The buildouts have always been right about the future, but often wrong about who gets paid the process (e.g., which companies benefit, and which wipe out).

Subscribe now

1) Historical builds: patterns and learnings

The overall arc:

1. Transformative Innovation: A stunning innovation transforms productivity

2. Explosive Demand: Usage grows exponentially.

3. Supply Constrained: The innovation’s deliverability and accessibility is choked by lack of infrastructure.

4. Capex Buildout pt. 1: Capex is invested rapidly to fill that supply gap.

   1. This is the point in the cycle we’re in today. Capex is being invested rapidly to fill the GPU compute gap.

5. Capex Buildout pt. 2: Excessive capex build out leads to oversupply

   1. Notably, from a timing perspective. Demand does get built, but usually takes an extra few years; but the builders have taken on too much debt, so they can’t withstand the timing crunch…

6. Prices Fall: Utilization trails supply, leading to prices and revenues falling

7. Wipe Outs: Builders wipe out

8. Ancillary Long-Term Beneficiaries: beneficiaries are those companies that thrive on these low prices (cheap rail, cheap electricity, cheap fiber, cheap oil).

These case studies dissected in far more detail below.

In most of these cases, the builders behind this capex bubble wiped out.

The common dynamics in failed cycles were:

• Supply overbuild: Supply (fiber, railroads, oil wells) exceeded forecasted demand. Building efficiency increased, so build per dollar rose. Demand eventually grew into the capacity, but it took an additional 3-10 years. So the drop off in revenues, fused with debt leverage, wiped out the builders.

• Excessive building competition: without coordination, too many builders went after the same demand.

• Commoditized output: miles, bits, oil barrels. All commoditized without differentiation.

• Liquidity & leverage: a timing problem on fixed debt obligations. Many balance sheet wipe outs.

Exceptions, where the builders didn’t fail, were:

• State funded, no private capital at risk (Interstate highways, Apollo, China’s infra build).

• Regulatory monopoly (Electrification).

  • Electrification’s builders survived because regulated local monopolies eliminated competitive replication and let them keep the margin from falling generation costs, while Insull-style load engineering matched demand to capacity. A wipeout still happened, but only in the leveraged holding-company pyramids stacked on top, not in the operating utilities.

• Diversified Majors’ balance sheets (specifically with Shale).

  • In the US Shale buildout - the Independent E&Ps (aka upstream drillers) wiped out. 114 bankruptcies in the following two years. However, the Integrated Majors (Exxon, Shell, BHP) survived with some scratches (took impairments and write-downs, but given their scale, they survived fine).

The beneficiaries

The beneficiaries ended up being those companies that thrive on the low prices: cheap rail, cheap electricity, cheap fiber, cheap oil.

So history rhymes. Let’s use it. Where are we in the cycle, who are today’s builders, what’s genuinely different this time, and who ends up owning the future all this concrete and silicon is buying?

2) AI Buildout Today

Where are we in the cycle? Who are today’s builders? This time is different?

Where are we in the cycle?

In this point in time, this seems to be a very healthy buildout. Hyperscalers are ramping up combined capex by +75% vs. last year and they’re all saying it’s due to supply-constraints on excess demand. Frontier compute build continues to run at full utilization. Looks like electrification, not fiber.

But… cycles never turn on the surface. They turn on the margins. There are three margins worth watching.

• Marginal Pricing: the price of the commodity at the edge. Blackwell-class compute is rationed. However, two generation old compute is not. H100’s cost $8/hr in 2023, and dropped to ~$1-$2 by 2024 (today at ~$3/hr). This doesn’t seem like a differentiated good, it’s a commodity, finding a clearing price. This looks quite like bandwidth in 1999 and oil in 2014. And given exchanges are launching GPU compute futures, this is strong evidence of commoditization.

• Funding Source: Buildouts seem to share a common pattern. They’re born on cash, and die on credit. This is true for every single case study in history. Today, AI capex is consuming ~94% of hypserscalers’ operating cash flow after dividends and buybacks, up from ~76% in 2024. In Sept & Oct 2025 alone, tech companies borrowed $75B in debt for data centers (this is more than double the annual average of the past decade). Meta financed its Hyperion campus with $27Bn of off-balance sheet debt. Oracle’s total debt has pushed past $130Bn. Morgan Stanley estimates an additional ~$800Bn of private credit will be needed to fund data centers through 2028. This buildout definitely started from the cash flows of some of the most profitable companies in history, that’s truly unprecedented and a real cushion. But the marginal dollar has now flipped from cash to credit. And in every prior cycle, this exact flip was what broke the arc.

• Quality of the “demand”: Contracted demand is different from funded demand. OpenAI announced ~$1.4Tn of compute commitments over eight years (against ~$13Bn of 2025 revenue). Then, this Feb, reset that figure to ~$600Bn through 2030. It’s a bit of a tell, about the reliability of the demand we’re building compute data centers for…

Zooming out, we’re mid-buildout. The demand is very real. Yet, the two leading indicators that preceded crashes at every prior turn (softening prices, and hardening debt leverage) are both flashing early. Still, we’re in set up phase, so this doesn’t feel like the top.

Who are the builders?

• Hyperscalers: Microsoft, Google, Amazon, Meta. Large, diversified players. Somewhat like Exxon in shale, as in, if a glut comes, it costs them impairments and hurts returns on capital, but not existence.

• Neocloud Cohort: CoreWeave, Nebius, Lambda, and many others. Selling a single commoditized product (GPU hours), funded with leverage collateralized by those same customers, selling to a handful of customers who are mostly the same customers everyone else has. CoreWeave for example has $35Bn of total debt (vs. a market cap of $40Bn, and $3Bn of cash on hand); it’s flagship debt facility carries an ~11% variable rate, secured against its chips. These companies are growing rapidly, both top-line revenue as well as profit margins. As long as the pricing spread continues, they’re well positioned. But if this cycle produces its version of shale’s 114 bankruptcies, I’d wager most of those come from this category.

• LLM Developers: OpenAI, Anthropic, xAI. They sit on both sides - they’re are the demand behind this build out, and have a hand in building themselves (e.g., Stargate and Anthropic’s $50Bn data center program). Closest analogy seems to be the joint-stock railway companies of the 1840s, with investment underwritten based on continued exponential growth rather than current usage. While these railway JVs collapsed, one key difference between that analogy and the LLM developers is differentiation vs. commoditization. Frontier models are not interchangeable. As long as the frontier stays differentiated, these labs can maintain pricing power, and they have a unique advantage vs. history. However, if open-weight models close the gap, and intelligence commoditizes, they start to look like fiber builders.

• Notable single player mentions:

  • Oracle: A single levered bet. It’s borrowed aggressively, pushing debt to >$130Bn, to fund a single customer, OpenAI’s $300Bn five-year contract. Bond markets don’t love this, S&P downgraded Oracle’s debt, and its yield has shot up to 7.18% (vs. 5.38% in fall of 2025). Reports say JPM is struggling to syndicate the Stargate debt. If OpenAI delivers, Oracle’s in a great position; but if things swing the other way…

  • Nvidia: Smells a bit like Carnegie Steel, selling rails to every railroad builder. And a bit like Lucent and Nortel, who vendor-financed their telecom customers in the late 90s. Unlike Lucent and Nortel, Nvidia’s stakes in OpenAI, xAI, Anthropic and the neoclouds are smarter, funded from profit (not debt). But it still carries a similar problem - the seller (Nvidia) is funding the buyers, making end demand harder to read.

• and the utilities, of course, win regardless. Whatever happens to chips, the substations probably get paid well, due to regulated guarantees.

This time is different?

When everyone thinks you this time is different, it usually isn’t…

Honestly though, I do think there are some very real differences:

• GPU obsolescence (I’ll call this ‘the Treadmill’ later). GPUs become economically obsolete in roughly 5 years - so a glut would self-liquidate in a sense. Most everything else in history was far longer-standing (fiber, railroads etc).

• Demand is real. Utilization is high. This was true of both fiber - but after full build out, <5% of fiber was lit. Frontier compute is currently running on full. On the fiber vs. electrification spectrum, this seems more like electrification. But if we overbuild, we tilt the other way. And even if token volumes grow exponentially, it could fall short of the capacity that’s priced in - if price per tokens continue to fall as fast as they’ve been (~10x every year or two).

• The funders are the richest companies in history. Hyperscalers are diversified. Like Exxon. But the marginal dollar has turned into debt.

• Regulatory Step-In. Compute is a geopolitical arms race. Nobody backstopped fiber. Washington’s taken an equity stake in Intel, OpenAI’s offered the US Government a 5% stake. China’s infra sprint, Apollo program, and electrification all did just fine, because of regulatory backstop. Compute build isn’t fully backstopped today, but it very well could have a national-security floor under it.

3) Where do we go from here?

End game scenarios

There seem to be three scenarios that come to mind:

• The Fiber Ending. Demand grows, but slower than contracted capacity. Token prices fall faster than volumes can compensate. Debt leverage at various choke points (including private credit) forces restructurings. Majors take impairments, and also buy assets out of receivership for cents on the dollar, and emerge stronger. Notably, not an AI-pessimistic scenario. Just as the internet didn’t fail in ‘01, only the builders did.

• The Electrification Ending. Demand keeps pace with capacity. A wipeout happens, but not due to an oversupply crash, rather due to the financial engineering stack. SPVs, GPU-backed loans, and private credit funds’ running maturity mismatches melt collateral. Given so much of this buildout has already migrated into off-balance-sheet vehicles, and is owned by insurers and pensions, this ending has more surface area than most people appreciate. The losses here would hit fixed-income portfolios more heavily than equity markets.

