation exemption.”

   2. COIN has also launched pre-IPO perpetual futures for non-US customers which gives retail access to private companies and it’s another big benefit of tokenization.

3. Agentic Commerce: It owns Base, the chain that carries ~98% of x402 machine-payment volume.

   1. I’ve separated agentic commerce out throughout this article aside from in the valuation section where it’s grouped in with total S&S revenue. The revenue isn’t material enough yet to be given a separate line item in COIN’s reports or in my valuation work.

   2. Earliest signs so far we’re seeing are agents buying inference and compute.

   3. Brian Armstrong has been explicit in saying that Base has a huge competitive lead vs newer entrants like Stripe (Tempo) and Robinhood’s L2 with about a 2 year head start.

4. Crypto Adoption: It runs the exchange that still captures the bulk of institutional and retail BTC/ETH trading activity that gives it the foundation and relationships that funds and reinforces everything else built on top. COIN is now much more than a single product exchange now though and is discussed by management as the “Everything Exchange” with spot trading, derivative trading, traditional equity trading, and prediction markets.

   1. COIN also now holds ~12% of all on-chain assets. That’s a huge lead vs other peers.

5. Other Bets: Beyond the four core layers, COIN holds a growing set of smaller, higher risk, higher reward bets that represent free optionality on where crypto infrastructure heads next:

   1. Prediction Markets: Already scaled to $100M ARR. If this becomes a recognized asset class the way sports betting did and options did…COIN is already a regulated leader building out this product. This growth has come from existing COIN users with zero marketing spend yet.

   2. Derivatives: The Deribit acquisition has given COIN a lead in the crypto options and futures niche. Margins here are much higher than spot crypto trading. It also diversifies revenue away from the whole “consumer retail cyclical trading” bear COIN case.

   3. Coinbase One: Crossed 1 million paying subscribers in Q2 2026 even during a down market for crypto trading which is more evidence that membership has value beyond just trading volume.

And if history repeats itself in any way, owning the stack tends to be the best investment. NVDA is the cleanest example. It didn’t just sell GPUs…it built CUDA (the software layer), NVLink (the interconnect layer), and much more. In other words, it captured value at every layer of the AI compute stack rather than just the chip. Apple did the same. Amazon did it with AWS (owning compute, storage, and the platform layer).

With COIN, you have the same structure (though less monopolistic) forming across crypto.

The bet of course isn’t that COIN becomes crypto’s NVDA. It’s that a company sitting on the dollar (stablecoins), the settlement layer (Base), the venue (the exchange), and the asset-migration (tokenization) should be worth more than the sum of its parts once those layers start reinforcing each other.

Now here’s the valuation:

Valuation

Total Transaction Revenue:

This is the bucket of revenue that the market already knows about. It’s the cyclical part of the business and it’s not the core bull case I have for COIN in anyway.

It’s still obviously important to model in the numbers because it’s a significant chunk of today’s revenue and will be a significant (but hopefully shrinking relative to the rest of the business) section of the overall revenue by 2031 onwards.

I model ~$7.2B in transaction revenue in 2031 as a bull case. TTM transaction revenue was ~$3.32B so that’s a CAGR through to 2031 which I see as viable based on:

• Consumer spot notional $1.0-1.4T at a consumer take rate of 0.25% (down from today as fees compress). This is the volume that will be the lumpiest by far.

  • That’s ~$3.6B

• Institutional spot notional of $3-5T at 0.05%. This is the larger volume that will exist when retail are quiet.

  • That’s ~$2.4B

• Other transaction revenue is derivatives, non-spot pairs, and newer products like scale like prediction markets which are already at a $100M ARR.

  • I forecast ~$1.2B here

Total: ~$7.2B

These numbers rely on:

1. Digital asset turnover being higher in 2031 than it is now on average.

2. Consumer take rates going down but volume outrunning that.

3. “Other” transaction revenue has real potential.

A base case scenario has COIN generating $5.4B in transaction revenue which is only ~9.6% CAGR through to 2031.

Total Subscription & Services Revenue:

This is broken down into three separate items in my model:

1. Stablecoin revenue

2. Agentic commerce revenue

3. Other revenue

Stablecoin Revenue:

This is the bulk of the opportunity and where the risk to reward makes sense if the below numbers materialize.

US Treasury Scott Bessent forecasts stablecoin revenue to be ~$3T in 2030. We’ve also got forecasts from Bain & Co at ~$3.8T.

In my model, stablecoin revenue will be the following:

My model forecasts $10.8B in revenue in 2031. That’s:

$3T total stablecoin supply by 2031 (aligned with Bain & Co and Scott Bessent)

30% share of stablecoin supply for USDC → $900B suggesting USDC stock

$900B x 3.4% (yield) x 35% (blended share of USDC economics for COIN) → $10.8B

The above would be my base case.

A more bullish case would assume COIN is nearer to 45-50% of USDC economics.

i.e. $900B x 3.4% (yield) x 45% → $13.8B.

That 35% assumes distribution dilutes as USDC goes into banks, other apps, and non-COIN wallet.

Agentic Commerce Revenue:

McKinsey estimates the 2030 market to be upwards of ~$1T in the US and $3-5T globally. Directionally I think this is spot on. Timeline wise it’s much harder to gauge particularly with recent regulatory failings (i.e. CLARITY).

The TAM for COIN is therefore not this $3-5T figure. Most of that will settle on cards, bank rails, Amazon, Shopify, Stripe, Visa, etc. COIN’s realistic slice is the machine native layer: agent wallets, USDC settlements, Base, x402.

So in terms of forecasts for COIN themselves it’s a question of how much agentic volume flows through COIN’s rails combined with a realistic blended take rate.

• Bull Case: Agents become a default checkout path and USDC/Base/x402 become default settlement options. I give this is a 20% chance.

  • $500B-$1T * 0.5% - 1% → $2.5B - $10B

• Base Case: Stablecoins win a meaningful share of agent-to-merchant micropayments and some cross-border / always-on checkout, plus agent trading grows.

  • $100B - $400B * 0.4% - 0.8% → $400M - $3.2B

• Bear Case: Agents pay for APIs, compute, research, and some digital goods on x402/USDC but retail checkout stays on cards.

  • $20B - $50B * 0.3% - 0.6% → $60M - $300M

This shows you the ranges that are present depending on just how much and how quickly the financial system changes. To reiterate, I believe directionally we will head towards the bull case figures. The timeline is much harder to anticipate right now.

The big, stock moving numbers only really appear if COIN becomes a core settlement and wallet layer for agents.

Total Revenue:

In total this puts my total revenue between ~$16.5B and ~$23.3B. I place my bull case revenue ~$22.7B with base case ~$16.1B and a bear case at ~$9.6B.

That means my bull and base case are currently above analyst estimates with a bear case below.

Net Income:

I assumed a net income margin of 43.2%.

In FY24 COIN already did 39% in net margins with Visa at ~50% margins. The bull case is that COIN starts to look like a rail and not an exchange…hence a bull case with those kinds of margins.

Bull case:

That gives me a net income figure of $9.8B (bull case) which at 304.9M diluted share count gives us an EPS of $32.14 in 2031.

Base case:

A base case assumes $16.1B in revenue at ~33.5% net margins because of a lower stablecoin mix (more transaction mix) and general scale.

This kind of result for me seems unlikely given how I see the agentic world playing out…but it’s important to model out a case where I’m wrong.

This gives me a net income figure of $5.39B and an EPS of $18.

Multiple

S&S revenue today is ~41% and our bull case model puts us towards a 68.4% mix shift.

This makes S&S the dominant revenue generator for COIN and that’s deserving of a far higher multiple than if transaction revenue were the key driver.

Transaction revenue is a volume x take rate business. It spikes when retail trades and collapses when they don’t and the fee simply keeps getting competed down. The market therefore prices that part of the business like a broker:

• IBKR at 31x NTM PE

• SCHW at 14x NTM PE

• CME at 20x NTM PE

If we average those out, and consider that a base case has transaction revenue growing no faster than 10% per year, I think assuming a 16-18x multiple for this segment of the business is fair.

S&S revenue is deserving of a higher multiple because the cash flow generated here is more recurring, and structurally more durable than transaction revenue. Companies get valued higher for something that can generate increased cash flow with no material increase in incremental effort…and that’s where the stablecoin side of this business is heading.

V and MA for example trade in the 26x range and I think given COIN’s growth here and it being more crypto native, I think 26x-30x would be a fair range for this side of the business.

So to come up with an overall PE multiple, I simply take 68% (S&S revenue per my model) x 30x + 32% (transaction revenue per my model) x 16x which gives us an overall PE of 26x.

Bull case:

So 26x x $32.14 (EPS as above in bull case) → $836 which is a 322% increase of 27.1% per year.

Base case:

Let’s take a multiple more weighted towards transaction revenue vs S&S which would bring us closer to a 20x multiple.

A 20x multiple on $18 (EPS as above in base case) → $360 which is a 82% increase or 10.5% per year.

Disclaimer

Disclaimer: The Accuracy of Information and Investment Opinion

The content provided on this page by the publisher is not guaranteed to be accurate or comprehensive. All opinions and statements expressed herein are solely those of the author.

Publisher’s Role and Limitations

Make Money, Make Time serves as a publisher of financial information and does not function as an investment advisor. Personalized or tailored investment advice is not offered. The information presented on this website does not cater to individual recipient needs.

Not Investment Advice

Yields remain high yes, but the underlying economy remains strong too…and we remain in the middle of the greatest technological revolution we’ve ever been through.

Now is the the time to be aware of the macro situation, but remain with a bullish bias.

That’s exactly what I’ve done as I’ve continued to hold and add to the highest conviction names.

Those who have been focused on yields and focused on a hawkish Fed have been distracted away from the fact that:

• Hedge funds have been buying US tech equities at the fastest rate since mid-2025.

• Intel (INTC) CEO has said memory prices have risen 5x with a warning shortage will worsen in 2027.

• AI token usage is increasing at a rate of 31% month-over-month.

• Nebius (NBIS) has raised on-demand GPU prices ~20%

• GPU depreciation has been evidenced to be nowhere near what bears initially forecasted.

Today remains the time to be cautiously bullish.

Macro Update

1. It looks like QQQ and SMH are breaking out to multi-weekly highs. As I said last week, “there is no clear trend at the moment and until then it’s important to have a bullish bias, but be very picky with what you’re buying and when you’re buying it.” Today, it looks like QQQ and SMH are forming a new trend higher.

2. Institutional investors have now been net buyers of US tech for three straight weeks. The buying has been across hardware and software which is improving sentiment after a difficult period from July.

3. Crypto markets remain the best trade over the last two months…timed extremely well with our COIN buying.

4. Yields remain elevated which is why caution and position sizing is still very important…but it’s been promising to see markets hold and push higher even despite yields so far. It’s a sign that yields aren’t the overpowering narrative for the markets yet.

   

Commodities

Copper: The market is full of inflection points right now and copper is one of them if you look at this chart below created by Bernstein and FT.

Copper importantly remains a growth play, but it also remains a growth play even if the Fed hikes or not. Governments and hyperscalers continue to invest heavily (trillions in projected CapEx) into grid infrastructure and defence and there are no signs of this slowing down.

Copper now sits ~3% away from highs and as Stanley Druckenmiller says on copper:

“It’s the tightest position I have ever studied.”

Silver: Silver remains in its consolidation phase after its historic run in August 2025. As long as we hold above the $48 level I see no reason to be bearish silver as it acts a nice macro hedge, and an extremely underrated growth play for the AI infrastructure buildout just like copper.

I currently hold SLV, HL, and CTGO.

Scandium: I’m currently up 41% on my SREMF position that is 3.3% of my portfolio.

SREMF (Sunrise Energy Metals) is a pure play on scandium which is one of the most underrated metals in the world. Here’s why demand to supply is set to hit ~3:1 and why I remain happy holding SREMF:

• It makes up a key component in Bloom Energy’s solid oxide fuel cells (SOFCs).

• Aluminium-scandium alloys create the most lightweight, but strong metals in the world.

• It’s a key metal in 5G/6G technologies.

Crypto

Bitcoin and Ethereum have performed incredibly well of late, both breaking out of downtrends and jumping above May highs.

Whether or not the reason for the last 2 years was retail sentiment driven or capital flowing from a hyped retail market to ETF, institutions, and alike…I’m not sure.

But to me it seems like crypto, and crypto related equities are moving more in line with the long term agent thesis now that I’ve outlined many times. Moves aren’t as hyped and so far it seems like more of a sustainable trend.

It finally seems like AI and the agentic wave is catching up to the reason crypto was invested in the first place.

For me, the risk of being on the bearish side of this trade as agentic count grows at 30% CAGR per month is not worth it with the upside potential.