• The Treadmill (no precedent). No bust at all. Obsolescence absorbs excess before it can crater prices. So the cycle continues into permanent capital intensity, capex forever, and tight margins on returns (compared to the cost of capital). A bit of the airline outcome - demand grows relentlessly, but roughly no returns accrue to those who funded the planes.

It’s still early in the cycle, so there’s no obvious ending. If I had to place bets, I’d say the fiber ending for the levered edge, the electrification edge for the financial stack. And given the Treadmill, a lot more upside for humanity, for consumers, and for those who thrive on cheap intelligence.

The long game: who thrives on cheap intelligence?

the most consistent lesson in history is that the greatest fortunes were made downstream by whoever thrived on the commoditized output. So the prescient question is “what becomes economical when a token costs 1% of what it does today?”

Some of those would be: agents that do entire jobs rather than answering questions. Software written on demand that can be thrown away after use. Tutoring and medicine, hyper-personalized. Exponentially better drug discovery. Robots with cloud brains.

Defining companies of the cheap-intelligence era will become far clearer into the glut. A good historical case would be Netflix. Benefited greatly from cheap bandwidth, making streaming possible. Netflix was founded in 1997, but didn’t launch streaming until 2007.

Notably there are two other places where value migrates when an input commoditizes: (i) the scarce complement, and (ii) the distribution. When bandwidth commoditized, value moved to content and aggregation. If intelligence commoditizes, the scarce complements are:

• electricity

• proprietary data

• physical-world integration

Distribution means whoever owns the surface where billions of people consume that intelligence daily.

Which brings me back to the same names. If forced to evaluate it today, I’d think the long-term winners are Anthropic, OpenAI, Google, and Microsoft. Not because they’re the builders, but because they own distribution, and consume intelligence across everything they ship. And this is the fuzziest part tbh - since history would mean the downstream names don’t even exist yet, and the builders are the ones that get wiped; but again, the companies in this paragraph own distribution downstream, not just the buildout….

The frontier labs will need frontier differentiation to survive. They’re the first builders in this 180 years of case studies whose output might not commoditize. That “might” is kind of the entire bet - and it’s going to be re-evaluated every model generation.

The zoomed out pattern

Looking across these 180 years. The buildout’s always been right about the future. Just wrong about who gets paid for it. Railway promoters were right about rail. Carriers called the internet correctly. Frackers were right about shale. And yet the investors wiped out anyways. Because miles, watts, bits, and barrels. They all did the same thing, their prices collapsed.

Compute’s next on that list. Maybe. So what I’m doing:

• Keeping an eye on the two numbers that drove the turn in every prior cycle:

  • (i) the price of the commodity (GPU compute) on the edge; and

  • (ii) the share of buildout funded with debt.

• Build a running list of the companies I’d want to own the day intelligence gets cheap. Prepare for that buying opportunity. It could be this year, it could take another few years. But when the time comes, be ready.

• Higher dry powder. Multiples are steep today, and many long-term winners are still to be created. Keeping reserves to buy when the glut comes.

Subscribe now

The headline: Trimming the froth, and holding more dry powder than usual.

Shiller PE Ratio for the S&P 500 (1976-2026):

While good buying opportunities seems far and few between, many of the best companies are experiencing exuberant valuations. I’ve been trimming, and writing here with the intent to openly share my thoughts on how I’m viewing the markets. I’d suspect this isn’t anything novel, and most thoughtful allocators have have similar views. In one line: trimming the froth, and holding more dry powder than usual. The rationale: Strong fundamentals. Frothy prices. Thus, more dry powder & raising cash.

Remember, I adhere to a long-term view when investing - with a 40+ year horizon, and the mindset of buying up positions in good assets especially when prices are compelling (e.g., cheap).

Today, it’s the opposite, I’m cautious and fearful. Prices are quite frothy. And yet, fundamentals are very strong. Both on a macro level (employment, overall corp. earnings, inflation, and GDP) as well as company level (top-line growth, profitability). Yet, the prices for these assets are higher than high, the market is pricing in extraordinary forecasts, and the AI exuberance is driving irrationality.

And so.. it’s pruning season. I’ve been trimming the overgrown positions.

Subscribe now

CrowdStrike, as an example.

To be more precise, as an example, one of my favorite holdings has been CrowdStrike (NASDAQ: CRWD). Fundamentally, the company’s performance is exceptional:

• In Q1, annualized revenues grew to $5.6Bn, and growth clipped >25% YoY (fourth quarter in a row of accelerating growth).

• Strong gross margins, record positive cash flows, and raising guidance (outlook).

• Best product in cyber. Defensible. and cyber is one of the few software budgets still expanding (due to combo of both AI and geopolitical risk tailwinds).

• Top-tier management.

Suffice to say.. I quite like the company. Fundamentally excellent asset.

And yet.. the price is exceptionally frothy.

• Today, the company trades at a ~28x growth-adjusted revenue multiple (I’ll call this ‘revxg’ going forward)

  • e.g., $194Bn market cap. vs. $5.56Bn annualized revenue growing 25.6% = 27.7x

• I’ve historically bought in (heavily) when the shares have been discounted

  • in 2024, when the outage occurred, I bought heavily from late July through mid-August.

  • similarly at the start of the SaaSpocalypse this year, I bought in from Jan-Mar).

  • In these periods, I was paying ~15-18x revxg, expensive still, but reasonable for a best-in-class asset in a growth category.

• But when CRWD’s price surpasses >25x revxg, I start getting wary.

• This week, the stock has traded up another whopping +9% ($764/sh. vs. $700/sh last Friday) - due specifically to (i) Iran de-escalation, and (ii) the stock-split. This much of a pop is absurd in my view.

• Today that means the revxg multiple is nearing 28x

• And so, I’m trimming my position - today I’ve sold off ~90% of my holdings in CRWD.

• I’ll sit on the sidelines, and wait to buy back in when the markets normalize again.

Other specific stocks & overall portfolio

Another example is GOOGL - now this one I’ve not sold, it’s not over the top, but I’m very cautious, and as much as I love the company’s exposure to the AI wave, I’m certainly not buying any more shares. DAL is another good company, and yet one where I’ve effectively exited my entire stake.

Now that’s not the case across the board. There are other categories where I’m gearing up to buy up:

• In my oil & gas holdings for example, I trimmed down a bit when the Iran squeeze drove those prices up - now as oil prices subside, I think they’re valued quite reasonably so I’ll be increasing my positions.

• NFLX is cheap and a good asset, so I’m slowly building a position in there.

• UNH has been one of my largest positions (largest now after trimming CRWD), and one I’ve high conviction in and will continue to hold.

Put simply, I’m shaving some excessively priced positions, and finding it difficult to find compelling buying opportunities right now.

The net result: I’m ~50% dry powder (whereas I usually target minimal dry powder and staying maximally invested).

Broader Market Dynamics

The case described on the asset-specific level, holds true for the market as a whole as well in my view. Fundamentals are strong, and yet prices are frothy.

Strong Economic Fundamentals:

• Real GDP Growth sits at 2.1% YoY, quite healthy.

• Unemployment, albeit slightly elevated vs. 2022 levels (3.4-3.5%) remains strong at 4.3% today.

• Corporate earnings growth is experiencing one of its strongest periods since, with aggregate YoY EPS growth of +30% in Q1’26 - this is driven especially by (i) AI infra boom; (ii) resilient consumer demand; and (iii) energy sector tailwinds.

• Consumer health. Looking at credit card delinquency rates here. At 2.9% for Q1, this is mildly elevated relative to the preceding decade, but doesn’t show signs of real strain.

Frothy Prices (and expectations):

Shiller PE Ratio for the S&P 500 (last 50 years)

• S&P 500 today is trading at 27.3x trailing earnings

• Market have priced in an earnings growth rate of 36% for the next 12 months. This would be incredible, especially this late in the economic cycle

• The Shiller PE ratio for the S&P 500 sits at 41.4x, significantly elevated.

So again, fundamentals are strong, and yet prices are frothy.

There are other signs of irrational exuberance:

• SpaceX IPO… say less.

• Stock-financed M&A: in current quarter, 49% of US M&A deal value is being financed with stock. For comparison in the tech bubble this was 2/3, and in ‘21-’22 it was 45%. In normal periods this is usually ~25-30%.

• AI exuberance:

  FT had a great article on Sunday titled “AI ‘exuberance’ risks ending in lengthy investment bust, BIS warns”. It runs a comparison vs. historical major innovation booms (including canals in the 1830s, railways in Britain in the 1840s, the roaring 20s, and the dotcom boom of the late 1990s. The BIS summary, it’s [AI] “a genuine technological breakthrough that attracted capital in excess of what commercial returns could ultimately justify”. I could not agree more.

Looking ahead

So where this leaves us:

• On Actions: Personally, as detailed above, more cautious than usual, and higher dry powder than usual (although still invested and still looking for opportunities).

• On Market Posture: While I think the markets are irrationally priced, GDP and unemployment factors do not point to a recession. Yet, prices point to a higher than usual possibility of a price correction. No guarantee, as markets can stay irrational for long periods of time.. but still risky prices. The missing piece for a correction is a catalyst - it takes a clear catalyst to drive a correction, and under Warsh’s stewardship fused with the quasi-peace agreement with Iran, there may be more smoothness than volatility on this front today.

With Anthropic and OpenAI’s IPOs on the horizon, I’d expect the exuberance to continue for a little bit longer. So holding cautiously.

Educational opinions only, not investment advice. No perspectives here on where the markets will go. Just food for thought

I believe the market conflates both (i) macro panic from Iran war; and (ii) structural AI disruption (SaaSpocalypse). The software stocks that are still depressed despite the ceasefire - I’ve built up more positions off this morning’s ceasefire announcement.

Subscribe now

Ceasefire and reactions

What just happened? Trump announced a two-week ceasefire with Iran.

• Oil dropped: down -18% (from ~$111 > $92 bbl)

• Equities surged: +3% or so

While the markets are overall buying up everything in reaction - I think the smarter focus is on specifically where to buy up.

Side note: Notably I do think that Iran is still going to aim to retain control of the Strait of Hormuz; and layering in tolls there. The west is not going to like this.. so doubtful this is a pure clean resolution…

SaaS: The Double Discount

For the past 5 weeks, software and growth names have been in double-whammy land. Both (i) the AI SaaSpocalypse continued and (ii) the Iran War driven macro volatility.

The AI SaaSpocalypse: genuinely structural. Impacts per-seat pricing. Threatens moats. Compresses enterprise SaaS multiples. Though.. as I wrote about in my last piece, I do think there’s nuance there.

The Iran War driven macro volatility: This hammered many many risks. But now has largely dissipated. Of course, it’s not entirely gone. But I do think that the markets haven’t breathed the deserved life into SaaS yet.. and I think that there’s still a bit of a window to buy up on the software cos you like - which is what I’ve been doing.

Separating the headwinds

By sector - macro headwinds (Iran War) vs. SaaSpocalypse (AI disruption)

What I intend to illustrate with the above is that not every beaten down software name will recover. But the golder ones (the ones that are more macro compressed but AI resilient) are still compressed with room to grow.

Notably, here I am talking about long term 40-year investment horizons. My usual style. Not the recent short term trades.

At a high level:

• The good - I like and buying: cybersecurity and regulated SaaS are the sweetspot. Their drawdowns were primarily macro-driven - investors de-risking broadly, not making a considered judgment about AI impact. Cybersecurity demand actually grows with AI adoption (bigger attack surface). Regulated industries have compliance moats AI can’t easily abstract away. These categories got thrown out with the bathwater. Again more detailed in the last write up.

• The perma cooked - I’m not going back in on: horizontal SaaS & dev tools. Their drawdowns are primarily AI-structural - the per-seat model is being undermined by agents (e.g., less seats on per seat pricing businesses). More competition for tools. Build-vs-buy shifting hard toward build. The ceasefire doesn’t help them. The macro headwind fading is noise relative to the structural headwind that remains.

The framework from my last piece still applies for identifying the diamonds in the rough:

• Toll booth businesses: a high-moat enterprise that controls a bottleneck in the value chain.

• AI-native platforms: built in the modern era, and thus capturing the commercial value of AI rather than being disrupted by it.

• Picks and shovels: AI infrastructure and observability.

What’s changed now is the timeline. Q1 earnings start in under two weeks. The macro fog lifts and fundamentals get scrutinized with fresh eyes. So I think those will be the catalysts for the pops. I’ve already started buying up but may do a bit more before earnings season drops.

And notably - there’re risks

Of course, I could be wrong. Notably again this is for the high quality assets I want to hold for 40+ years. So I’m not going to be portfolio pained if the geopolitical volatility comes back in play and causes prices to compress.

But the risks to this write up most prescient:

• The ceasefire collapses. It’s a two-week deal. If Strait transit fees become a sticking point or there’s a military escalation, oil rockets back above $110 and the macro headwind intensifies rather than fades.

• Friday’s CPI shocks. March CPI is reporting Friday, expected at 3.1% headline - fully reflecting the oil spike. Even with oil falling now, a hot number could cook the Fed. Pushing the “Fed is done” narrative regardless of the ceasefire.

  • Of course, this is a short-term risk in this sense - because a '“hot” CPI print only further cheapens the entry point for good SaaS names.

• The AI disruption is worse than I think. Of course I could be wrong. AI could eat everything software.. it’s possible the structural repricing is just the start of a long decline. That Q1 earnings reveal deteriorating fundamentals even in names I’d consider resilient. If that’s the case, the “double discount” is really just a single discount that isn’t deep enough…

Sizing

Of course there’s always risks. But this is how I’m placing my chips. Overall it’s always, aim for 51% right and position sizing. Strong opinions loosely held.

I’m sizing accordingly - not all-in, keeping reserves, watching CPI closely before fully committing.

1. The SaaSpocalypse is overall accurate: AI platforms and AI technologies certainly impact many SaaS companies’ business models.

2. But imprecise - winners are also being re-priced, and in some cases mispriced. Investors are “throwing the baby out with the bathwater”.

   • These same high quality assets have historically been very richly valued. This type of value dislocation is rare, and this is the holy grail of generating alpha.

3. Evaluating winners? AI impact is a double-edged sword. Yes, many software companies are facing a tectonic shift to their business fundamentals and risk being decimated. However, others are less exposed, and may in fact benefit from these new developments. Assess these, and use this window of time to buy up (while the markets are hammered).

Of important note: they may certainly continue to drop. If you build high enough conviction, that just means they’re on steeper sale, and so… buy even more.

So, my view is not “buy up all software because it’s down”. Nor am I “calling the bottom”. It’s a narrower argument, that more resilient businesses are being unfavorably mispriced, and those specific ones are worth buying.

Subscribe now

Understanding the SaaSpocalypse

Simply, yes. Yes, AI structural forces are absolutely valid. The exact characteristic that made SaaS so richly valued, compared to any other industry, is being challenged - stickiness. This is through many vectors, summarily moat disruption, weakening pricing power, retention at risk, competitive threats (from AI players, from new competition, and from build vs. buy).

“Some of the software companies being disrupted by AI are facing very warranted decreases in their valuations.”

— Orlando Bravo, Co-Founder, Thoma Bravo (March 2026)

More precisely, these structural forces are:

• Disruption of Moats: AI agents can abstract away interface lock in. Many moats (incl. data moats, and systems of record) are at risk of disruption.

• Less Seats: Per seat pricing is being threatened - as headcount reductions eliminate seats.

• ^ = NRR at Risk: Net revenue retention (a key metric of SaaS health) is being hammered on both fronts - interface lock-in (thus retention) + expansion (expanding seat counts)

• Build beats buy? The ease of build your own (a la the buy vs. build decision) has improved, putting pressure on pricing.

• Flattening the playing field: Competition is on the rise - both (i) directly from foundational AI companies (e.g., Anthropic) as well as (ii) indirectly from ‘vibe coding’ allowing competitors to roll out competing features quickly.

Simply, this means that the premium value the market assigned to software is being put into question. Traditional hedge fund investors will call this their TV re-mark (terminal value, the value of the long-term of the company being haircut). The result:

• IGV (Software index) down -30% from recent highs

• ~$2 trillion of SaaS value decimated

• EV/Sales multiples compressed to 4.2x (vs. ~6-8x historically, and much higher at times)

Bessemer’s emerging cloud index, does a great job highlighting specifically enterprise (B2B) SaaS performance. More here.

But.. markets do tend to throw the baby out with the bathwater

Even the most savvy institutional investors are subject to short-term performance dynamics to their investors (LPs). This means. When the markets see an industry re-rating is ongoing (the multiple at which an industry trades being re-evaluated by the market due to broader forces). When every name in a sector is bleeding. When the spectrum flips from greed-driven, to fear-driven. This creates too much risk for short-term institutional investors to ride out. And certainly, most retail investors jump ship as well.

But, per Benjamin Graham’s (Warren Buffett’s teacher) famous maxim: “In the short run, the market is a voting machine, but in the long run, it is a weighing machine.” If you can keep your emotions aside and build conviction in the fundamental value of the business you’re buying, and you’re right, the stock prices will converge to their true value. When they deviate, we call this value dislocation. And this is the time to buy.

When oil prices briefly turned negative in spring of 2020, every upstream driller got hammered. Why? Thematically because they had high debt, and sustained low oil prices meant insolvency, and would lead to restructuring and wiping of equity value. But not all upstream drillers were equally at risk. Take Pioneer Natural Resources (PXD), it dropped -65-70% from ~$155/sh to ~$55/sh. Ultimately, when the dust cleared, its share price soared to >$240/sh. (and eventually got acquired by Exxon for even more). @$240, this would be nearly a +4x increase in just 2 years. Had one been discerning enough to look at their clean balance sheet, and solvency, the gains were surreal.

I believe today a similar dynamic exists in SaaS land.

Framework for finding the diamonds

The key questions are as follows:

1. Impact of AI: Has the fundamental business truly been impacted? Or is it resilient? Or perhaps even advantageous?

2. Quality: Is the company naturally a good asset in its category? Is it a product leader, a category creator? Does it have a good financial track record? High growth? Sticky revenues, good gross margins, healthy unit economics?

3. Management: Does management have a good track record of stewardship - can we trust management to navigate successfully through this choppiness? Excellent management tends to take advantage in such situations.

4. Valuation: Is the pricing reasonable compared to the intrinsic valuation?

Ultimately, the goal here is to find high quality assets at a discount. Not mediocre assets for cheap.

Questions #2-#4 are standard and relatively easy. Much literature exists on software asset quality historically, and how to evaluate good management. Valuations are an important point to always consider. Question #1 is the prescient new question - how is this company impacted by AI and the new paradigm?

Some considerations, certainly not comprehensive:

Company characteristics

• Pricing Power “Toll Booths” - some SaaS derives pricing power from clean interfaces (this is no longer going to hold). Others do it via data they hold (this is also being hammered). The third category is “toll booths” -or- “system of action not system of record”: these are the characteristics that benefit.

• AI-native or AI-wrapper? A thin wrapper (an LLM powered chatbot) and bolt on AI is hard to defend. Others are deeply AI-native (modern data pipelines, re-training models on proprietary data sets) - these are the types of companies that will successfully take advantage of AI. It’s a hard decision, because it requires significant R&D investment.