Micro Update

• South Korea’s chip exports hit a record high in September mainly led by SK Hynix (SKHY) and Samsung.

  • Total exports rose 14.9% in September (nearly 2x August’s growth rate). There are still no signs to me of any fading momentum.

• The release of Astra is just more reassurance of the bull case.

  • Rivals either have to close the gap to Astra or fall behind. That leads to more investment, better products, a larger market, and more discovery and knowledge. Token prices will eventually fall but the volume of usage will outweigh that so aggregate demand continues to grow.

• TSMC’s (TSM) chip production is set to fall short of demand in 2027.

  • JP Morgan has estimated that NVDA’s Vera Rubin will take up ~60% of TSMC’s output in 2027 (vs 15% in H1 2026). Supply is only getting tigher.

• Morgan Stanley has modelled that Bloom Energy (BE) will be ~20% of the entire fix for America’s data center power shortage.

• Nebius (NBIS) raised on-demand pricing for 4-year old H100s.

  • The revenue impact isn’t that significant but the signal is very bullish. Older GPUs are holding value longer than expected and pricing is keeping up with that fact.

  • Coreweave (CRWV) even announced on their earnings call that they signed an A100 (a 2020 GPU) contract extending through to 2029.

Big news for the week that needs discussing individually:

The CLARITY Act failed to pass last week (September 15th) which caused an initial sell off in crypto assets and crypto related equities like COIN, HOOD, and SECZ.

However, regulators have moved quickly and we’ve seen since:

• September 17th: SEC approved limited on-chain trading of tokenized stocks under a temporary exemption.

• September 17th: CFTC allowing developers to build passive derivatives software without registering as brokers.

• September 18th: CFTC filed new rulemaking for crypto transactions and markets.

Hence why Bitcoin is now up 17.5% in the last month despite a failed Clarity Act.

I’ll have more news in my “What’s Mispriced” section below regarding Coinbase (COIN) but this price action is further evidence that we’re now at an inflection point in the financial system.

Agent count is multiplying. Risk assets have been performing poorly aside from Bitcoin and Ethereum…this matters and it’s very important to understand why.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Coinbase | COIN (Low Risk)

Paid subscribers will know that I’ve been buying COIN aggressively lately. It’s now a +5% position that I’m up only 10% on but happy with an average cost here.

Technically, anything under this $212 level for me is a nice play to buy COIN if you have long term conviction in the play. A break above $212 will be some nice assurance that we’re heading higher over time.

I have my deep dive and valuation coming out on COIN on Wednesday afternoon. Make sure you’re subscribed for it:

Subscribe now

Read more

For those that haven’t seen the past few days OpenAI and Anthropic are aiming towards having 3rd party evaluators (from unknown organizations) to slow down the “Pace the Frontier” as Dario explained.

There’s countless takes I can make on this and countless takes I’ve read on X over the weekend.

But my aim here is to strip this all back and evaluate the investment implications with you as well as do our weekly macro and micro check in with my look at what’s mispriced.

TL:DR Macro Summary:

1. Bonds and oil markets remain the central to almost all media outlets narratives. In my opinion, bonds and oil aren’t anywhere near as important as what is actually happening with AI CapEx over the long run.

2. Comments from Dario and Sam Altman over the weekend will add another layer of distraction to what is actually happening. I think the next few days/weeks will see continued chop and likely more dips as enthusiasm around the AI trade amongst many will soften.

3. Being in a cash position above 10% and ideally higher I think is sensible right now. Thematically, there is no clear trend at the moment and until there is I think it’s important to have a bullish bias, but be very picky with what you’re buying and when you’re buying it.

4. With a long term mindset, the GDP forecasts, the compute we still need ahead, my bias remains firmly on the bullish side.

Before I do this let’s check in on the markets:

An update on my Web-App

Before we move on…just a quick note on my web-app which is ~3-4 weeks away from full deployment.

Some incredible features all paid subs will have access to:

• Full investing universe broken down by theme

• How I’d structure a concentrated ETF by theme

• A guided valuation workflow (or a view of all my personal models)

• My portfolio and watchlist and the ability to create your own portfolio and watchlist on the app

• All my daily notes

• And much more

Commodities

The below is a chart of SPY vs.. GLD, SLV, and COPX.

There’s not much to add here on top of what I said two weeks ago:

Silver (and gold…but mainly silver) is one of the safest ways to play the AI CapEx boom.

Investors are underestimating the demand behind silver as we head into the inference stage of the AI trade. If you’re bullish electricity, grid capacity, switchgear, turbines, solar, etc… then you need to be bullish on silver.

Silver is the backbone material running through all of this infrastructure buildout and the supply demand imbalance in my opinion will at some point give us another copper like move.

The safer way to play this move is through SLV. The higher risk way to play this move is by owning the miners. I personally own Hecla Mining (HL) and Contango Silver & Gold (CTGO).

On the other hand, silver and gold also acts like a hedge against US treasuries at current yields. History has told us high yields and sticky inflation is bullish hard assets (gold and silver).

Crypto

This chart below is important in backing up my thesis on Ethereum and it’s role in stabelcoins, tokenization, and everything agentic AI.

Over the last month alone, ETH (relative to market cap) absorbed 2.4x the inflows of BTC.

Advisors and institutions are positioning in ETH for the tokenization and stablecoin era that we are on the verge of.

A throwback to Brian Armstrong’s X post from March 9th 2026:

Other Key Macro

With 10Y and 30Y yields still high, the AI doomsday narrative continues to be focused on these levels relative to 2008.

As I’ve said countless times…this is all a distraction to what is actually happening.

Look at the rest of the market.

HYG (High Yield Corporate Bond ETF) is holding up fine. For those that don’t know what this is it’s the iShares iBoxx $ High Yield Corporate Bond ETF which tracks the US high yield (“junk”) corporate bonds.

It’s essentially a credit-market gauge and this chart says that risk appetite in credit is still there. Investors still believe that most issuers can keep paying.

Or more simply…”The market is still comfortable for now.”

Stop focusing on the yield narrative that you see everywhere and start focusing on what the underlying market is actually saying. For the most part, it’s saying the economy isn’t in as bad of a position as most of the narrative is suggesting.

From a more technical standpoint, the SPY is still holding above the 50D MA. I’m not convinced, given market spooks around the pace of the frontier that we will hold this but let’s see.

This is the unknown that I think is very key to track at the moment. My bias most definitely remains to the upside. I will not be selling much or trimming much (unless I feel I need to raise more capital which depends how much we do fall from here).

But I definitely remain on the side of the fence that is bullish micro will overpower a mixed/bearish macro.

Pacing the Frontier

I’m going to write this section as concisely as possible. I could write for ages on the different perspectives, and the different viewpoints in my head and those I’ve read but you can just go to X for that (for both my opinion and everyone else’s).

Here’s the investment implications from slowing down the pace of the frontier… assuming this actually happens (that’s a whole different debate).

Near Term: Near term I see narrative driven bearish sentiment. Enthusiasm around the AI trade will wrongly soften. Retail will likely panic and misunderstand that pacing the frontier (again IF it happens) means lower CapEx forecasts which is not the case.

That alone I think will cause a near term sell off which I will be sitting patiently on ready to buy dips as and when I deem necessary.

Mid-Long Term: Putting third party evaluators on models to slow down releases will not change demand, CapEx, or any cycles. In my opinion, this is purely a way to gain a regulatory moat, and a way to manage legal risk.

It’s strategic moves by Anthropic and OpenAI and not bearish at all for the overall future demand of AI.

In fact, this should create a smoother for longer AI cycle (as Gavin Baker correctly described it) which will decrease the risk of an intense boom-bust cycle.

Micro Update

• H100 prices are at the same level as they were three years ago.

  • This is more signs of the insatiable demand of AI. Compute is still so scarce that even NVDA Hoppers are not scrap because inference and agents can keep the old clusters busy..

• CrowdStrike (CRWD) CEO came out with a response that pacing the frontier does not secure production. The problem just moves from human speed security to machine speed security. I will have an article out mid-late this week on the cybersecurity industry but my main opinion is…demand for cyber is very high but be aware of the the very high multiples attached to this demand.

• Jensen Huang (NVDA) has said that AGI has arrived.

• Robinhood (HOOD) took a minority stake in Crypto.com.

  • HOOD also released that it generated $4M in Chain revenue on September 2nd which topped SOL, ETH, and TRON.

• Iren (IREN) CEO commented on the “Pace the Frontier” news with the following:

“Even if models never improved from here (which they will), just rolling out what they can already do would take more compute than the world can build for years.”

This isn’t an exhaustive list this week. But the very important thing to understand is that I do not think you should fade what is clearly happening.

Agent count is increasing parabolically. Research agents are now generating 3.1 work days for every 1 human work day. Thousands of agents can solve thousands of tasks, particularly tasks that have not originally been solved. Think:

• Drug discovery

• Materials discovery

• Mathematics and algorithms

• Etc

Every discovery opens up the world to a whole new TAM.

Granted the other side of the equation is terminal values of some companies will shorten which will compress multiples… but long term my opinion is that the world’s GDP is heading one way and that’s far higher.

Being bearish on the market now because of the current headlines is fading the 1st/2nd innings of the greatest technological revolution in history.

Short term, expect volatility and remain patient. Long term, keep your bias.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Read more

This piece is a look at the entire uranium and nuclear stack and the best way to play this theme.

Here’s a TL:DR overview of exactly what’s in this nuclear and SMR piece:

• Nuclear is back from 2011-2020 slump because decarbonization, energy security, and AI need dense, 24/7, carbon free power.

• Uranium prices will likely stay propped up and even rise to +$100 again as the utilities keep under-contracting uranium (as they have for the past 10-12 years). Add this to huge demand vs supply imbalances and I remain very bullish on the uranium market for the next decade.

• The best way to invest in the nuclear theme is by understanding the entire stack (metal → fuel cycle → technology).

• Some of the biggest bottlenecks are in the conversion and enrichment process of the fuel cycle.

The physics behind nuclear has always been there.

It’s just been a numbers game combined with political and public appetite, lower natural gas prices, and a history of Western projects that ran into huge cost overruns that made nuclear not worth it over the last 3 decades. Hence we saw nuclear go from generating ~17.5% of global electricity in 1996 to 9% today.

But now we’re in the middle of decarbonization goals, reliable domestic power needs, and AI’s power demand… the physics of nuclear actually starts to make sense.

Nothing else in the entire grid can pack in so much carbon-free power into so little land, and run that 24/7. Solar and wind can compete on speed and cost per installed kilowatt but with land use, capacity, and the need to be online 24/7… nuclear trumps everything.

Despite this, the numbers behind the installed base of nuclear isn’t there yet at all. The world has 400 GW of operable nuclear today but new Western builds remain very scarce relative to China. The “nuclear renaissance” as we call it seems visible in policy, cheques, and contracting…but isn’t close to being visible in actual completed plants.

But the part of the nuclear market that is showing forward demand perhaps before anything else is the price of uranium.

An update on my Web-App

Before we move on…just a quick note on my web-app which is ~3-4 weeks away from full deployment.

Some incredible features all paid subs will have access to:

• Full investing universe broken down by theme

• How I’d structure a concentrated ETF by theme

• A guided valuation workflow (or a view of all my personal models)

• My portfolio and watchlist and the ability to create your own portfolio and watchlist on the app

• All my daily notes

• And much more

Subscribe now

Uranium Prices

A core factor to understand if you’re bullish on nuclear stocks is why uranium prices are elevated and why they can keep going higher. We’re currently sitting at ~$89 with $136 being the high in June 2007.

Here’s why I forecast uranium prices to keep rising:

1. We’re burning far more uranium than mines produce.

   1. The buyers who keep reactors running have not locked in enough uranium for the decade ahead. Global reactor requirements sit near 69,000 tonnes of uranium per year. Current mine output is 60,000 tonnes per year.

2. Utilities have compounded the problem by under-contracting for years. Nuclear operators have at most 174 million pounds covered for 2026-2035 against ~360 million pounds of anticipated need. That’s 186 million pounds of no contract and no assigned supplier.

   1. This 186 million pounds isn’t any speculative demand. It’s fuel for reactors that already exist. There will be a time when that book has to be filled and buyers will hit the market together competing for a scarce supply. In that sense, under contracting is just concentrating a future wave of buyers.

3. Demand is massive.

   1. The World Nuclear Association sees reactor requirements (total volume of uranium that all operating reactors worldwide need to purchase each year as fuel) rising from 69,000 tonnes of uranium today to ~150,000 by 2040 as a base case (upper case is +200,000).

   2. On capacity, that’s 400 GW today to 746 GW in 2040 (base case) and 966 GW if governments get more aggressive on nuclear targets.

4. Kazakhstan still accounts for ~2/5ths of world production.

   1. Kazatomprom (historically supplied ~40% of all uranium) has approached peak output with production set to decline over the next 15-20 years.

   2. Not many new mines have been built recently and a project still takes a decade or more from discovery to delivery. Supply simply isn’t keeping up with demand.

The Opportunity

From an investment standpoint, the structural shift that matters is who’s buying nuclear, and how.

The original nuclear renaissance was focused on huge contracts with utilities and governments with 10+ year timeframes. That’s changing now.

Most of the new demand that is emerging doesn’t necessarily need 1,000 MW and most importantly it can’t wait 15+ years. An industrial site swapping out gas heat does not need a gigawatt reactor. It wants a few hundred MW it can add in as the plant grows. Same with a retiring coal plant for example. The demand for 1,000 MW+ units is much less now.

The demand is being fragmented into 50 to 300 MW blocks and this is exactly one of the core catalysts behind the next-gen nuclear and SMR’s.

It’s how nuclear gets sold to hyperscalers and industrials instead of only to state utilities.

Small Modular Reactor TAM

The International Energy Agency counts private SMR plans at ~25 GW with forecasts at 40 GW by 2050 as a base case. With policy support, funding, and good execution it sees these numbers closer to 120 GW.

Here’s where it gets slightly more interesting down the road:

Some reports suggest that if factories learn to develop SMR’s on a budget, and at scale, SMR’s have the potential to address 700 GW of industrial demand in North America and Europe. That would be in the region of $500 billion to $1.5 trillion in investment.

To put that into perspective, the world has about 400 GW of nuclear running today. So 700 GW of just North American and European industrial SMRs is almost 2x the entire global existing fleet. That’s roughly a 17x increase from the IEA’s base case assumptions in 2050.

Realize that is an upper-bound opportunity / an extremely bullish case rather than a forecast.

The Numbers:

The SMR market is currently valued ~$6-7B with the IEA seeing a market ~$25B by 2030.

This makes the valuations of pre-revenue companies (you’ll see below) pretty uninteresting at the moment given most are +$4B.

But here’s a thought experiment on where the market can go a 1-2 decades down the line:

At a mid-cycle build cost of ~$6,000-8,000/kW, and IEA current base case estimate of 40GW by 2050…the total construction bill for that fleet of SMR’s would be ~$240-320 billion. IEA’s more bullish take of 120 GW puts us ~$670B of cumulative SMR investment by 2050.

Note this excludes the extra bullish case of potential 700 GW.

The Timeline

The timeline for the uranium/nuclear trade is the biggest bearish case for me. It’s improving year on year…but nuclear is still a 3-5+ year story with SMR’s being a 5-7+ year story.

Aside from the technology advancements, and uranium supply of course, policy is the main thing that can compress the timeline.

On May 23, 2025, Executive Order 14300 told the NRC to change how it licenses and oversees reactors such as with stricter timelines in months, rather than years…and a change of process rules that tended to be hurdles for new builds.

Washington is backing this regulatory push as well. So far, the DOE has allocated $17.5B in loans to improve supply chain issues. We’ve also seen $2.7B in loans for domestic uranium enrichment (more on this below).

None of this makes nuclear a 2027 or 2028 story necessarily…but it does help alleviate some of the clear bear cases we’ve seen over the last decade…essentially that the US will talk about nuclear for decades without actually building anything material.

How to Invest

Let me take you through the nuclear roadmap quickly because understanding the roadmap is key.

A reactor developer sells a license, some engineering, and potentially some royalty on modules. The dollars are actually the uranium in the ground, the conversion, the enrichment, the heavy forging, the vessels, the constructions, and at the end the power contract.

In other words, the listed developer is the thinnest, and most binary part of a huge industry.

They’ll probably be deserving of the highest multiple just like Agility (CCXI) or Unitree in the robotics space will be deserving of the highest multiple, but whether that translates into the best (and safest) investment returns vs the actuator/bearings plays, or perceptions plays in robotics is the more important question.

It’s exactly the way to think about it in the SMR market too.

The way to think about the investment potential here is via a stack. Not through 1-2 pure play SMR tickers.

The Underlying Metal (Uranium)

Cameco | CCJ: CCJ is the blue chip of the group operating McArthur River and Cigar Lake. It sells significant portions of output on long-term contracts at fixed or base-escalated prices meaning it’s a de-risked way to play the more general uranium theme. You also pay up for the safety of this with a stock trading at 73x NTM PE.

Denison Mines | DNN: This is a pre-production developer making it far riskier than CCJ despite sharing similar geography. Construction began in March making it the first new larger scale Canadian uranium mine in decades. Their target production date is mid-2028 so it’s a slightly higher risk way to play the underlying layer of the theme but one that has a big catalyst coming in the next 2 years.

Uranium Energy Corp | UEC: I think of UEC as a similar story to CCJ just with slightly higher risk in that production hasn’t reached full capacity yet and they do not hedge uranium prices meaning investors have purer exposure to uranium spot prices.

NexGen Energy | NXE: NXE owns the Arrow Deposit that by a lot of metrics is the highest grade uranium deposit in the world. It has a grade of 2.37% U3O8 (most uranium mines operate in the 1% range). The forecast numbers in terms of mine life, capacity, and total production are impressive but first production is realistically not till 2031-2032 so it remains another higher risk way to play the uranium theme.

Energy Fuels | UUUU: UUUU operate out the White Mesa Mill in Utah which is the only conventional uranium mill operating in the US. The big story is that UUUU is pivoting into rare earths to produce dysprosium, terbium, samarium, europium and gadolinium. This makes it a uranium / rare earth hybrid play.

Fuel Cycle

This is where some of the most interesting companies sit…but most of them aren’t US public companies aside from the two mentioned below.

Firstly, understand that there is a key conversion from the underlying metal (mining) → conversion → enrichment → fabrication.

The conversion and enrichment stages of the nuclear fuel supply are the ones with the tightest bottlenecks.

Conversion: Raw uranium (U3O8) is what comes out of a uranium mine. After ore is crushed and chemically processed you have a dry powder that is roughly 80% uranium by weight. It’s useless as a fuel though and it’s in the wrong physical form for enrichment. It therefore gets converted to Uranium Hexafluoride (UF6) which is a gas that centrifuges can use for enrichment.

Enrichment: Natural uranium is ~0.7% U-235. Conventional reactors need 3-5% and advanced SMR’s need 5-20%. The industrial process used to achieve this is a process called gas centrifuge enrichment which spins UF6 at 50,000 rpm so the heavier U-238 is separated from U-235.

There are active mines (~40 globally), active nuclear reactors (+400 globally), but there are only 5 companies that make up ~95% of the entire conversion and enrichment markets. That’s one of the reasons why conversion prices have jumped ~400% and enrichment prices have jumped ~300% over the last 18-24 months.

The only Western conversion providers are:

• Cameco (CCJ) is a conversion provider.

• Orano (French) is a conversion provider.

• Solstice Advanced Materials (SOLS) is a conversion provider.

Enrichment is mainly state dominated by Rosatom (Russia), Urenco (UK), Orano (France), and CNNC (China). The remaining Western capacity falls to Centrus (LEU), Cameco (CCJ) and Silex Systems (SILXF).

Solstice Advanced Materials | SOLS: SOLS owns Metropolis Works in Illinois which is the only conversion plant in the US. CCJ is the other conversion play (I wrote about it above in the underlying metals section).

Centrus Energy | LEU: LEU is the only Western company currently producing enriched HALEU at commercial scale in the Western world. The total backlog now stands at ~$4.5 billion through to 2040 with a $900M DOE task order and new supply agreements with Oklo and X-Energy. Advanced SMR’s physically cannot be built without HALEU.

ASP Isotopes | ASPI: ASPI is a complex name in the enrichment layer. It uses Quantum Enrichment which is a laser based isotope separation technology. The issue and complexity with ASPI is that its main business is actually specialist isotope production. It makes ASPI a slightly more diversified holding company with a high-upside laser enrichment option.

Silex Systems | SILXF: This is a higher risk bet on the enrichment process getting disrupted. It has a SILEX laser enrichment process that exploits the quantum mechanical properties of U-235 molecules by firing a laser to get them to react differently and allow separation in far fewer stages than the centrifuge process. If this process materializes, the energy consumption per SWU will translate directly to lower fuel costs (enrichment is ~30% of total nuclear fuel costs). Silex are working with Cameco on commercializing this technology.

Lightbridge | LTBR: LTBR doesn’t sit cleanly in this layer but it’s closer here than others. It’s trying to replace cylindrical rods by using a spiral-twisted metallic alloy that conducts heat far better than conventional ceramics. This ultimately increases power output and increases margins. It’s still completely pre-commercial but in August 2026 they were selected for DOE’s Launch Pad programme which can provide an accelerated regulatory pathway.

Nuclear Engineering

BWX Technologies | BWXT: BWXT is an underrated play in my opinion. It has quite a defensible niche in the stack in that it is the only large commercial nuclear equipment manufacturer in North America. It manufactures nuclear reactor components for defense (US Navy and aircraft) and for SMR. It has a backlog of $7.3 billion.

GE Vernova | GEV: Another blue chip, and arguably the leading play in the SMR niche as well. It’s building BWRX-300 which is technologically the most advanced Western SMR project today. It also acts as a large, and very diversified energy infrastructure company so you get a defensible energy play with a built in SMR option for the upside. It won’t have the upside potential of some of the pure plays but GEV at 44x PE is historically on the cheaper side.

Fluor | FLR: FLR are the EPC contractor (the firm you hire to take a reactor design and actually build a working power plant behind it).

Pure SMR Plays

This is where the potentially outsized upside is more likely but it’s also where you have to respect the timeline the most. I have a very small position in SMR that I’m in no rush to currently build out. My only aim over the next few months/years is to bring my average cost down on SMR as much as I can and then wait. It’ll likely be a bumpy ride as a stock that won’t hold up in a market downturn, but with the size of the position I have that’s completely fine for me.

NuScale | SMR: Commercializing a light-water small modular reactor of ~77 MWe and it already has US NRC certification. TVA and the Romanian Doicesti project are the two large catalysts right now to watch.

X-Energy | XE: Developing Xe-100 which is ~80 MWe. It also has TRISO-X-Fuel meaning the business can generate revenue from reactors and from fuel. Only recently public and pre-construction.

Oklo | OKLO: Building Aurora Powerhouse which is a sodium-cooled reactor. Commercial operation is targeted for ~2030.

Nano Nuclear Energy | NNE: Developing KRONOS. It’s earlier stage but has been awarded a NRC construction permit which should improve regulatory approval time.

Mini-Deep Dives Into My Favorites Within the Stack

Read more

It continues to be a battle between the state of the macro (bearish high beta stocks & bullish commodities) and the state of the micro which continues to shine positively through all higher beta AI infrastructure names.

Commodities

Last week we saw some slight consolidation amongst the commodities and some slight bullish moves across the AI infra names.

iShares MSCI Brazil ETF (EWZ) is showing some signs of a breakout too which is a positive sign for the commodity markets.

Macro environment

It’s early days right now but my stance sits more on the more bullish side (through to the end of the year and most definitely if you look out to 2027/2028) mainly because yields are more of a distraction relative to everything else going on.

Historically the traditional playbook guides to rising long-term yields as a crisis signal. But aside from the headlines look at what is actually happening:

• Credit markets aren’t moving. Junk bond spreads (the premium risky companies pay to borrow) are near record tights meaning there isn’t much default risk right now.

• Inflation expectations, although slightly high right now, are sitting ~2.5%. Inflation expectations aren’t pricing in anything too worrying right now.

Investors clearly need lower rates at some point soon because the market can’t be fully driven by the earnings growth of AI companies. That is currently the reason why returns are subdued but I currently don’t think the macro environment is a strong enough reason for a major pullback.

It’s key to understand that although the yield chart isn’t that pretty…it’s just one part of the story that retail seem to understand most cleanly. The more downstream environment isn’t too bad right now.

Micro environment

On the micro we have:

• S&P PEG at 30-year low.

• KB Securities (largest investment bank in South Korea) said the memory market will see the “tightest supply conditions in history.”

• Nvidia (NVDA) forecast to grow +70% in FY28 even despite being supply constrained.

• Broadcom (AVGO) forecast 100% growth in both FY27 and FY28 (for AI semiconductor revenue) whilst being supply constrained.

• Broadcom (AVGO) CEO: “Demand for lasers, whether it is EML lasers, CW lasers, is far surpassing supply out there in the industry.”

• Credo Technology (CRDO) growing revenues by 115% with the optical transceiver market growing from 60 million units (2026) to 175 million units (2030).

The above are just new narratives from the last 7 days.