• AI Beneficiaries - take Netflix. They’re a content business with a taste profile and IP moat. I doubt AI means Claude will replace Netflix. In fact, AI lowers their production cost (they’re a beneficiary). Yet, they saw their stock plummet ~40% (its since rebounded in the last month so probably too late on this one).

• Picks & Shovels - the AI infrastructure and observability companies. As enterprises deploy thousands of autonomous agents, the need to monitor, secure, and store the ‘thought trails’ of those agents creates a massive category of spend.

More resilient sub-industries within SaaS

• Cybersecurity - AI expands the attack surface faster than enterprises can defend it, which means the demand for cyber would grow. Naturally, some companies in the industry will fall behind and collapse, so not all cyber. But I’d look for those that are innovating for AI application endpoint detection.

• Data infrastructure - every enterprise AI initiative requires a strong data layer.

• Regulated Industries - this moat is one that is resilient beyond data or network moats. Healthcare, insurance, finance, and government facing industries tend to have a deeper resilience. Some examples include financial infrastructure or EHR systems.

I’ll refrain from listing any specific names. The intent of this literature is to share frameworks, and share how I think in terms of a long-term mindset. This is not a stock picking newsletter.

A footnote on position sizing

Importantly, it’s worth keeping in mind that the market could very well continue to drop (and just as well could recover tomorrow). Really, it doesn’t matter. But, as far as sizing positions go, I don’t buy up my entire target position from the get go. I keep ‘reserves’ in store to buy up more if the prices drop. Balanced with an initial investment sufficiently large that if the stock doesn’t drop any further, I’m comfortable with the exposure (and feel I’ve bought enough).

“Be fearful when others are greedy, and greedy when others are fearful.”

- Warren Buffett

TL;DR

The S&P has dropped -7% from it’s all-time high just two months ago, on the backs of an escalating Iran conflict, driving an oil shock, and followed up by increasing Fed hawkishness. What’s next?

In here I’ll go deeper into the forces at play and why I’m expecting further decline in the market before it bottoms out. Driven by the magnitude of this oil shock compared to history. Yet, if the market does indeed drop further, how this mounts pressure on the US to find a stable resolution. And finally what I’m doing with all this information.

For me:

1. Stay Steady. I’m buying, building positions in equities I like, and looking for more high quality assets that are available at a discount.

2. Short Overall. A temporary position, but on the whole, I’ve turned net short on the market. For now.. Sharing my thinking in detail below. Monitoring for quick reversal (if favorable war resolution or oil relief).

Subscribe now

Setting the stage

Where we are today

A time of increasing volatility and situational uncertainty

Before all this. We entered the year with the market hitting all-time highs (the S&P 500 surpassed >$7,000 on Jan. 27th, up +16% YoY), and richly valued (with a ~40x Shiller CAPE ratio. Fueled by AI excitement, strong corporate earnings, resilient consumer spending, expectation of an increasingly dovish fed, and underpinned by overall exuberance as investors chased returns in high growth AI / big tech equities. This put us in a starting position of a market priced to perfection, with no margin of safety.

Since which.. we’ve seen two fears grip the market. Initially, (i) an AI-triggered SaaSpocapylse, and since then (ii) the Iran War - which has led to high alert about geopolitical volatility (obviously), brought about an oil shock, resurged inflationary fears, and resulted in the Fed taking a more hawkish stance (lowering expectation of future rate cuts).

This means that in less than 2 months, the S&P 500 has shaved off -7% of it’s value and overall the market has wiped out ~$5-$6 trillion in value. Though still richly valued at a 35x Shiller CAPE Ratio (for reference, the trailing 10 year average is 32x, and 20 year average is 27x).

A quick overview of the forces in play, which have been pressuring the market to wipe it’s gains.

Tldr: frothy prices → Iran War → oil shock → inflation resurgence → hawkish Fed

• The Iran War dragging on: Initially with a threat of strikes on Iran, perceived to be a negotiating tactic by Trump, this has escalated significantly. From ‘unlikely’ strikes → to a joint US-Israel strikes posited as a ‘72 hour surgical operation’’ → to now, a war that’s dragging on, 3 weeks in already, and bringing other countries into the mix, with no certainty on resolution. World War 3? Or will it resolve by Tuesday.

• Oil Shock: The primary pressure point Iran has targeted is oil and gas. Effectively closing the Strait of Hormuz (through which ~20% of global oil transmits), daily oil transit through the Strait of Hormuz has dropped -87%. Israel struck Iran’s processing plants at South Pars. Iran struck energy targets across Kuwait, Saudi, Qatar, Bahrain, and the UAE, as well as Hormuz bypass routes at Oman’s ports. As of Friday, Brent oil prices reached $112/bbl, up over 70% since the war began.

• Hawkish Fed: On Mar. 18th, the Fed held rates steady at 3.5% - 3.75% for the second straight meeting. It raised inflation expectations to 2.7% (vs. 2.4% previously), and touched on a weakening job market elevating broader economic risk. The Fed officials also shifted rate cuts outlook (the dot plot) hawkishly, with now 7 officials seeing no rate cuts this year (vs. 6 members before), and another 4 or 5 officials lowered their moved from projecting two cuts to just one.

.

My positioning

Stay steady is my core advice

“Be fearful when others are greedy, and greedy when others are fearful.”

- Warren Buffett

As a reminder, I always consider a long-term view (40+ year holding period), and find periods of market uncertainty as an opportunity to buy high-quality assets at a discount.

Investors who build real, compounding wealth do not do so by selling into drawdowns. They do so by staying steady through periods of volatility and fear. I’m buying businesses I like. And if the market continues to drop, I’ll buy even more.

Specifically, high-quality businesses with durable competitive advantages. In the past year, everything’s felt too expensive. Now, some of these businesses have potentially been materially impacted, but many are trading at discounted valuations while underlying fundamentals hold strong. Particularly, I’m looking for undervalued businesses in these three sectors:

• Software

• Airlines

• Consumer Discretionary

For the long-duration, the playbook is clear: stay steady, stay discerning, and use fear-driven dislocations to improve the quality and cost basis of the portfolio.

..yet, I myself have taken on a large short position

Despite the fact I don’t suggest timing the market, I’m transparently doing the opposite. I’ve been building a short position (as a hedge against my investments) over the last few months. And within 7 minutes of the Fed announcement on Thursday, ramped up that short position to a degree that’s ~120% of the value of my portfolio. Meaning my net exposure is -20%.

Notably, this is a hedge, I’m still holding (and building up) my core positions.

I’ve been assessing how long to hold that short offset, and whether I’d consider increasing my position (taking a heavier short). As such I’ve laid out my framework of thinking here.

The possible scenarios and probabilities are laid out in the next section. Summarily, my assumptions lead to a probability weighted framework pointing to a weighted average trough that’s -8% to -14% below current prices (down to <$6,000). I believe the US gov’t is naturally pressured to turn this around and alleviate market pressure (“intervention offset”), and expect this to really ramp up at a level of ~$6,000-$6,200 (S&P price level).

While holding onto my short position, I’m closely watching two triggers - war resolution, or oil price relief. If either show optimistic signs, I’ll likely immediately close out my short position.

This was written over the weekend.. perhaps the reversal is already underway given this morning’s announcements of Trump delaying his threat on Iran’s power plants.

.

Pricing this trade

01 | Scenarios & Probabilities

I believe the market still has more room to drop before it bottoms out

The framework I’ve applied here is as follows:

1. What are the possible scenarios?

2. What is the expected trough for each scenario? (e.g., how much further might the market drop?)

3. Assign a probability to each scenario

I’ll illustrate the paths as the following four scenarios. What’s actually happening is a much more complex chessboard, so this is a reductive view of some possibilities.

• Swift Resolution: Ceasefire in days. Hormuz reopens (still gradual). Oil prices come back down to $70/bbl. Fed focuses on combatting job market and increases rate cuts.

• Constrained Grind: War persists, but as background noise. Oil sees relief (whether with the Strait of Hormuz opening, or alternative supply routes, tapping deeper into strategic reserves, etc.). Growth may slow but we avoid recession territory.

  • Base Case: If I had to bet on one path, I’d expect this and then the intervention offset amplifying when the S&P falls into ~$6,000 - $6,200 territory.

• Prolonged Escalation: A deeper decline before recovery. Ground troops are deployed. The war persists through Q2, Hormuz stays shut, and oil stays >$100/bbl. Perhaps CPI inflation re-accelerates, hiring stalls, or consumer confidence craters. Stagflation may enter the chat..

• Full Crisis: World War 3 and/or an oil crisis more parallel to the 1973 OPEC Embargo. Strait of Hormuz stays shut and potentially major Gulf production infrastructure destroyed. Oil stays >$150/bbl for an extended period. Fear mongering case, and naturally would cause a lot more fear, uncertainty and doubt.

So in 3 of the 4 scenarios, which I believe are cumulative 85% probability, the S&P drops to below $6,200 (another -5% slide vs. today). So I believe there’s more room to fall before a resurgence.

Notes to this framework:

• Assumptions matter: these are my weights based on my readings. Yours may differ and clearly would imply very different outcomes. Form your own view.

• Reality is deterministic not probabilistic: Aka the future is not a blend of these 4 scenarios, it’s going to be one path. Don’t let spreadsheet math lead to overconfidence.

.