When you understand that the yield story is perhaps not as bad as headlines are suggesting (of course something we need to watch) then it really comes down to the fact that we’re entering a year where hyperscaler CapEx is set to be $1.3T.

This large number tends to create the bubble label but it shouldn’t. The size of the spend is not the test people need to focus on. It’s whether paid demand is still running ahead of what can actually be built and that answer remains an astounding yes.

That creates flows throughout the entire supply chain from optics, to memory, to cooling, to power that are quite hard to comprehend because the numbers are so different to what investors have been used to for decades.

The one part of this system that isn’t debatable therefore is that we’re in an undersupplied market. We’re being held up because of atoms (turbines, grid queues, HBM etc etc). It’s why players like Anthropic are still signing ~$517B worth of capacity deals in the last eleven months.

There is a time when an overbuild becomes a risk. It’s something to monitor but for the time being I don’t think we’re close. Being bearish on AI infrastructure is likely the wrong side of the trade if you have a 2+ year timeframe and are willing to look past headlines about rising yields (which may I make clear again are important to monitor…I just don’t think they’re more important than the above).

Crypto environment

And then we also have the crypto markets which are continuing their breakout.

For the last two years, BTC hasn’t done anything relative to the entire AI market around it. We had ETFs get launched, a more crypto friendly administration, and institutional custody mature…but the underlying prices did nothing.

This is slowly changing.

The reason is key. BTC is no longer a retail, leverage, and risk on trade.

The world is starting to realize quickly that AI agents need a financial system that:

• Settles at machine speed, 24/7, without a bank cut-off.

• Can make programmable payments (pay, stop, split, escrow) without a human in the loop.

• Hold and post on-chain collateral.

And the financial system we have to date (with limited banking hours, with quarterly reconciliations etc) simply isn’t fit for agentic purposes whilst Bitcoin, Ethereum, and the crypto ecosystem does.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Last week I touched on:

• Aeva Technologies | AEVA - Still remains a watchlist position.

• Quantinuum | QNT - Still remains a watchlist position.

• Nvidia | NVDA - Still remains a position I’m happy holding.

This week we’re looking at the below:

Coinbase | COIN (Low Risk)

The wider COIN thesis is in line with my narrative on crypto above.

In terms of valuation, here’s the brief valuation write up I wrote to paid subscribers on August 25th:

Read more

So what does all this mean and what is the market telling us?

Macro Take

Let’s start with the bond market.

Warsh came out hawkish at Jackson Hole. This is the kind of speech that, in theory, should reassure bond investors that the Fed is serious about the inflation fight. Core PCE is still at 3.3% (well above 2% target) and yet 10-Yr yields rose.

The reason is unclear but I think there are two ways to read the reaction.

Either:

1. Traders don’t believe the Fed actually has the stomach to hike enough times to bring inflation sustainably below 2%.

2. Or aggressive rate hikes slow growth and slow growth means lower tax receipts which means bigger deficits. Bigger deficits mean the Treasury has to issue more debt regardless of what the Fed is doing at the short end.

Point number 2 is the clear bear argument here. Interest expenses are escaping control. I think watching the 10-Yr is an important consistent check in over the coming weeks.

What this means for Gold & Silver?

Paid subs will know that I’m very bullish on silver. Part of that reason will be explained below (macro). Part of the reason is because I think the market massively underappreciates the silver demand needed in the AI buildout relative to silvers supply.

Warsh’s hawkish speech initially triggered a selloff in precious metals. That seemed like the obvious move (tighter policy/stronger dollar → sell gold/silver).

But longer term if the Fed hikes aggressively, one of two things will happen:

1. Growth slows, the economy softens, tax revenues fall, and deficits widen. That is bullish gold/silver.

2. Inflation stays sticky, confidence in Fed’s credibility weakens, and foreign investors question whether holding US Treasuries at current yields is worth the risk. That is bullish gold/silver.

That’s why the bond market is moving upwards. Investors are demanding a higher premium to hold US government debt and that has historically been a catalyst for gold and silver.

Hecla Mining (HL) is one of the silver positions I hold. It’s currently sitting nicely above the 200D, 100D, and 8 and 21 daily MA’s and above the key support we have at $16. Structurally, it looks like we have a nice floor in either at $16 or ~$12.50.

Contango Silver & Gold (CTGO) is the slightly higher risk silver play that I own. It’s predominantly a gold miner right now but it’s also a future silver story too through Kitsault Valley. Kitsault is still in exploration/drilling phase making this a gold miner with a silver option attached.

Micro Take

The whole market right now seems to be a battle between the more bearish macro takes, the oil markets, and the more bullish micro takes we’re seeing across the AI/semiconductor market.

Here’s a breakdown of some of the more important narratives we’ve seen over the last week from earnings calls, management commentary, and institutional research:

1. Nvidia (NVDA) had a huge Q2 growth estimate and FY28 forecasts were 70% vs original Street estimates of 44% (despite being capacity constrained still).

2. Nvidia (NVDA) projected $1.3T of hyperscaler spend in 2027. That’s vs the original forecasts of $800B and probably more important than anything else in the market right now. There are so many different angles to focus on right now in the market that it can get confusing if you let it get confusing. It’s incredibly important to zoom out sometimes and understand that upward revisions in total CapEx keep on rising. $1.3T in 2027 is a huge uptick vs previous numbers and making bearish arguments on the supply chain is therefore weak.

3. Chinese optics names like Innolight (300308) put out incredibly strong numbers with last 2 quarters revenue growth figures up 192% and 175% respectively with forecasts for H2 to be even higher.

4. SK Hynix (SKHY) CEO said memory shortage is expected to continue through to at least 2030. We also had Sandisk (SNDK) CEO state that “we see structurally massive demand for NAND until 2030.”

   1. Goldman Sachs forecast memory is set to be ~40% of the semiconductor dollar amount for humanoid robots.

5. Elon Musk said SpaceX (SPCX) could reach ~$3.5T in annual revenue by 2033.

The AI infrastructure has (a while ago) entered a phase that requires a different analytical framework to perhaps what most investors are applying.

The easy instinct is to look at capital spending numbers like the $1.3T figure, the GPU orders stretching years into the future, the memory shortage through to 2030, and then mention the world “bubble”.

The important metrics are:

1. Is earnings growth looking sustainable?

So far there is absolutely no solid data points pointing against this. Earning are continuing to beat across the board and trends are only heading one way.

2. Is supply outrunning profitable demand?

Yes. Anthropic is heading towards $120 billion in ARR whilst losing market share. Innolight is growing optics revenue +180%. These are not the result of price cuts.

The part of this that a lot of investors are still missing is what happens when intelligence inevitably gets cheaper. The intuitive bear argument is that open source models and efficiency improvements will eventually compress margins and reduce compute demand as cheaper alternatives enter the market.

The historical evidence from every prior tech revolutions suggests the opposite.

When the cost of a resource falls dramatically, consumption tends to rise faster than the price decline. It’s known as Jevons Paradox.

Cheaper intelligence doesn’t shrink the compute market. It expands the number of tasks worth computing in the first place.

Frontier models and cheap open source models aren’t competing for the same fixed pool of use cases - they’re together expanding what’s economically worth automating.

What’s Mispriced?

This is a section looking at the potential opportunities I’m seeing at the moment. I’ll split them by how I see the risk profile (high risk → low risk).

I’ll tend to aim for at least 3+ opportunities in every article I publish here…but that depends on what the market is offering.

You’ll also likely see overlap week to week. I am not a trader that changes his mind every few hours, days, or even weeks. If I see an opportunity, that opportunity sometimes stays around for weeks.

Aeva Technologies | AEVA (High Risk)

I wrote the following note to subscribers about AEVA on August 10th.

I since messaged very regularly on AEVA updating on no entry given the technicals showed no signs.

AEVA remains in that basket of a potentially very good opportunity here but until we see some consolidation on the chart or a reclaim of the $16.50 support level, I will hold off buying.

An opportunity nonetheless… and one to watch here at sub $16.

Subscribe now

Quantinuum | QNT (High Risk)

Read more

As Jensen Huang says:

“Biology has the opportunity to be engineering, not science. When something becomes engineering and not science, it becomes less sporadic and exponentially improving. It can compound on the benefit of previous years.”

That idea alone is set to create some of the most important AI companies in the entire world.

I hope that opening got your attention about my view of this theme. I’m going to outline in this article:

1. Why this theme is set to be revolutionary

2. The potential monetary inflow into this theme

3. How to invest in this theme

Why Digital Biology Is Set To Be Revolutionary

The fundamental bottlenecks in medicine today are:

1. It’s slow

2. It’s expensive

3. It’s highly uncertain in the lab

4. It’s even more highly uncertain in living systems

It’s why traditional biology and drug discovery has been constrained by the complexity of biology. Hypotheses, data analysis, and eventually failure rates of +80% after years and huge costs.

The digital biology approach inverts this.

Specialized AI models trained on molecular, preclinical, clinical, and patient data can analyze billions of hypotheses instantaneously in parallel.

And then think about the next layer…agentic systems can then read literature, formulate questions, design experiments, interface with robotic labs, analyze results, and iterate this Design-Make-Test-Analyze loop over and over again. Productive wet lab teams pre-AI may be able to fully explore ~1,000-2,000 molecular hypotheses per year but an AI native setup can explore an order of magnitude more whilst feeding every outcome (success or failure) back into the model.

The nature of biological data is itself is arguably the most valuable data in the world. Previously failed testing programs were just treated as a sunk cost but with AI those failed outcomes turn into a goldmine. The models are able to internalize why something failed under specific conditions which offers the edge in avoiding that same mistake again.

The end patient and the lab still remain the ultimate end judge (and will be the new bottleneck), but the time, cost, and attrition of reaching this stage is set to compress massively.

I think the most important part of this entire feedback loop to understand from an economics point of view is that once the percentage of successful trials starts to rise, the compounding loop from there becomes quite aggressive.

And the faster these trials become successful, the faster this loop starts snowballing.

Trials will become more successful because of better data analysis and speed of iteration but also because aging and disease will change from a metaphysical concept to a set of defined, measurable, targetable engineering problems.

It essentially becomes a computational problem rather than a science and lottery problem.

The loop reinforces itself in the sense that better models yield better medicines which yield more capital and knowledge.

And eventually in a world where the algorithms become a commodity, the scarce assets (the proprietary biological data mainly) become extremely valuable.

I want to make it clear therefore that digital biology is set to do far more than accelerate existing pharmaceutical research and development.

It’s set to completely redefine the model for creating medicine. That market today is worth ~$1.8 trillion today. I suspect the end market will be worth multiples higher than that quite soon.

Just a quick note:

I’ve been working very hard on a web-app as you all know so far. We’re currently in beta testing and have a development team working on the final stages for this version 3 set to be released soon.

It’ll add a whole new dimension to my service over the next couple years. You can access the site here to have a look.

Here’s what it’s in the app:

• Guided Valuation Model (EV/Sales, EV/EBITDA, P/E, or MMMT Valuation Model) for every stock in my universe - see screenshot above.

  • This allows you to input your assumptions (or follow my assumptions) and come out with a target price.

• Stock Baskets: As well as showing my personal portfolio as evidence of my track record, I’ll be showing you concentrated stock baskets of the best themes in the markets.

  • These aren’t ETFs. These are 5-10 stock baskets for all themes in the market with individual stocks and weightings.

  • I think of it as a way of introducing you to other high conviction plays I have that I don’t necessarily personally own due to capital. I also think it’s a great way to increase your exposure to a theme that you may like more than I do.

• This is all on top of daily notes, live portfolio, technical screens, and research notes.

A Thought Experiment On Just How Big This Market Can Get

This is hard to model but I’m going to do my best to make the assumptions as clear as I can.

Note: These are a thought experiment. Do not take these are market forecasts.

Short Term

The first way I think about this is looking at the current biotech market today and the effects of a rotation out of AI hardware and into biotech changes the market size today.

The top 15-20 AI hardware / semi names have a combined market cap in the $13-15 trillion range right now.

The biotech market is currently valued at ~$1.25 trillion.

Now the next step is where assumptions come in. I think it’s highly likely one of these two happen:

1. Investors and institutions see potential slowing of the AI infrastructure play and rotate capital elsewhere…of which some goes to healthcare and biotech. If that’s the case I think a fair assumption is ~2.5% rotation from AI infrastructure directly into biotech (I think this is already happening in the early stages).

2. Investors and institutions start to see biotech as one of the purest AI plays out there over the next decade that may create some of the most important companies worldwide (just like I stated in my opening). If that is the case I think we’re looking more towards a +6% rotation out of AI infrastructure and into biotech.

At 6% rotation out of a $14 trillion market means we’re looking at $840 billion in capital.

At 5x inelasticity (fair for the biotech market) that leads to ~$3.5 trillion of amplified buying power.

If ~40% of that gets concentrated into the top 10 names like LLY, TMO etc there is a $1.4T incremental market cap impact on that group of multi-billion dollar companies.

And then on the mid cap to smaller cap names we’d likely have ~$2.1 trillion impact but the current market cap of this range of stocks is no more than $200 billion so the multiple uplift could be in the range of 8-10x.

That’s purely just a thought experiment to evidence how large potential capital flows may impact a smaller, more inelastic market in biotech today.

When you combine these shorter term capital rotation forecasts with the below assumptions, the numbers get quite staggering.

Longer Term

Forecasts today have the biotech market generating ~$1.7T in annual sales with a market sitting ~$1.25T and 2030 forecasts at $2.2T.

Those assumptions are based on the current environment of ~8-9% of drugs entering Phase 1 actually making it through to approval. It’s also based on ~$2.6 billion in spend per successful medicine, and ~10-15 year total timeline. Most of that capital spend is unfortunately taken up by failures rather than successes.

If the thesis plays out like I outlined in the first section and we see the industrial feedback loop work smoothly by spotting early failures, and feeding back data into better models…then the same R&D spend becomes far more productive.

Conservative Assumptions

A conservative take would be to say Phase 1 to approval success goes from 8-9% to ~12% and then gradually towards 15% and time compresses to ~5-8 years. That alone would increase new annual launches by ~75-100% and cut the cost per drug down to sub $2B.

If the same R&D budget (which actually could be increased substantially if clinical outcomes become less of a lottery) produces ~2x the amount of approved drugs per annum, the drugs reach cash production several years earlier and fewer dollars are burnt in Phase 2 / Phase 3 trials and also fewer dollars are held up in projects lasting many years.

Over time my numbers point towards a 2x in earnings and free cash flow which can re-value the market from ~$1.3T to $2.6T, excluding multiple expansions.

Multiple expansions is the easier argument to make. A sector that runs like a compounding production system on some of the most valuable data in the entire market with rapidly improving ROIC will be deserving of a multiple at least 1.5x higher than today.

The narrative changes from a high risk lottery bet to a engineering bet.

A 1.5x increase in the multiple then takes us to ~$4T.

A reminder that’s a base case.

A reminder also that this is excluding what I said above with capital inflows from other areas of the market.

Bullish Assumptions

A more bullish case (which I suspect will happen but just further down the road) assumes the loop industrializes.

Phase 1 approval rate jumps to ~18-20% with compressed timelines ~4-5 years in bio-marker rich niches of biotech. Because of this cost per approved drug reduces to ~$1B and annual launches can therefore head towards ~2.5-3x the rate today.

At this point, the main bottleneck will become the end judges (the humans) and the scale and speed at which we can run trials and manufacture drugs at scale. Discovery becomes faster but unless clinical execution and manufacturing can ramp up there is still a clear constraint.

Current forecasts suggest $2.2T in annual sales by 2030. A 2.5x in approvals in an ideal world means a 2.5x in sales but that won’t be the case as the more drugs approved likely means a lower % of blockbuster drugs. What is likely does mean though is that there will be a faster flow of well branded assets that reach patients earlier, with more years of protected / patented cash flow ahead of them.

That means the net present value rises before we even consider an increase in sales.

So valuation impact then comes from:

1. More approved assets

2. Cheaper produced assets

3. Longer NPV per assets

If earnings therefore increase ~3x based on above assumptions and the multiple increases ~2x we’re looking all of a sudden at a $7.8T market cap.

Most of this value won’t be spread evenly among names. But I think it’s becoming clearer and clearer those 20-30 names that either are:

1. Pure plays on drug discovery

2. Pure plays on specialized compute

3. Pure plays on manufacturing

4. Pure plays on the platforms

5. Pure plays on commercial reach

I’m going to lay out the investment picture here now. I’ll dive a little deeper into my 4 favorite names and list ~20 more that I like today.

How to Invest

If I were to create an ETF on this theme, here’s what it would look like (click on the link to download the spreadsheet).

Read more

I’m going to highlight in this article how I see the stock moving from a $60B MC at $221 today to much higher whilst also showing you a range of situations that may happen and the associated stock prices so you can best gauge your estimates.

The goal though isn’t to give you a single price target. It’s to give you my personal view… and also a range of likely situations in 2030.

As with all modelling, a lot of this work requires assumptions. Sometimes I may be too aggressive and sometimes I may be too conservative. But based on current trends, management guidance, and industry wide forecasts I think I’ve got pretty close to what will likely happen.

Subscribe now

The facts we can be slightly more sure on:

Contracted Capacity

Management has guided us towards 5 GW of capacity to be contracted by year end 2026, with more than 1 GW of additional capacity planned for deployment each year beginning in FY27.

“We continue to build our future capacity pipeline through our own and colocated sites, and today raise our year-end contracted power target to 5 gigawatts. Our future capacity pipeline effectively makes Nebius one of just a few companies in the world able to build more than 1 gigawatt of new capacity a year. And we plan to do so in 2027.”

Connected Capacity

This path therefore produces ~5.2 GW of connected capacity by the end of FY30 if execution remains in line with management forecasts.

“Almost all of that power will come online over the next 3, maybe 3.5 years out of the contracted power.”

A lot of the numbers used in this work are based on this 5.2 GW of connected capacity so here’s a bit more depth on it.

So starting from YE 26, that points to the bulk of today’s 2026 forecasted 5 GW being connected in mid-2029 to 2030 starting from today’s base rate ~800 MW-1 GW.

5.2 GW is therefore not the most aggressive interpretation of the current commentary.

The risk to this forecast is grid interconnection which is structurally more of a nation wide risk. The grid simply can’t meet new demand quickly which is why we’re being forced into the behind-the-meter power solutions and power generation systems that do not rely on the grid.

Right now, we’re seeing grid-interconnection queues above 2,000 GW with an average wait time of ~5 years (application to operation). Only ~13% of projects that entered application before 2019 actually reached operation.

The more positive take for NBIS is that site selection and strategy is built around the interconnection issue.

75% of contracted capacity is sourced from owned, BTM generation (Bloom Energy fuel cells and Vineland), a natural gas power station in Missouri, and a site in Pennsylvania. None of these are stuck in the 2,000 GW queue.

Site selection and strategy is therefore the moat.

Annual Contract Value

More recent landmark contracts are valued at ~$20-25 million of ACV (annual contract value) per MW. Worth noting that original levels were ~$12 million.

My personal base case that reflects:

• Customer mix

• Ramp timing

• Difference between contracted and fully revenue generating capacity

Gives me a deliberate haircut (I always model conservatively) of a blended ACV per MW of $14M.

5.2 GW * $14M = $72.8B ARR by 2030 (this is my base case we will work from)

That’s roughly a 9.1x increase in ARR from YE26 to YE30 which is entirely capacity driven with a conservative ACV.

I’ll get into the valuation section after this but on the face of it a 9.1x increase in ARR in 4 years whilst trading at 8.8x NTM sales isn’t expensive.

ARR Forecast: Bullish Case

I think my assumptions above are a solid base case.

I do want to include a more bullish take so here it is. It’s quite in line with a model that Wells Fargo put out where we reach 7.5 GW of active infrastructure by 2030.

From here, Wells Fargo also used a more positive revenue per MW than my assumptions.

“These deals were for an average of more than $1 billion each. They represent a yield of $20 million to $25 million per megawatt.”

“You should expect the deals we closed in this quarter with more than $20 million per megawatt deals and less than 2-year payback period.”

Industry benchmarks for a fully utilized GPU data center are closer to $20-22M per MW which could push our 2030 ARR up to +$150 billion.

I understand this is a bull case but this is the far more likely scenario vs the below bearish take.

ARR Forecast: Bearish Case

Wolfe Research have put out numbers suggesting we see $41.2B in ARR by 2030 based on 5 GW (potentially not all fully online or fully utilized on day one of 2030 though).

An exact $ per MW is not clear but based on 5GW of active capacity we’re likely looking in the region of $8.24M which seems incredibly low to me. It’s based on:

• Normal utilization rates, ramp timing lags, and some capacity that is reserved or not yet at peak density.

• Mix of older, lower-priced contracts still in the base.

Seems like a heavy haircut to me, but one we should consider and one I’ll work from for a potential share price in each section as well.

Methodology 1: EBITDA

Note: This is all based on a $72.8B ARR.

Full disclosure here: I’m aware a pure EBITDA look may not be the best representation of valuing a capital intensive business like NBIS but when FCF is likely negative or very slim through 2030…you can’t run a clean looking DCF model.

EBIT / ROIC is a nice way of looking at the economics but it’s not a common valuation method to compare to peers today. In my next method, I do look at ROIC.

EBITDA is the purest metric we can use right now. It’s just important we take into account and understand the downsides of using this and that these share prices are likely a little inflated.

Group margin currently sits ~41% with management indicating we should see margins stabilize ~45% once we reach scale.

I’ll be using 45% as my base case longer term assumption but will model between 35% and 45% through to 2030.

$72.8B * 45% → $32.8B

$72.8B * 40% → $29.1B

$72.8B * 35% → $25.4B

An EBITDA of $32.8B which results in a 124% CAGR over a 4 year period.

124% CAGR (4Y) is arguably one of the highest growth rates we’ll see amongst all stocks I follow and deserving of a high growth associated multiple.

• EQIX trades at 23x NTM EBITDA based on 13% CAGR

• DLR trades at 23x NTM EBITDA based on 13% CAGR

• CRWV trades at 8x NTM EBITDA based on 71% CAGR (CRWV has far higher net debt and is lease heavy vs NBIS owned sites).

Here’s a bull to bear case:

Bull: 25x * 32.8B → $820B EV

Base: 15x * 32.8B → $492B EV

Bear: 10x * $32.8B → $328B EV

Net Debt

Assumptions for net debt are as follows:

Net debt is currently ~$2B. Forecasts and maths based on ~$30-38M per MW assume gross invested capital will be ~$156B - $198B.

Here’s what we know on the further details here:

• Customer prepayments are expected to cover ~50-60% of associated CapEx on large deals.

• Convertible notes (~$15B) will fill residual gaps.

• Operating cash flow should become a growing internal funding source.

So, if ~50% of CapEx is covered by prepayments + growing OCF, external capital needed will be in the $70-90B range.

Before I go into the numbers, I just want to be clear that NBIS’s strategy so far has been stellar in terms of their cost of capital. They’re already starting to look like a hyperscaler.

Most neoclouds still fund growth primarily through equity which is pricey and creates this repeated cycle of dilution → deployment → depreciation.

NBIS is already at a stage where they’re accessing asset-backed debt which is collateralized by investment-grade customer contracts. The recent $775M debt facility is collateralized by contracted cash flows from investment grade customers which means lenders are effectively underwriting the credit of counterparties such as META and MSFT rather than NBIS itself.

NBIS management indicated that this can essentially be repeated across the entire $40B of backlog.

Then as discussed we have prepayments directly funding new capacity as well which should mean this dilution figure isn’t as bad as we see tend to see in the neocloud industry.

I think a conservative range is therefore to assume net debt in 2030 is between $40B (bull case) to $80B (bear case).

So if we take the higher end of this range off our EV’s we get the following equity values:

$820B - $80B → $740B

$492B - $80B → $412B

$328B - $80B → $248B

These are equity values far above today’s $60B range.

Dilution

The next stage of assumptions is dilution. I’ll model my dilution out here and then re-use in the below methods too.

It’s also really important to note that the economics of this business sway between being extremely good and not great on quite slim changes.

I.e. if NBIS do manage to work towards +$100B in ARR then operating cash flow starts to cover far more of the CapEx than I am assuming here which in turn means less debt and less dilution.

This is entirely a story based on how quickly NBIS can convert to connected capacity, and from there how quickly they can get that blended $ per MW up close to +$20M per MW and beyond.

Modelling share count is obviously a challenge but today’s starting point is 274 million. We have:

• $5B convertible note that will:

  • 2030 notes convert at $313 → 7.98M shares

  • 2034 notes convert at $325 → 7.7M shares

• Pre-funded warrants at $2B at implied stock price issuance adding ~8.9M shares.

• Ongoing SBC which could add ~5M shares (conservative take).

That’s an additional 22M shares of known dilution added. I’m going to apply another 55M share dilution to account for the additional equity capital raises I expect will be required for the 4.2 GW buildout from here assuming average issuance prices at above today’s levels.

This gives me 351M shares as a base case to work from for the 2030 outstanding number of shares. 351M shares in 2030 is a 6.3% dilution per annum from here which again I think is on the more conservative side of where we actually end up.

Share Price

Therefore we have:

$740B / 351M → $2,108

$412B / 351M → $1,174

$248B / 351M → $707

EBITDA methods will likely give NBIS an inflated share price so I think taking these with a pinch of salt is the best bet.