Just some observations

02 | Iran, oil, and the Fed

• On Iran: First, Iran cannot sustain a closed Strait indefinitely. 75% of the region’s oil exports and 59% of LNG go to Asia. Meaning that while the closure is aimed at the West, Asian countries are also catching strays. As the Strait stays closed, diplomatic pressure from Beijing and New Delhi will intensify. Second, the physical infrastructure damage to Iran’s upstream oil & gas (incl. the South Pars gas field) is degrading Iran’s positioning. Nonetheless, Iran has a deeper pain tolerance than the west, where the US faces steeper immediate pressures, with more to lose, heavier impact from the oil shock, and upcoming mid-term elections.

• On oil: While Brent at $112 reflects the war premium, the physical market is even tighter. Dubai and Oman crude are both >$150, which may represent the real clearing price short term for barrels in Asia. The current price levels persist in spite of the IEA reserve release (400 million barrels) and the US SPR drawdown (172 million barrels). The underlying disruption is massive, with Hormuz daily oil transit down 87% from 20 MMbpd (million barrels per day) to 2.5 MMbpd. Saudi officials have warned crude could test $180 if disruptions persist through late April.

• On the Fed: In this recent Fed meeting, Powell said risks to employment “are to the downside, which would call for lower rates,” while risks to inflation “are to the upside, which would call for higher rates or not cutting anyway.” The implication being the Fed is taking a more hawkish stance, and focused on combatting inflation first. If relief comes, it’ll more likely be led from the geopolitical side (war ending / oil relief), not the Fed.

.

The drastic comparison

03 | Gravity of this oil shock

This is THE steepest oil shock in history. Neither oil prices, nor market reactions reflect the magnitude yet

The data speaks for itself:

This is a far more drastic supply disruption than we’ve seen in history, yet the oil price shock and impact to the market thus far is relatively mild by comparison. Likely a combination of some expectation this is temporary, as well as mitigating factors (US energy independence, and strategic reserves).

Other characteristics worth noting on both 1973 and 1990, regarding impacts to the market:

• 1973: the market dropped -48% following the 1973 OPEC Embargo. The US economy was overvalued coming into it (“Nifty Fifty” market). Inflation had already been a persistent issue in the preceding years. The oil embargo spiked oil prices, triggered runaway inflation, and a deep stagflationary recession

• 1990: the market dropped -20% following the Gulf War. The US economy was already fragile, with a savings and loan crisis weighing on the financial sector. The Gulf War was a catalyst, spiking oil prices and tipping the US economy into a mild recession.

Let’s consider the 1973 OPEC Embargo

The most drastic oil shock is 1973. Here, the OPEC members embargoed oil exports to the US and its allies. Proving Middle East conflicts could weaponize energy dependence. The result was drastic: oil prices quadrupled, inflation surged, and we experienced a recession with the S&P losing -48% of its value over the following 9 months.

Of course, we’ve learned since then, and there have been many other oil shocks subsequently which have had might light impacts on the market.

What’s different?

• US energy independence: US is now a net exporter of oil. In 1973, the US imported ~35% of its oil. Today, domestic production exceeds consumption. Relative insulation of the American market can be evidenced by the fact that WTI ($98/bbl) trades at a $14 discount to Brent ($112).

• Strategic Reserves: Strategic reserves have been established, and are being deployed, dampening the impact. The SPR did not exist in 1973, it was created in direct response to that crisis.

While a 1973-style recession is certainly unlikely given these structural changes, it’s always worth studying history; and it’d be naive to say there aren’t broader structural parallels (inflation, economic fragility, on the backs of a market that’s priced to perfection).

.

The optimistic offset

04 | Intervention Offset

It’s important to consider that the more the markets are impacted, the more pressure is going to mount on US leadership to come to a stable resolution

“Preferences are optional and subject to constraints, whereas constraints are neither optional nor subject to preferences”

- Marko Papic, Geopolitical Alpha

This is a framework I love. It comes from Marko Papic’s Geopolitical Alpha:

The crux of the idea is to assess what politicians can do more than what they want to do. Material constraints (economic, political, legal, demographic) dictate outcomes because they are more predictable and more binding than preferences.

The “intervention offset” considers the level of economic or market pain at which political constraints pressure action, regardless of stated goals. We saw it in April 2025 with “Liberation Day”: when tariffs drove the S&P down ~19%, the administration reversed course within weeks. The constraint was the market decline itself, amplified by congressional anxiety and corporate lobbying. This constraint has oft led to the phenomenon of the TACO trade.

Some constraints to consider:

Considering Trump administration intervention offsets in the past (especially Liberation Day, as well as other ‘TACO trades’) I’d estimate the intervention offset really mounts at ~$6,000 - $6,200 (S&P 500) / ~10-15% down from peak. At this level, expect some combination of: accelerated diplomatic efforts, expanded SPR releases, heavy search for alternative oil, or much heavier direct action to re-open the Strait. This is the structural reality of how American presidents respond to economic pain.

“President Donald Trump put his threat to destroy Iranian power plants on hold for five days after what he described as ‘productive’ talks”
— Bloomberg, 3/23/26

We’re already seeing some of it play out, with Trump starting to walk back aggressiveness this morning.

.

What do we do with all this?

My advice:

• Don’t panic. Selling into fear is the single most reliable way to destroy long-term compounding.

• Stay discerning. Use this volatility to your advantage. High-quality assets are on sale. Software companies, airlines and consumer discretionary names have all been hammered. For patient capital, this is where the best entries are made.

What I myself am doing:

• Short-term, net short, monitoring closely. I’m betting on further downside, and hedged accordingly. Monitoring closely for reversal (war resolution and/or oil relief).

• Buying and evaluating good buys. Building positions in my favorite existing names. Searching for more discounted assets, particularly in software, airlines and consumer discretionary. High quality businesses, with durable advantages allowing them to weather this storm and come out stronger on the other side this storm.

• Oil assets. Of note, I’ve ~30% of my portfolio in oil & gas assets already. Holding these, although may trim some if prices get exuberantly overvalued.

As always, sharing my assessments as food for thought. Good investing is not letting emotions take control of the situation, and good alpha comes from being right 6 of 10 times, fused with good position sizing, and staying nimble. No one can say if the market will go up, down, sideways, or start doing some loopdy loops. But for now, this is how I’m playing it..

.

Educational opinions only, not investment advice. No perspectives here on where the markets will go. Just food for thought

Subscribe now

My main position I’ve built up on this past week is Nvidia. It’s an existing position for some time, and sharing some thoughts on why I’m more excited now.

While Nvidia has lagged other chip peers (e.g., TSMC, AMD, Intel) and AI players (e.g., Google), I believe the same dynamics investors are penalizing it for (China loss) is a large advantage for it (Western AI monopolization).

In very short, I’m buying up on NVDA (and TSMC) because:

• Nvidia has replaced its China business with higher quality revenue from new sovereign buyers.

• AI demand is exploding, which is monopolized by Nvidia (with 90%+ share).

• The market has weighted a discount on Nvidia due to Taiwan risk, which is being mitigated by TSMC’s Arizona expansion.

• but, re-evaluate in Oct/Nov if China rare earths risk spikes

Subscribe now

Nvidia’s synonymous with AI

If AI is the new electricity, then Nvidia is the power plant

AI workloads (ChatGPT, Claude) are run nearly exclusively on Nvidia chips. (Some exceptions include TPUs and Gemini, which is why I’m also long GOOGL). Nvidia’s dominance of the market is stunning - over 90% of AI computing in the Western world runs on Nvidia’s hardware. And the software that developers use to program these chips (a platform called CUDA) is Nvidia’s proprietary system.

01 | Nvidia’s quietly broken up with China. The upcoming Xi <> Trump summit will make it official. And that’s a good thing for Nvidia

Some history

China used to be about a fifth of Nvidia’s data center business. But over the last two years, the U.S. government kept tightening rules on what chips could be sold there. Based in a concern that China could use powerful AI chips for military purposes.

Nvidia kept trying to thread the needle by designing weaker versions of its chips that might satisfy Washington while still being useful enough for Chinese buyers. In December, Trump approved the latest attempt: Nvidia could sell a mid-range chip (H2000) to China, but with a 25% tax and extensive security reviews.

This fell flat. China’s government never approved the imports on their end. Probably because they didn’t want to depend on a supply line that America could cut off whenever it felt like it. A reasonable calculation.

In February, Nvidia’s CFO (Colette Kress) disclosed that Nvidia had sold zero chips to China. By early March, Nvidia stopped making China-bound chips entirely and redirected that factory capacity toward its next-generation product, where Western customers are lining up out the door.

Now, the summit: Trump is set to meet with Xi Jinping in Beijing on Mar. 31 for 3 days. Notably this is a first for a US President since his last term (in 2017). Both Trump and Xi’s leverage has significantly weakened.

For Trump. The Supreme Court struck down his tariffs in February. The Iran war is consuming significant attention. For Xi, the Iran War and Hormuz disruption is hitting China hard, as half of China’s oil comes from the Middle East.

My prediction is that beyond a few headline-friendly commercial deals (agricultural purchases, maybe energy commitments)), Xi will not get meaningful relief on semiconductor export control - meaning the chip split holds and really strengthens.

This solidifies Nvidia’s loss of China revenue, and I think that’s a good thing due to Western AI monopolization (which I’ll touch on later here). China was a headache for Nvidia anyhow. Every quarter brought new uncertainty: would the rules change? Would there be another write-down? Nvidia was selling weakened chips at thinner margins into a country that was simultaneously trying to build its own competing chips. That’s like supplying your future competitor with a slightly worse version of your product while Washington randomly changed the terms every few months.

That headache is over. The factory capacity that used to make China chips? It’s now making chips for customers who are actually buying - at higher prices, with simpler logistics, and without political risk.

02 | AI is expected to get 10x cheaper to run - which will explode demand for it

This is a counterintuitive element in Nvidia’s favor. While the cost of running AI is expected to improve by ~10x. This does NOT mean Nvidia’s revenue decreases by 90%, it means that AI demand explodes and is a further tailwind.