More bullish case (based on connected capacity and $ per MW)

Assumptions:

• ARR: $150B (see above)

• EBITDA margin: 45% (see above)

• Share count: 351M shares (see above)

($150B * 45%) * 25x / 351M → $4,808

($150B * 40%) * 15x / 351M → $2,564

($150B * 35%) * 10x / 351M → $1,496

More bearish case (based on connected capacity and $ per MW)

Assumptions:

• ARR: $41.2B (see above)

• EBITDA margin: 45% (see above)

• Share count: 351M shares (see above)

($41.2B * 45%) * 25x / 351M → $1,321

($41.2B * 40%) * 15x / 351M → $704

($41.2B * 35%) * 10x / 351M → $411

Note: The business is valued a lot like this today on a pure 17x NTM EBITDA multiple for the most part. That won’t be the case forever.

There will likely be a case soon where margins, where FCF conversion, where CapEx comes under a lot more scrutiny and that will come if/when we get more of an prolonged drawdown.

It’s why I don’t love this above approach when looking out for 2030 onwards.

Methodology 2: EBIT

A longer term DCF approach is ideally the best way to value NBIS at today’s inflection stage because of the multi-year capacity buildout ahead. The issue is based on current forecasts we’re likely looking FCF negative until ~2030-2032.

The main goal with this approach is to quantify the capital required to get to this capacity peak with a realistic funding mix and depreciation.

Here’s a clean walkthrough for you (based on the same assumptions of 5.2 GW and a blended $14M / MW:

Read more

AAOI has become a popular stock amongst the retail community, and for good reason.

There’s not a single quality stock I know about in the market that has the potential revenue ramp up that is possible with AAOI over the next 18-36 months.

It reminds me of NBIS back when it traded at ~$75. It was trading at 2.4x ARR when ARR was projected to 8x over the following 12 months. NBIS then went from $75 →$298 before pulling back slightly now to $186.

AAOI is at a very similar revenue inflection point now…just on an even more bullish scale.

My personal AAOI position:

I’ve now held AAOI for a while with an average cost at $133. It used to be in the $150s but I took advantage of the recent dip we saw down to ~$80 to slowly lower my average cost basis.

It’s now my fifth largest position behind LMND, OUST, NU, and AMZN - you can see the entire portfolio (and much more) as a paid subscriber.

Subscribe now

I’m inclined to make AAOI an even larger part of the portfolio but I’m just being a bit cautious as we approach new highs on the indexes again.

Contents

• AAOI Background

• Optics/AAOI TAM

• AAOI Earnings Report

• AAOI Valuation

Company: Applied Optoelectronics

Ticker: AAOI

Headquarters: Sugar Land, Texas

CEO: Dr. Chih Hsiang (Thompson) Lin

Market Cap: $10.72B

NTM Fwd Sales Multiple: 5.44x

AAOI Background

This will be a very brief AAOI overview.

AAOI fabricates its own Indium Phosphide laser chips in-house and also manufactures the optical transceiver modules those lasers power. Because AAOI grows its own laser material (InP), it controls the entire supply chain whereas almost every peer is constrained.

Lasers are, in the words of the CEO: “The biggest bottleneck right now for the transceiver business.”

AAOI currently has two segments:

1. Data centre transceiver business (100% of my investment thesis)

2. Cable television (CATV) amplifier business - a smaller, profitable business but one I have little interest in looking out for the next 3 years.

The company is in the middle of the most aggressive manufacturing expansion and volume ramp up in its history (and arguably one of the most aggressive expansions in the market) looking to 25x their manufacturing space. The product roadmap is currently 100G and 400G, with 800G as the near term inflection and 1.6T as the next standard as well.

The wave after that is co-packaged optics (CPO) where the laser source moves from a pluggable module directly onto the switch package. AAOI’s ELSFP laser module (the external light source that CPO needs) is targeting ~400,000 units per month by 2028.

This ELSFP would be an earlier story but as per the CEO:

“We just can’t make enough of them to be involved in their current first-generation deployments because there’s just not enough capacity. We have to prioritize our ability to make lasers for our own transceivers first.”

I’ll try to make it abundantly clear throughout this article, but I think you can gauge from what I’ve said so far… AAOI is a company not short on demand.

Customer demand even exceeds the planned capacity targets by another 20%.

This makes AAOI currently purely an execution bet - an execution bet that the manufacturing ramp up in Houston delivers on schedule. If it does, and forecasts around gross margins, ELSFP volumes, and 800G/1.6T volumes are accurate…then AAOI is, in my opinion, one of the better bets in the entire market today at sub $11B MC.

Before we move on, a quick word about TrendSpider - the charting platform that I use.

TrendSpider is running a weekend Sidekick AI Sale. You can save up to 45% on TrendSpider and receive a free one-month Sidekick upgrade.

If you manage your own portfolio, here are a few ways TrendSpider can fit into your research process:

- Group Strategy Tester to run one or multiple strategies across an entire watchlist, portfolio, sector, or index, then compare the results in one sortable table

- Claude Opus 5 in Sidekick for advanced market research, chart analysis, backtest interpretation, portfolio analysis, and custom indicator workflows

- New Segments & KPIs data covering business and geographic segments, profitability, capital expenditures, and company-specific KPIs for approximately 1,500 major U.S. companies

If this sounds useful for your process, check out TrendSpider's Sidekick AI Sale and save up to 45%.

Disclosure: I may receive a commission if you sign up or make a purchase through links in this post.

The TAM

The optics market in general is massive and we’re in the very early innings.

Let’s start big picture.

Goldman have put a forecast out for the total addressable optical networking market to be worth ~$154 billion which is 9x from today. Whether this holds true or not, it’s likely a directionally correct signal for what connectivity infrastructure is worth in a world going through the largest tech revolution of all time.

TrendForce have put out numbers too on the combined value of the CPO and NPO market to expand from ~$100 million in 2025 to $39 billion by 2030 with the majority of that growth being in 2028 and 2029. They also put out forecasts for the pluggable transceiver market to generate ~$26 billion in revenue by 2030. This means their estimates total to ~$65 billion (40% of Goldman’s and also two years further into the future).

For the sake of being conservative, most of the analysis below will work from TrendForce.

Now AAOI does not compete across the full $65 billion / $154 billion market. They compete in pluggable transceivers today (mainly 2027-2028) and are building into CPO laser supply for 2028 onwards.

On the transceiver side, AAOI is guiding towards 930,000 units per month of 800G and 1.6T by the end of 2027. With that equating to ~$471M/month ($5.6B annualized) that’s ~20% of the $26 billion forecasted pluggable transceiver market by 2030.

On the CPO side, AAOI is targeting 400,000 ELSFP laser units per month by 2028. Estimated ASPs are ~$500 per module which equates to ~$200M per month from CPO lasers alone ($2.4B annualized) against a $39 billion forecasted CPO market.

This gives AAOI ~6% of the potential CPO market which gives them some room for growth.

Put this into perspective though in ~2030. At a current market cap ~$11B, a business generating $6-8 billion in annual revenue across transceivers and CPO laser supply (against a total TAM of between $65 and $154 billion) is trading ~1.4-1.8x that potential revenue run rate.

Q2 2026 Report

I will dive into the numbers in this report because that’s what we do in earnings reports, but I think it’s important to caveat with the fact that the numbers in H1 2026 are not central, nor that important, to my investment thesis in any way.

The reason for this is because AAOI is a H1/H2 2027 onwards play in terms of volume ramp up.

That’s when we should start to see the $471M / month numbers which is a 12x ramp up from FY25 revenue.

Most of what I care about today is:

1. Manufacturing: The story now is predominantly about manufacturing execution. Demand is through the roof and the cap for AAOI right now is whether or not they can hit that 930,000 modules per month goal by the end of 2027.

2. Updates of 800G ramp up: Management have made it abundantly clear their timelines for 800G/1.6T ramp up. It was never going to be a Q2 story. It was always Q3 and Q4 and then ramping up nicely throughout 2027 so numbers in Q2 were never going to be backing up this thesis.

Anyway, here’s a look at Q2 numbers:

Demand:

• Q2 was the fifth consecutive record revenue quarter.

• Revenue grew 86% YoY and 27% QoQ.

• 56% of revenue came from data centre products and 42% from CATV.

  • Most of the data centre revenue came from 400G and 100G with 11.9% of data centre revenue coming from 800G (still immaterial but this should start to change in Q3).

Profitability:

• EPS beat estimates and management guidance.

• Returned to non-GAAP profitability for the first time in years with non-GAAP net income of $5.5M vs a guidance of -$2.5M - +$2.8M.

Guidance:

• $1.1B guided for FY26 meaning we’re looking at ~$485M guided for in Q4 (2x what is guided for in Q3). This is where execution risk becomes quite material for AAOI. Management wouldn’t be guiding here if this wasn’t feasible but Q3/Q4 will be almost a make or break for investors in terms of their trust in AAOI management. AAOI management do have a slight reputation over overpromising and underdelivering in the past. I sense this time is different, but it’d be naive not to be aware of this as a risk.

• Long-term gross margin target remains at ~40% with ELSFP laser contribution being ~50-65% gross margins.

Important notes from earnings call:

“We continue to expect by the end of this year that we will be capable of producing over 650,000 pieces of 800G and 1.6 terabit products per month. By the end of next year, 2027, we continue to expect to grow our production capacity to be able to produce over 930,000 pieces of 800G and 1.6 terabit products per month, with over half of that output coming from Texas.”

As I touched on above, the most important part for me for Q2 earnings was the manufacturing capacity trajectory.

One year ago, AAOI had 65,000 square feet of manufacturing space. Today they have 1.6 million square feet in the Greater Houston area (that’s 25x in footprint).

Here’s the ramp up projections management have given us:

Q2 2026: 200,000 monthly capacity for 800G/1.6T units

End of 2026: 650,000 monthly capacity for 800G/1.6T units

End of 2027: 930,000 monthly capacity for 800G/1.6T units

An 800G transceiver at current ASPs generates ~$400-500 per unit. A 1.6T transceiver is much higher.

So 930,000 units per month by the end of FY27 (predominantly 800G and 1.6T though unsure on the split) is where we start to see this potential for $5-7 billion in revenue.

It’s where management’s $471M/month figure comes from.

“We also discussed our plans to increase our manufacturing capacity for our external light source or ELSFP. That's for co-packaged optics or CPO". we anticipate ramping production later this year and into 2027, ultimately culminating in about 400,000 pieces per month in 2028.”

Modelling this into a revenue figure takes guesswork. Above, I estimated the ASPs to be ~$500 a unit based on:

• ELSFP lasers being 6x the die size of a transceiver laser.

• Requiring 300mW output vs 70-100mW for 800G/1.6T.

• DWDM variants are much more complex.

To me this implies material premium over a standard transceiver component which has an ASP ~$400-500 per unit. I think $500 is therefore a base/conservative number to work from.

$500 ASP at 400,000 units per month is $200M per month (or $2.4B annualized) on top of the potential $5.6B we have annualized in FY28 for the transceivers.

What’s more, the ELSFP business carries materially higher margins than the transceiver business meaning on approximately half the revenue, we’ll likely see a similar gross profit figure.

“For laser, the gross margins is about 55% to 65%. For ELSFP, the gross margin should be more than 50%.”

“Our ability to deliver revenue in general, and specifically when it comes to 800G products, is limited by our production capacity right now…if we could produce more, we could ship more right now.”

Just a final reminder that AAOI is a supply constrained company and not a demand constrained company.

Valuation

The stock is expensive if you look at today’s multiples relative to today’s revenue (~17x sales).

But the valuation compresses to sub 2x NTM sales on the mid-2027 run-rate management is guiding towards with a 12x ramp in revenue over 24 months from FY25.

Read more

Part 1 and 2 have already been released:

Part 3 focuses on:

• Uranium

• Brazil

• Application Layer

Uranium

The MMMT Uranium index is currently down 19% over the last year.

That makes it a slightly worse performing basket to hold than the likes of Global X Uranium ETF (URA) which is down 6% over the last year at the time of writing it.

As I try and make clear, the point of these baskets is to find a slightly higher beta way of holding the underlying theme rather than just buying the standard ETFs.

In Part 1 and Part 2, my baskets far outperformed the respective ETF’s so it’s no surprise that we have at least one or two baskets underperforming, especially in sectors that have been struggling over the last year.

Also, it’s worth nothing that my uranium basket is much higher beta than other ETFs because I have included SMR (a small modular reactor play) as a 15% holding.

This is also a play I own in my personal portfolio that I am down on a lot, but currently sized reasonably as to not materially affect the portfolio at all. I still have a lot of conviction in the SMR thesis…the time frame to start producing revenue is likely just a long time away which makes it a far riskier hold during market downturns like today.