Historically, we’ve seen the opposite - when a resource gets dramatically cheaper, people use vastly more of it.

• When cloud computing made server costs drop by 10x in the 2010s, companies didn't just run the same workloads for less. They ran 100x more workloads. Amazon's cloud business didn't shrink, it became a $100 billion juggernaut.

• When the cost of a phone call dropped by 10x, people didn't make the same number of calls. They started calling constantly, then texting, then video calling, then streaming.

• The same thing is about to happen with AI. At today’s prices, a company might run one AI assistant as a pilot project. At 1/10th the price, that same company deploys hundreds of AI workers running around the clock. The total amount of computing consumed goes through the roof.

And we can see this demand building right now. The companies that buy Nvidia’s chips are growing at rates that have almost no historical precedent. Anthropic (maker of the AI assistant Claude) grew from $1Bn in ARR to $14Bn YoY. Open AI ended 2025 at $20Bn in ARR. These companies run almost entirely on Nvidia hardware. Their explosive growth is Nvidia’s demand pipeline.

The infrastructure being built to serve them is staggering. Microsoft, Google, Amazon, Meta, and Oracle have collectively committed to spending $660–$690 billion on AI data centers in 2026 alone, +2x year-over-year. A joint venture called Stargate (backed by SoftBank, OpenAI, and Oracle) is investing $500 billion to build AI data centers across the United States.

The vast majority of this hardware spending goes to Nvidia. And all the software is built on Nvidia’s proprietary platform (CUDA), which creates an incredibly strong moat.

03 | New buyers are replacing China, and at better prices

Think beyond basic analyst models. Basic models assume that if China’s out, that’s lost revenue. However, in Nvidia’s case, there is an exceeding amount broader demand replacing it.

Due to national security and technological independence reasons, governments around the world are building their own AI computing facilities. Take a look at Saudi Arabia. It’s sovereign wealth fund (PIF, with $930Bn in AUM) created a company called HUMAIN specifically to massive AI data centers, all powered by Nvidia chips. The UAE as well, building the largest AI computing campus outside the US - a 5-gigawatt facility (for scale, that’s roughly enough electricity to power 4 million homes). The U.S. government cleared 70,000 of Nvidia’s most advanced chips for export to Gulf states last November (vs. to China, it was only the old models). Japan, France, and Germany are doing similar things…

The result is that China demand is being replaced by many other geos. Those buyers are buying Nvidia’s best chips, at full price. No tariff or regulatory risk. Dollar for dollar, feeding cleaner, higher-margin, stickier revenue than China ever was.

04 | Lastly, worth flagging TSMC diversification

A large supply chain risk is Taiwan. TSMC is the manufacturing bottleneck, with 72% market share of global foundry revenue, and estimated 90%+ for cutting-edge AI chips. Given this was all done entirely in Taiwan historically, and the geopolitical tension as well as US <> China fragmentation underway (key underpin of this thesis), this exposure poses a pretty big risk.

However, this is being significantly de-risked, with TSMC factories in Arizona being built, and far ahead of schedule. The first Arizona factory is already producing Nvidia chips on US soil today.

In total, TSMC has committed $165 billion to build 6 factories and a research center in Phoenix. When complete, expecting Arizona facilities to produce ~30% of TSMC’s most advanced chips.

The "Taiwan risk discount" that investors apply to Nvidia's stock is starting to erode, and it'll keep eroding every quarter that Arizona factories prove they can deliver.

05 | Key Risk: rare earths, and expiry of the Busan truce

Rare earths are a key input and a key risk to watch.

The Busan truce, the one-year pause on tech escalation that Trump and Xi agreed to last October, expires in November 2026. As part of that truce, China suspended a sweeping set of export controls on rare earth elements. These are the specialized minerals used in everything from chip manufacturing to fighter jets to electric vehicle motors. China controls roughly 60% of global rare earth mining and over 90% of refining. There is no near-term alternative.

When the truce expires, those controls snap back on, unless both sides agree to extend them again. And who knows if they will..

This doesn't break the thesis. But it means the thesis has a clock on it. The window between now and November is the window where the repricing should happen. After November, when the Busan truce expires, it’s worth considering closely as the risk profile changes.

Educational opinions only, not investment advice. No perspectives here on where the markets will go. Just food for thought

Everyone's reading this as an oil shock, but we need to go deeper

Preamble: As I posited two weeks ago in this article, oil prices are spiking (although dropped back below <$90/bbl today, still up from ~$74/bbl). My focus with these writings is not to emphasize the obvious (Iran war is prolonged, gas is impacted, jobs reports is weak etc), rather to draw attention to the needle-moving currents we tend to overlook. So today I’m sharing analysis around short-term reversal of de-dollarization underpinned by an LNG trap, how I expect this to continue, and how there’s a higher than priced in change of US HH LNG prices spiking as well.

LNG = natural gas (stands for “liquefied natural gas”)

Today’s thesis, TL;DR:

Europe’s summer LNG refill race creates a structural multi-month dollar bid.

1. The dollar reversal is structural, not reflexive, because of LNG market mechanics

2. The LNG constraint creates a specific, quantifiable, time-bounded dollar bid lasting through October

3. The same dynamic that strengthens the dollar short-term (stagflation) eventually undermines it, creating a trade with a clear exit condition

While in most of 2025, the dollar lost ground, the Iran war has reverse this heavily - with the Strait of Hormuz effectively closed and Qatar shutting its main LNG terminal. The dollar has strengthened rapidly, up +1.3% in a week, and with room to run.

The market understands immediate mechanics: energy is priced in dollars, a geopolitical shock sends capital fleeing to safe assets, rate differentials widen. That’s all true. What the market is missing is why this dollar reversal is structural, not a knee-jerk spike.

The reason is LNG (liquid natural gas). Underneath the oil shock that everyone is tracking is a natural gas supply disruption 50% larger than the 2022 Russian crisis - landing hard on Europe. Europe doesn’t have the capacity to quickly find alternative supply, cannot politically reverse its Russian import ban, and enters the critical summer storage-refill window, which is already running 15 percentage points below the seasonal norm. Oil markets have release valves: strategic reserves, OPEC flexibility, US production flex. LNG has essentially none. LNG cannot be rerouted by pipeline. New export terminals take 5-10 years to build. The result is that Europe will be competing for every available LNG cargo through October - requiring conversion from euros to dollars every single time. That’s not a one-week dollar trade. It’s a multi-month structural bid.

A secondary thesis is a downstream observation: US natural gas (Henry Hub) has barely moved while European prices surged 80%. That growing spread is pushing US LNG exports to maximum capacity. If that sustains, it’ll eventually feed back into domestic US prices. I’d put odds of US HH LNG breaking >$6/MMbtu (vs. $3.62 today) at 25%.

Lastly, a noteworthy timing constraint to consider alongside all of this: the same dynamic that strengthens the dollar short-term, is what weakens the currency long-term. This is stagflation. This dynamic today is energy-driven inflation colliding with a softening labor market - the textbook definition of stagflation. The mechanical dollar bid wins the next several months. The stagflationary drag eventually works against it. Thus, this is a trade with a clear short-term window, not an indefinite position.

Subscribe now

Briefly framing historical de-dollarization

De-dollarization has been underway with particular acceleration particularly in 2025 (with the USD weakened by ~6.5%). From another perspective, the dollar accounted for 71% of global reserves in 2000, which has fallen to 59% by 2025. This is due to a litany of factors including US’s weaponization of the dollar by seizing $300Bn of Russia’s foreign reserves in the midst of the Ukraine invasion; erratic US economic policy; sanctions; geopolitical fragmentation and attacks on US hegemony esp. BRICS; growth of rival economies (China & India); advent of alternatives to US’s SWIFT (China’s CIPS & India’s UPI). These trends long-term create global pressures to drive de-dollarization, and I do expect this de-dollarization to continue course over the longer arc. With the aforementioned stagflationary pressure only compounding.

What Actually Happened Last Week

01 | An Oil Shock That Hides a Bigger Gas Story

On March 3rd, Iran closed the Strait of Hormuz to commercial traffic. A fifth of the world’s daily oil and LNG passes through that 21-mile chokepoint. Cargo insurers pulled war-risk coverage almost immediately - without insurance, tankers don’t move. Persian Gulf oil exports reportedly collapsed from 22 million barrels per day to under 3 million almost overnight.

Then Qatar, which supplies roughly 16-17% (25 million tonnes per annum, or mtpa) of Europe’s LNG, shut its export terminal after drone strikes. European natural gas prices spiked 80% in a single week. Stock markets sold off hard. And crucially, bonds sold off too (the 10-year Treasury yield jumped 20 basis points to 4.15%). The labor market just shed 92,000 jobs in February. Socks down, yields up, is the fingerprint of a stagflationary shock.

The dollar gained 1.3% on the week. To understand why that move is likely just the beginning, you need to look at where the dollar came from, and what’s now underneath it.

The dollar index (DXY) peaked at ~108.5 post-election in Nov’24, then weakened steadily through 2025 - down ~6.5% to a trough around 102 by Feb’26. Last week’s Iran shock reversed that trend in five trading days. The question is where it holds and extends

The picture across these three charts tells one story. Oil is up sharply (the market has frameworks for that). Gas is up catastrophically in Europe, and it doesn’t have the same frameworks, because LNG structurally can’t behave like oil. And the dollar, having drifted lower for 12 months, has just reversed on a shock whose effects will run much longer than a typical geopolitical event. That’s the disagreement with consensus worth sizing.