Maybe the timing on this thesis has been slightly wrong, but the thesis I believe in.

Uranium Thesis

The supply-demand thesis for uranium is as clear as any commodity thesis I know.

Annual reactor demand globally runs ~200 million pounds of uranium.

Primary mine supply is 130 million pounds.

That’s a structural deficit ~50 million pounds per year that is currently getting filled from finite buffers like secondary supplies, enrichment underfeeding, and inventory drawdowns.

But here’s where it gets more clear cut:

Only ~65% of the uranium that reactors need annually is currently covered by long-term contracts. Utilities have been deliberately undercontracting for years…running down existing supply rather than committing to new supply at higher prices.

That ultimately worked when inventory was comfortable but it no longer works now with supply at record low and time between signing a contract and receiving delivery at record high of 2-3 years. That means it seems inevitable that there will be a buying wave upcoming and with that the price will move in unison.

The Basket

This basket in particular is quite concentrated amongst higher beta names, hence a reason why it’s underperforming vs the index.

We’ve got:

• Centrus Energy | LEU

• Cameco Corporation | CCJ

• Constellation Energy | CEG

• NuScale Power | SMR

• Denison Mines | DNN

Before we move on, a quick word about TrendSpider - the charting platform I personally use.

If you manage your own portfolio the way I do, here’s how TrendSpider fits into my process:

• Scanners for technical and fundamental setups

• Sidekick AI to dig through market data, filings, news, and the names I own or watch without bouncing between tabs

• Charting and automated technical analysis to track support, breakouts, and moving averages across my portfolio and watchlist

• Real-time options, sector rotation, market breadth, news, and sentiment to understand the wider context behind each position

TrendSpider brings the research, charting, and monitoring parts of my process into one place.

If this seems like a useful tool for automating your workflow, take advantage of this 50% discount offer.

Disclosure: I may receive a commission if you sign up or make a purchase through links in this post.

Brazil

The MMMT Brazil index is currently up 31% over the last year (in line with EWZ).

There a lot of valid lenses to look at the Brazil investment case through (Selic rate, politics, Ibovespa valuation etc) which can all lead to different conclusions on whether Brazil is investable today or not.

I’ll dive into them very briefly. But my main focus is on the fact that Brazil is becoming / has already become one of the most strategically important countries in the world.

The US has a critical mineral national security issue that is much bigger than news and headlines suggest. As we’ve known, China dominates the critical mineral supply chain. That doesn’t sound too damaging right now to most, but when you understand the critical minerals needed for technology for the next 10-20 years it becomes a national security issue:

• Humanoids: Rare earth permanent magnets (neodymium, terbium, dysprosium) power every joint actuator and motor.

• Quantum: High purity germanium and indium for qubit substrates, niobium for dilution refrigerator components.

• AI: Copper and fibre-optic rare earths (erbium and thulium) for data center interconnect, cobalt and lithium for power infrastructure.

The aim is therefore to “nearshore”. I.e. move dependence away from China by building supply chains domestically and building supply chains with allies closer to US borders. Outside of the US, Brazil and Venezuela are the key opportunities.

Venezuela has potential but it’s far less investable because of the infrastructure gap and the governance risk and whether this can be monetized at scale yet is still a big question mark.

Brazil on the other hand has:

1. Proximity

2. Scale

3. Relative political investability (italicised for a reason).

It’s rare earths are second only to China’s in terms of % of global reserves held. It holds the the largest niobium reserves and has huge deposits of lithium, graphite, nickel, copper, and phosphate.

Ultimately, this is all why the Brazilian Mining Institute expects the country to attract ~$21.3 billion in investment through to 2030.

However, there’s an obvious complication or two. But the main one is President Lula and it’s a key reason the Brazil trade isn’t working out just yet.

Lula is a big believer in not allowing foreign powers to exploit its countries mineral resources. The workaround so far has essentially been to bypass the federal government by making agreements with Goias and Minas Gerais (home to large rare earth and lithium deposits) but the federal Industry Minister keeps pushing back on these deals. Repeatedly it’s a one step forward one step backwards situation.

But it’s not all bad news amongst a big, unexploitable opportunity.

• The US Development Finance Corporation has backed Aclara Resources rare earths project in Brazil with a $5 million loan.

• The US Development Finance Corporation has backed Serra Verde with a $565 million grant.

  • After this grant, the US government took a 10% stake in USA Rare Earths (USAR), which is in the basket, which then acquired Serra Verde’s Pela Ema mine in a $2.8 billion deal.

  • It has begun letting its existing offtake contracts with Chinese buyers expire.

  • A 15 year set-price supply agreement has been signed with a consortium of private investors and the US government (mainly terbium and dysprosium).

The next significant catalyst is in October 2026 when Brazil’s elections happen. It’s probably the most important catalyst for this entire Brazil and rare earth trade in 2026.

Currently it’s quite even between Lula and Bolsonaro. For reference, Bolsonaro is positioning himself as the candidate that will turn Brazil into the solution for the West’s US mineral dependency on China and therefore likely open up those forecasted $21.3 billion investment opportunities through to 2030 that I mentioned above.

Risks are clear. Upside seems to me even clearer.

It’s one of the reasons NU is quite a substantial position in my portfolio. I personally think a lot of the bad news has been priced into Brazil (given all the metrics like Ibovespa valuation) so the downside just seems far less than than the potential upside.

Application Layer

The MMMT Application Layer index is currently down 2.5% over the last year.

I wrote about the application layer in a lot more detail here in June.

I’ve always been optimistic about the application layer because history has evidenced it time and time again as the place where capital accrues over the long-term.

Read more

1. All of investable biotech today is no more than $1.5 trillion relative to the semiconductor trade which is above $13.5 trillion. That means that even a small rotation out of the the semiconductor basket and into biotech could be very material. This last weeks price action is potentially early evidence of a slight rotation. Whether that is long term or not is a challenging prediction to make.

2. The potential for new drug discoveries using artificial intelligence is vastly under estimated by the market. Biology is becoming computational and that changes everything.

The wider healthcare industry has fought its way through some storms over the last few years. But we believe the above two catalysts are finally starting to converge at the right time today.

For the last 10+ years or so, healthcare has been innovating. It’s been chugging along with aging demographics, technological advances, and biological sciences advancements.

It may not seem like it in the stock market where everyone’s focus has been on the momentum trades of optics, robotics, and alike (until this week)…but if we zoom out a bit it’s clear the advancements here have been extremely material.

• Cancer mortality rate is now ~145 per 100,000 vs 215 per 100,000 in 1980.

• Reversal of prediabetes is now 51%.

• Cardiovascular plaque reversal is now entirely feasible when once considered a “one-way progression.”

• GLP-1s are increasingly viewed as the first longevity drugs.

AI is now permeating every stage of the healthcare cycle from diagnosis, to personalized care, to actual treatment. The gap between what is “biologically feasible” and what can realistically be done with capital, doctor skillsets, and the labour force is narrowing by the year.

The best part of this is that it’s a virtuous cycle that will keep accelerating.

New treatments unlock more new treatments.

Longer life expectancies lead to finding more diseases.

More diseases leads to more medical devices and more drug discovery.

Enormous value creation is set to take over the healthcare sector and I think we’re in the very early innings.

The AI Angle

Medicine is now moving from broader chemical intervention towards very precise biological programming.

The old drug discovery model was about designing a molecule, running a trial, and hoping it works.

It’s changing to something fundamentally different - closed loop systems where AI designs experiments, robots run them, data feeds back into the model, and the cycle repeats itself at a speed no human can match.

This means the bottleneck is changing from biology itself, to whoever controls the data, the infrastructure, and the discovery engine.

This shift creates three clear parts of the chain that are getting disrupted that make up the premise of this write-up.

All in MMMT’s personal portfolio (sizing differs):

1. Discovery: ABCL

2. Data & Infrastructure Layer: TEM

3. Blend of the Above & Commercialisation: LLY

I’ll be diving into them all below this as very mini deep dives but the wider thematic is the most important part I want to convey in this article and I think I’ve covered it above.

To sum it up quite nicely here I’ll say this:

The healthcare market over the next few years wins by:

1. Capital finally rotating to an undervalued theme and away from anything pure AI infrastructure.

2. Capital continuing to rotate into pure AI but the market starting to realize healthcare (especially ABCL, TEM, and LLY) is an AI trade now.

AbCellera Biologics | ABCL

ABCL operates the abLighter platform - an AI driven system that finds therapeutic candidates in 45 days. It’s attempting to change the industry from essentially a numbers game (screen enough candidates and hope something works) to a precision game.

Here’s the core premise for ABCL: Your immune system already knows how to fight disease.

When it encounters a threat, it produces antibodies. Some of these antibodies lock onto and neutralize the threat and therefore have powerful properties within them that can make powerful drugs.

ABCL’s finds these antibodies rather than building them from scratch in a lab.

Partners (LLY, ABBV, BIIB, NVS etc) bring the end target - the protein involved in a disease that they want to treat. ABCL then runs the discovery using AbLighter. From there, if successful, the partner takes it through trials and ABCL earns as those trials progress.

That’s business model #1.

Business model #2 is the more exciting model with more upside (though riskier). It’s the proprietary pipeline and it’s where ABCL owns the asset itself and the economics flow entirely back to ABCL rather than through royalty payments.

ABCL635 is targeting NK3 for menopausal symptoms. It has Phase 1/2 topline data expected in Q3 which, if positive, will reframe the entire narrative on ABCL to a TechBio firm with its own drug pipeline.

The downside risk with ABCL is more cushioned that perhaps it may feel for a small-cap biotech company. It’s currently sitting at $1.8 billion MC with $504 million in cash and committed government funding. So that’s ~30% of the entire MC sitting in cash with over three years runway at the current burn rate with a management team that has delivered so far on all timelines they have publicly projected.

Poor data on Phase II would most definitely threaten the stock in the near term, but the company would not be threatened too materially in the short term. The royalty stream from the 40+ partners continues and as long as these programs keep advancing, the $1.8B MC seems closer to a floor. It feels like a free call option on ABCL365 to an extent.

Tempus AI | TEM

In a sentence, TEM is the data infrastructure that makes AI driven drug development possible.

The whole premise of AI in drug discovery is pretty straightforward: train a model on enough high-quality data and it can identify patterns that would take humans decades to find, right?

The main issue is that the data required to do this well (real patient records, treatment histories, pathology slides etc) is very difficult to assemble and very expensive to get a hold off.

TEM has this data because they invested huge amounts in their foundational years before anyone knew how valuable it would be.

And this data essentially is changing what is possible for pharma. TEM uses this data to actively tell drug companies things they couldn’t otherwise know before committing hundreds of millions of dollars to a trial. CEO, Lefkofsky, explained this pretty neatly:

He said that a pharma company came forward with 25-30 clinical trials to work through TEM’s network. TEM only took 13 of them because they estimated the other 17 would likely fail before generating any meaningful results.

It’s ultimately all about the data they own.

Here’s how the flywheel actually works in practice for TEM (I’ll make this section fairly concise compared to the complexity of it…feel free to skim over it. I’ll go into a bit more detail on the valuation section).

TEM’s different business segments are ultimately built on the data foundation. They have access to the data and the clinical context behind all of that data.

1. TEM sequences patients. I.e. they run genetic tests on cancer patients, reading the DNA, RNA of their tumours.

2. TEM pulls in connected clinical data. I.e. TEM pulls in patients medical history (treatment records, doctors notes, lab results, scans etc).

3. TEM generates molecular and diagnostic insights. I.e. TEM combines the genetic results with clinical history to produce data with background evidence.

4. TEM pushes those insights back to physicians. I.e. The treating doctor gets a smarter report, not just a list of mutations…actual guidance.

5. TEM gets more data. I.e. Every new test adds another patients genetic and clinical data to the database.

6. TEM’s models improve. I.e. More data makes the AI predictions more accurate.

7. Pharma pays to access the data. I.e. Drug companies then pay TEM to interrogate this dataset to understand why drugs work in some patients and not in other patients.

8. This revenue funds more patient sequencing. I.e. The pharma licensing money subsidises the cost of running more genetic tests.

9. Repeat endlessly.

This is the foundation that Tempus is built on. They’ve now done this so much to the extent they have 500 petabytes of data across 45 million patients.

Take a second to truly appreciate this foundation. I know the path from here isn’t straightforward but just understand how far ahead they are from any companies now that are trying to replicate this business model.

It’s years/decades.

Data will be the moat for years to come. The companies that invested heavily at the right time (i.e. 10 years ago) will be the ones that truly reap the rewards over the next 10 years.

Valuation:

Of course it all comes back to valuation at the end of the day.

Aim #1 is always to own a quality business. Aim #2 is about owning a quality business at the right price.