Why the Dollar Reversal Has Legs

02 | Oil Markets Have Release Valves. LNG Has Almost None

To understand why the dollar move is structural rather than transient, it’s important to understand something about how LNG actually works. It’s genuinely different from oil, and those differences matter enormously right now.

When oil prices spike, there are release valves. The US and other IEA members can release from strategic reserves. OPEC can increase quotas. US shale producers can bring more rigs online. None of these work instantly, but they exist, the market knows they exist, and it discounts the spike accordingly. That’s why “how long until supply normalizes?” is the right analytical frame for an oil shock. LNG doesn’t work that way. Even fast-tracked projects take 4-6 years.

Right now, global spare liquefaction capacity is effectively zero - current global liquefaction capacity is roughly ~700 mtpa. Current utilization and are running at >95%.

The US, the world’s largest LNG exporter, is already running near full capacity. When Qatar shuts, there is no dial to turn up elsewhere. The shortfall gets absorbed through price alone, for as long as the disruption lasts.

The 2022 Russian gas cutoff is the reference point everyone uses. It nearly broke European industry. It triggered emergency demand rationing across the continent. It forced governments to sign decade-long LNG contracts in a matter of weeks and consumed two years of political capital. That disruption was 80 billion cubic meters a year. The current disruption threatens 120 billion. +50% larger. And crucially, unlike 2022, there are no unused alternative supplies waiting in the wings, those contracts were already signed to cover the Russian gap, and that capacity is fully spoken for.

Now layer in Europe’s starting position for this crisis.

The problem is the refill race into next winter. Europe needs to get from roughly 30% to roughly 90% storage by November - starting 15 points behind, with the largest LNG supply disruption in modern market history, and zero spare capacity to draw from. Every cargo Europe secures this summer gets paid for in dollars. That bid doesn’t stop in a week.

Russia is the obvious escape route, yet it’s politically sealed on both sides. The EU had already scheduled a ban on Russian LNG beginning in April. American LNG explicitly replaced Russian supply after 2022. Washington would view any reversal as a direct breach of solidarity while Europe simultaneously depends on US security guarantees. Europe cannot un-brick that door without fracturing both its energy policy framework and its alliance credibility.

A dollar bid from an oil shock fades when the oil shock fades. A dollar bid from a summer-long LNG refill race, where Europe competes for every cargo with no alternative supply, keeps refreshing every week until October.

How the Dollar Strengthening Plays Out

03 | How the Dollar Strengthening Plays Out

Mechanics

The dollar was already recovering from its 12-month drift lower. The Iran shock added the catalyst. What makes the move durable rather than a reflex is the plumbing underneath it. Specifically, who has to buy dollars, how much, and for how long.

The first two bids are already running and don’t require any particular decision from the ECB or anyone else. They operate through pure mechanics.

Europe needs to refill from ~30% to ~90% storage by November. European storage capacity is approximately 1,100 TWh (roughly 100 BCM equivalent). Getting from 30% to 90% means filling ~660 TWh. At current TTF prices of €67/MWh, that’s approximately €44 billion of natural gas purchases over the refill season. A significant portion of that is dollar-denominated LNG procurement. Even if only 30-40% is dollar-settled spot LNG (vs. pre-existing euro/local-currency contracts), that’s ~€13-17 billion of forced EUR→USD conversion over six months.

The third is my speculative overlay: worth holding, worth watching. But it requires the ECB to make a choice that its own institutional history suggests it’s unlikely to make.

The Oil <> USD Gap Left Open

Another number worth considering: Barclays estimates a 10% rise in oil prices produces a +0.5-1.0% DXY gain through the rates channel. The market pricing in tighter Fed policy, wider real rate differentials, dollar-denominated energy settlement. That relationship is well-documented across multiple energy shock cycles. Apply it to the current setup: oil is up 36% in five trading days. On that math, the DXY should have moved somewhere between 1.8% and 3.6%. It moved 1.3%. There are two ways to read that undershoot.

That 0.5-2.3% gap is the position. Not a vague call that “the dollar has room to run,” but a mechanically derived estimate of what the rates channel hasn’t priced yet - because it takes weeks, not days, to fully transmit. The market is treating this as a geopolitical reflex that will fade. The rates channel is indifferent to the geopolitical narrative. It responds to the energy price, and the energy price hasn’t moved back.

The pessimistic read is that the undershoot reflects genuine market skepticism about duration - smart money discounting a rapid resolution. Additionally, the weak jobs print means the market is likely discounting for some economic decay. That’s a legitimate risk, and it’s addressed in the China valve section below. But if the LNG constraint holds for even half the summer, the rates channel catches up. The math does the rest.

The Secondary, Domestic Play

04 | What Happens If the US Feels It Too.

The Henry Hub chart tells an interesting sub-story. European gas is up 80% in a week. While US domestic gas is up modestly. It’s the same molecule, just not exported. That gap exists because US LNG export terminals are a physical bottleneck: you cannot ship more gas than the terminals can liquefy, regardless of how wide the price spread gets. Right now that cap is protecting American consumers. The question is how long it holds.

If the Gulf disruption sustains for months, US LNG terminals run at maximum capacity through the entire summer and fall. Gas that would normally replenish US storage goes to Europe instead. The US enters the 2026–27 winter heating season with thinner domestic reserves than usual - and at that point, the European pricing pressure starts to bleed through. US shale production can flex up, but it takes months, not weeks.

This is a real but lagged and conditional risk. The honest probability is roughly 25%. High enough to flag, not high enough to treat as load-bearing. The most likely trigger isn’t this spring but next winter’s storage season, and only if the disruption hasn’t resolved by then. The more immediate political consequence is softer: a sustained Henry Hub elevation above $5–6 creates the domestic noise that eventually pressures Washington toward export restrictions. This is the tail risk that would abruptly invert the dollar thesis, not the price spike itself.

Duration

05 | Stagflation: the driver strengthening the dollar short term, is the same driver that weakens it long term

It’s worth being direct about a tension in the thesis. Stagflation, rising inflation into a weakening economy, is initially dollar-positive through the rates channel. Inflation expectations rise, the Fed is seen as more likely to hold, real rate differentials widen in favor of the dollar. That’s the short-term dynamic, and it’s real.

But stagflation is ultimately corrosive to a currency. When an economy deteriorates (unemployment rises, consumer spending softens, and corporate earnings erode) capital eventually starts flowing away from that currency regardless of what the central bank does. The 1970s showed this: the dollar weakened structurally through much of that stagflationary decade even as inflation ran hot. The mechanical bid wins first. The economic decay wins later. Notably, today consumer spend and corporate earnings are still strong. Still, the dynamic today is energy-driven inflation colliding with a softening labor market - the textbook definition of stagflation.

Where’s the handoff? Roughly: as long as the Fed holds rates and the energy shock is being absorbed as an inflation story, the dollar stays strong. The inflection comes when the labor market deterioration becomes the dominant narrative. E.g., when the market stops pricing “inflation risk” and starts pricing “recession risk.” The jobs numbers have shown some softness: the US lost 92,000 jobs in February, the three-month average is 6,000, and the economy hasn’t added net jobs since last April. If that trend continues, the stagflationary drag on the dollar accelerates. The window for the strong-dollar thesis is likely measured in months, not years - which is precisely the window the LNG refill race fills.

Or put simply, monitor:

If you make this bet on the US dollar strengthening, monitor:

• Corporate earnings and consumer spend - if signals of market-acceptance of stagflation start hitting the print, or any real risk of recession occurs, it’s time to close out the trade.

• Similar for unemployment - if unemployment and/or job losses continue to rise.

What Could Break This

06 | China’s Release Valve & Other Risks

China as Shadow Release Valve: [Serious Primary Risk]

• This risk deserves serious treatment. The core assumption in the thesis is that global spare LNG capacity is effectively zero, no dial to turn up. Yet, China is a hidden dial.

• The possible mechanism: China receives Russian and Iranian energy via overland pipelines and non-sanctioned sea routes - supply that never touches dollar-settled markets. Meanwhile, China holds long-term LNG contracts with US and Qatari producers. If Chinese domestic industrial demand is soft (which it currently is), China can redirect those contracted cargoes to Europe at spot premiums, pocketing the difference. It becomes a black market: Russian/Iranian molecules flow into China from the east via yuan-settled deals; Western LNG molecules flow out of China to Europe via dollar-settled markets. Net effect: China functions as an off-balance-sheet liquefaction terminal that isn’t captured in any spare capacity statistic.

• In 2022–23, this is exactly what happened. China quietly arbitraged the crisis at scale. If it happens again, the “no spare capacity” pillar of the thesis softens materially. The dollar bid doesn’t disappear, as Europe still converts currency to buy gas. But it softens the price spike, and with it the urgency and duration of the bid. Watch Chinese industrial utilization as the leading indicator. If their economy is accelerating, China needs those cargoes itself and this valve closes. If growth stays sluggish, the valve opens and China becomes the market’s invisible release valve.

• Monitor Chinese industrial utilization. If it stays soft, the “structural” dollar bid might be much weaker than I expect.

US Restricts LNG Exports [Tail Risk]

• Also a very serious risk to monitor.

• Henry Hub’s relative calm is notable. But if US LNG exports run at maximum capacity long enough, domestic prices eventually feel it. With midterms approaching, the political pressure to restrict exports is real and bipartisan. Since restricting exports protects domestic heating bills. A US export curb would simultaneously collapse the dollar-energy thesis and leave Europe with no viable alternative. Not the base case. But it’s a genuine tail risk that would trigger abrupt reversal across the whole thesis.