If one of those isn’t investment worthy, there is no trade.

Here’s why I think TEM is so undervalued:

At ~6.5x sales, the market still prices TEM like a diagnostics lab for the most part… i.e. Guardant Health (GH) trades at 15.9x NTM sales today.

The way to think about TEM’s valuation is to split it out into two separate businesses.

Diagnostics: I think $2.4 billion in revenue by 2028 for diagnostics is feasible if volume continues to steadily compound and ASP moves closer to $2,000. That assumes ~26% CAGR from 2025.

At a 22% operating margin (already operating at 65% gross margins), that’s ~$530 billion in operating profit which means operating profit for that business would be growing at ~25% CAGR. Put it at 1x PEG (25x multiple) and we have a $13.25 billion business there.

Data & Applications: This is where it gets slightly more interesting for me here. My estimates put us at $820 million in FY28 revenue. This is a SaaS like business so +30% margins are very likely (and perhaps still conservative).

As a SaaS like business with +30% margins, I think a 30x NTM PE isn’t overly bullish at all. 30x PE on that gets us a $7.4 billion business which I think is an absolute base case.

Combined that’s an EV of +$20 billion vs $10.3 billion today.

Eli Lilly | LLY

LLY was once (and still is) a GLP-1 story.

Read more

Part 1 was released on Sunday:

Part 2 focuses on:

• Robotics

• Optics

• Solar

Three themes that have been hit hard by the recent market sell off.

Robotics

I think many people appreciate that robotics is a huge megatrend coming up.

But few people realize how much of a national security issue for the US right now it is…which is central to the investment thesis today.

China is currently way ahead of the US. Their 2026 humanoid output is +100,000 vs the US which is ~14,000.

That issue stems because of a few different reasons, but ultimately gaps this visible do attract policy, urgency, and capital into the theme.

The most obvious issue today is the rare earth supply chain. The US isn’t doing anywhere near enough to prioritize this issue right now but that can’t remain which is why USAR to me is a clear thematic opportunity.

But zooming out, the robotics megatrend is so vast.

I wrote about it in depth in the article attached above but a simple summary would go along these lines:

• Backbone

  • Actuation/Bearings: Without actuators robots are just an intelligent statue. There’s only a few key players here.

  • Sensory & Perception: LiDAR, IMU, tactile and force sensing, analog and semiconductor sensor components. There’s many angles to play this.

  • Software/AI Integration: This is where the US leads now. The AI foundations are strong but the physical components that make up the robots is where the US is behind quite significantly.

  • Motion Control: The sophisticated chips that orchestrates a robots every move calculating how much each joint must rotate and at what speed.

  • Communication: The invisible layer that keeps everything connected.

• Frontier:

  • Medical Robotics: More than just a few niche surgical aids now. It’s set to transform healthcare from surgical systems to diagnostics, to rehabilitation exoskeletons, to pharmacy automation and logistics bots.

  • Defense Robotics: Low cost robotic systems are now destroying high value manned assets forcing this shift towards unmanned drones.

  • Logistics Robotics: This will likely be the first large industry to be fully automated. AMZN alone already has over 1 million humanoids deployed.

  • Humanoids: Once quite an un-investable segment of robotics until CCXI went public.

  • Professional Robotics: The smaller, specialized, task-specific machines optimized for professional applications.

The current basket includes 13 names with 40% of the basket allocated purely to OUST and NVDA.

OUST is in my personal core portfolio as my 2nd largest position so I’m very bullish there.

You can see my prediction with OUST here:

The rest of the basket is made up of some frontier and some backbone plays like OII, DGII, AIT, MBLY, PRCT and FLS.

I’ve weighted all those plays quite small but the two that have my most attention at the moment to increase weighting for are:

1. PROCEPT BioRobotics | PRCT: PRCT is a surgical robotics company with one product that treats enlarged prostates. The numbers are very interesting on this one if they execute and if they can scale. Most MedTechs like this really struggle for the initial years because operating and gross margins are so thin. It’s only when scale hits that numbers start to look quite interesting. PRCT appears to be at that inflection point now.

   1. PRCT trades at 2.3x NTM sales whilst set to grow revenue at 29% in FY26 and 24% in FY27.

   2. If PRCT reach EBITDA profitability next year (which is expected), then we should expect ~$170M in EBITDA in FY29 based on $680M in revenue and a 24% EBITDA margin.

   3. At $170M in EBITDA and a 20x EBITDA multiple (TMDX today trades very cheaply at 14x), we could expect an EV ~$3.4B which is 3x from today.

2. MobilEye Global | MBLY: MBLY is building the perception and decision making systems for autonomous vehicles and humanoid robots. It funds its frontier R&D through a highly profitable ADAS chip business. It recently acquired Mentee which is targeting commercialisation in 2027.

   1. Revenue growth has been slow in FY23, FY24, FY25 and likely FY26. But FY27 is when the company is likely to ramp up again to +22% estimated growth which would mean a 3.3x sales multiple would be cheap.

   2. If MBLY meets estimates of $3.8B in revenues in FY29, then they’d be growing revenues at 25% CAGR.

   3. Compare that to a partial competitor like NXPI which is growing ~7.9% CAGR for 5.2x sales, and we start to see the re-rating potential for MBLY down the road.

Last chance to grab TrendSpider at a steep discount.

Get 38% off + $98 in free upgrades if you sign up by July 8th at midnight — once it’s gone, it’s gone.

TrendSpider is the charting platform I personally use, which has helped me spot the most in my investing.

If you manage your own portfolio the way I do, here’s what TrendSpider provides:

- Scanners that surface technical and fundamental setups before they’re obvious - the kind of early entries that turned OUST and NBIS into two of my biggest winners this year.

- Sidekick AI to dig into market data and portfolio holdings without jumping between five different tabs.

- Charting and automated technical analysis to track support levels, breakouts, and moving averages on every name you hold or are watching.

- Real-time options, sector rotation, market breadth, news, and sentiment to understand the broader market context behind any single trade.

This is one of TrendSpider’s biggest sales of the year. If you’ve been thinking about trying it, this is a good opportunity to sign up with a discount.

Optics

Copper can’t handle AI-scale bandwidth at 800G and 1.6T speeds which is why photonics is the solution.

The optics theme was obviously one of the biggest winners in the entire market over the last year with the MMMT index of just 5 stocks now returning 303.1% vs the SPY at 20.2%.

This recent pullback with the likes of AAOI down ~42% in the last month is nothing more than a correction in what I deem to be a long term buildout.

The major inflection point we’ll likely see as per most earnings calls is mid 2027 as 1.6T scales and as CPO broadens past NVDA and AVGO.

Markets will likely price this ramp up before it actually arrives, which is what has happened to an extent so far, but looking at individual valuations I am of the opinion we are far from being priced in.

Here’s my optics piece from December 2025…nicely before all optics plays surged in January and February.

My core photonics play is AAOI. It’s a stock I am down ~21% on in my core portfolio, but it’s one of my highest conviction plays right now that I’ll personally keep building out sub $130.

A few reasons for this:

1. If AAOI hits projections, it’s ~12x in revenue growth from $455M in 2025 to $5.6B in 2028 (that’s being conservative on management assumptions because of the timeline). There’s no other company in the market forecast to grow that quickly especially at 6.5x NTM sales today.

2. All hyperscaler customers want to buy all transceivers AAOI can possibly supply. It’s a supply constraint and not a demand constraint.

3. If you have partial trust in management, then looking at current financials for AAOI actually makes no sense. This is a 2027/2028 ramp up story likely like no other in the market. We saw it with NBIS earlier in the year and I personally believe AAOI is in the same boat, if not even more bullish than NBIS was before the recent run from $92 to $300.

Solar

Solar is an interesting, and very underappreciated trend. The MMMT Solar index is up 69.9% in the last year vs SPY at 20.2% but it’s been a very volatile ride.

Read more

Part 1 is looking at the following baskets:

• Critical Minerals

• Grid Infrastructure

• Miners

Here’s a snippet from my web-app:

Critical Minerals

YTD the MMMT Critical Minerals Index has returned 54.6% vs SPY at 9.0%.

It’s been a volatile basket to hold with stocks like MP surging to $100 last summer before falling back to the mid $50s today. USAR has been another volatile play (which we hold in the core portfolio) jumping from $14 to $44 and back to $19 in the space of 8 months.

We’re now ~12 months on from China’s restrictions on rare-earths and permanent magnets which sent the likes of USAR and MP soaring. However, the market so far in 2026 has seemingly forgotten about the technological revolutions ahead in AI, physical AI, and even quantum and the rare earth requirements for this.

Valuing these plays at the moment is a challenge. It’s likely why USAR and MP and LYC have struggled recently.

But despite what the stocks are doing the underlying thesis remains clear:

Brazil remains the big opportunity which is one of the main reasons I am building my USAR position. It’s by far the most “nearshore” ally to the US and depending on the presidency ballot in October this year, the outlook for Brazil as an ally could be extremely advantageous.

Flavio Bolsonaro has already attempted to position Brazil as this ally to the US and the US’s solution for mineral dependency.

This is one reason why I hold USAR in my core portfolio.

USAR acquired Serra Verde which owns the Pela Ema mine. Pela Ema is the first large scale producer or rare earths outside China. It’s also signed a 15 year supply agreement with the US government making it quite a clear bet on this rare earth thesis.

I also have Sunrise Energy Metals (SREMF) in the critical minerals basket. It currently trades at $12.49. Here’s what I sent to paid subscribers back in March when SREMF traded at $5.50.

TLDR Summary for critical minerals:

Markets are not pricing in the physical bottlenecks of the AI buildout. Huge amounts of copper, gallium, silver, gold, platinum, scandium, and niobium are needed for the continued data centre buildout, and the physical AI world we are entering into.

It appears today that the US is technologically superior to China in models but China owns over 70% of the worlds critical minerals supply chain. That’s not a strong position for the US to be in.

Last chance to grab TrendSpider at a steep discount.

Get 38% off + $98 in free upgrades if you sign up by July 8th at midnight — once it’s gone, it’s gone.

TrendSpider is the charting platform I personally use, which has helped me spot the most in my investing.

If you manage your own portfolio the way I do, here’s what TrendSpider provides:

- Scanners that surface technical and fundamental setups before they’re obvious - the kind of early entries that turned OUST and NBIS into two of my biggest winners this year.

- Sidekick AI to dig into market data and portfolio holdings without jumping between five different tabs.

- Charting and automated technical analysis to track support levels, breakouts, and moving averages on every name you hold or are watching.

- Real-time options, sector rotation, market breadth, news, and sentiment to understand the broader market context behind any single trade.

This is one of TrendSpider’s biggest sales of the year. If you’ve been thinking about trying it, this is a good opportunity to sign up with a discount.

Grid Infrastructure

Grid resilience is one of the most underappreciated levers for technological dominance over the next decade.

Forget just AI - think technology in general (robotics, quantum, everything). The most developed economies today are massively constrained by a very outdated grid that is in need of a complete rebuild.

So much so, that the White House invoked the Defense Production Act (Section 303) in April 2026 which specifically focuses on transformers, substations, high voltage breakers. It’s evidence that the domestic capacity is now not sufficient to meet the baseline demands moving forward.

IEA (International Energy Agency) projects are spending ~$550 billion in 2026 alone (that’s a 20% increase YoY).

Total (add in all end-use electrification like EVs, electrified industrial processes etc) electricity investment is now nearing $2 trillion per year.

Now let’s get into the specifics of what this means from an investment angle.

The wider theme is obvious, but how quickly this is actually achievable is a completely different story.

With how fast technology is moving, and how large the demands are for electricity, the grid buildout appears structurally far too slow to keep pace with the demand. In the US alone, data center buildout is projected to go from 21GW in 2026 to 84GW annually by 2030.

Given that grid interconnect is the constraint here, perhaps the bigger opportunity therefore lies in the companies enabling buyers to generate or access power on their own timeline, and not to do with the companies selling generation to the utilities.

Capacity simply isn’t growing fast enough to cover the 21GW to 84GW of US data center demand so there needs to be an alternative.

The biggest winners will be those that allows a buyer to generate power without relying on the utility. The answer today is Behind-The-Meter fuel cells…i.e. Bloom Energy (BE) or gas engine names like Innio (INIO).

The other angle is SMRs but the target for them is likely a 2028-2030 story…perhaps even later. I put SMRs in the Uranium basket and also as a long duration hedge on the grid problem. I haven’t included it in the power grid upgrade basket yet.

Miners

As you’ll know, I’m fairly aggressively building my positions in silver, copper, and scandium miners and I continue to believe this is:

1. One of the most important divergences in the market today.

2. One of the safer ways to expose yourself to the technological revolution without investing in the tech directly.

Read more