Warm Spring Across Europe [Watch Factor]

• Weather is the swing factor that determines whether a structural constraint becomes an acute crisis or a manageable stress. A mild spring and early summer reduces demand, buys refill time, and compresses the urgency that drives price. 2022 showed this clearly: an unusually warm autumn gave Europe the breathing room it needed to avoid rationing. This is the least predictable input and the most powerful natural hedge.

The Longer Tail

06 | The Dollar Wins The Battle. Not the War (the decade)

There’s a separate, slower-moving dynamic worth keeping in the background. If this disruption accelerates yuan-denominated energy settlement at scale, it chips away at something structurally larger than any single crisis. This mechanism is China quietly brokering Russian and Iranian supply to energy-desperate buyers, priced in yuan.

Every Asian energy importer has just watched a US-dollar-settled Gulf supply chain disappear in a week. Some of them are drawing conclusions about concentration risk that won’t show up in FX screens this quarter but will compound over years.

The data on this is already in the rearview mirror. The dollar’s share of global foreign exchange reserves has fallen from 71% in 2000 to 59% today - a 12-point structural decline over 25 years that predates this crisis and will outlast it.

The dollar wins this battle, with high conviction. The question worth holding separately is what the accumulated weight of these episodes looks like over a decade of energy market geopolitics. Each crisis that routes supply through yuan-settled channels, each central bank that quietly diversifies reserves, each Asian importer that rethinks concentration risk → none of it matters in the next six months. All of it matters in the next ten years. Both things can be true at once. The investor who sees only the near-term trade, or only the long-term structural decline, will be surprised by whichever one they’re ignoring.

The dollar was losing ground. An energy shock reversed it. What the market is pricing as a temporary geopolitical reaction is actually a structural bid that refreshes every week Europe tries, and fails, to replace gas it cannot find.

My Trading Strategies

Implications

• Betting on the dollar short term: While I’ve heavily structurally hedged my portfolio against de-dollarization, I’m reversing these hedges for the time being. It’s a short-term view, and I’ll monitor and reverse this as soon as signs point to the opposite (e.g., a re-reversal).

• US HH LNG exposure: taking small positions, tiny ones really, on US HH LNG futures for this summer and this winter.

• [Later] China arbitrage: Assuming China will act as a release valve on LNG capacity, I’m monitoring for this. No trades yet, but when the time comes, I’ll go long on European industrials as well as the euro.

Educational opinions only, not investment advice. No perspectives here on where the markets will go. Just food for thought

By popular demand, keeping this one shorter and less technical. Aiming to get in the habit of sharing ruminations weekly, so please reach out with feedback!

Today is about copper.

TL;DR: Copper is due for a strategic repricing

We’re on course for a supply shortfall of 10 million metrics tons of copper by 2040, a structural shortfall that mining investment cannot close because permitting timelines run a decade, and the easily accessible deposits are gone. Demand is being driven simultaneously by the converging of three inelastic forces in a single commodity cycle: AI data center infrastructure, NATO defense rearmament, and broad electrification. China controls the processing checkpoint (40% of global smelting capacity, and 66% of copper concentrate imports). The US designated copper is a critical mineral in 2025, yet policy has not kept up.

The investment thesis is not ‘buy copper miners.’ It’s that copper smelters and refiners outside China are the definitional Halo asset (Heavy Asset, Low Obsolescence) - by processing resources not information and they are trading at commodity cyclical multiples (6-10x EV/EBITDA) rather than strategic infrastructure multiples (12-16x). That is the mispricing, a gap of ~100-150%.

01 | Reading the Room Correctly

The concept of the Halo effect seems to be entering the investor zeitgeist. A great opinion piece on FT by Robert Armstrong here (The Halo Effect). Notably, citing Halo stocks (Heavy Asset, Low Obsolescence) as an emerging investor framework. The assets that fit are utilities, energy, materials, industrials. Protected not just by physical ownership but by permitting and regulatory moats that AI cannot quickly overcome And the kicker: “The real dividing line may not be how many assets a company owns, but whether it processes resources vs. information.” And that’s the copper thesis in elevn words.

The broader market context is worth sitting with. The S&P 500’s forward P/E has compressed from 23x in October to 21.4x today - while nine of the ten most closely watched valuation indicators are now in the top decile of their historical distributions. The CAPE sits at 39.64 - 99th percentile since 2000. The Buffett ratio (market cap to GDP) is at 2.18 - also 99th percentile. The Q ratio at 2.54 - also 99th percentile. Average household equity allocation is 54.9% - also 99th percentile. The one exception is trailing P/E, which is at the 75th percentile. By any honest reading of the tape, the market is priced for perfection across every valuation lens, and that has historically been a leading indicator.

Meanwhile, professional hedge funds have been net sellers of U.S. equities at the fastest pace since March 2025. Nvidia reported revenue up 73% YoY (the strongest growth in four quarters), and yet the stock dropped. That diversion between earnings reality and market reaction is extremely telling - the anxiety is no longer about whether AI demand is real. It’s about whether AI build out can catch up to the market - which is priced to perfection.

Broadcom (AVGO) reports earnings on Wednesday. The Street expects 26% EPS growth on AI networking hardware. Former NY Fed President Bill Dudley noted that the four major hyperscalers are expected to ramp capex to more than $600 billion in 2026 alone - creating what he explicitly called “inflationary demand for labor and materials.” He wasn’t thinking about copper when he wrote that. He should have been.

The Supreme Court struck down IEEPA-based tariffs in a 6-3 ruling last week, concluding that a persistent trade deficit doesn’t meet the statutory bar for national emergency . The immediate market read was relief. The more important read for this thesis is that the administration’s durable legal mechanism for protecting strategic materials just narrowed to one tool: Section 232 - the national security route that survived court challenges on steel and aluminum. Copper was designated a critical mineral in 2025. The Section 232 filing is the logical next step and, post-IEEPA ruling, the most politically and legally viable one.

02 | The Supply Gap: Numbers that don’t move

In January, S&P Global published Copper in the Age of AI: The Challenges of Electrification - co-chaired by Daniel Yergin, whose prior work on oil geopolitics turned out to be directionally correct in ways that were deeply uncomfortable for anyone who read it and did nothing. His book, “The Prize” (1990) is one I highly recommend. The study’s headline figures are being treated as interesting data points rather than as investment mandates.

"Economic demand, grid expansion, renewable generation, AI computation, digital industries, electric vehicles and defense are scaling all at once - and supply is not on track to keep pace."

Global copper demand is expected to reach 42 million metric tons by 2040, a 50% increase from current levels. Mining production is expected to peak in 2030 at 33 million metric tons, then decline as ore grades fall and the easy deposits are exhausted. Even a more-than-doubling of recycled copper scrap (from 4Mt today to 10Mt by 2040) doesn’t close the gap. This leaves a structural deficit of 10 million metric tons annually by 2040, or 25% below projected demand. The near-term version of this is already arriving: the International Copper Study Group forecasts a 150,000-ton deficit in 2026, flipping from a 178,000-ton surplus in 2025. J.P. Morgan’s 2026 refined deficit estimate is 330,000 metric tons, and while it’s price has already reached record highs, surpassing $13,000/mt, it doesn’t sufficiently reflect the supply-demand gap.

A quick note on the apparent contradiction: copper was up 40% in 2025 while simultaneously running a nominal surplus. This run up was due to tariff-driven front-running of US copper imports ahead of potential Section 2323 action, which filled CME-deliverable warehouses, and created a durable price premium on CME versus LME (that Chicago / US centric vs. London / global focused pricing). the physical market was tight even when the aggregate balance sheet looked loose The 2026 deficit removes even that ambiguity.

The supply chart deserves one further note: the curve above assumes the full announced project pipeline actually gets built and permitted on schedule.. this has never historically happened. The realistic supply trajectory is probably worse than what S&P Global models, which means a wider deficit. Mine supply growth was revised down to approximately 1.4% for 2026, roughly 500,000 metric tons below estimates made at the start of 2025, before a single new project was delayed.

03 | The Three Engines: A demand structure with no historical precedent

Every commodity super-cycle has a primary demand driver. Oil had post-WWII industrialization, then OPEC leverage in the 70s, then China’s infrastructure decade in the 2000s. Iron ore had one driver, China building cities. Copper’s current situation is categorically different. Three large-scale, structurally independent demand forces are scaling simultaneously.

Engine One: AI Infrastructure

A large hyperscale data center campus requires up to 50,000 metric tons of copper - transformers, busbars, power distribution units, cooling loops, networking cabling. The Stargate initiative carries $500 billion in committed AI infrastructure investment. OpenAI alone is telling investors it plans to spend $600 billion on compute by 2030. The four major hyperscalers are expected to collectively ramp capital expenditure above $600 billion in 2026 alone - which, as Dudley observed, creates inflationary pressure specifically for “labor and materials”. He was thinking about inflation broadly, but the materials he was referencing are disproportionately copper.

Nvidia’s Q4 earnings print this week (revenue up 74% YoY, guidance beat by the widest margin in two years) was rewarded with a stock decline. The market’s anxiety isn’t about AI demand. It’s about whether buildout can catch up to pricing. This divergence is a signal: AI infrastructure spending is running ahead of most demand models, while the physical commodity that infrastructure requires is being priced as a cyclical trade rather than a structural one.

Engine Two: Defense Rearmament

NATO’s trajectory towards 5% of GDP in defense spending is no longer aspirational political language. It is becoming budget reality across EU member states at a pace not seen since the Cold War. Joint Chiefs Chairman General Dan Caine is raising alarms directly with President Trump: US interceptor stockpiles are depleted to roughly two weeks of capacity against an Iranian missile campaign, and the USS Gerald R. Ford is tracking an 11-month deployment, a record. Iran accounts for 4% of global oil production, and 26% of crude transits the Strait of Hormuz. The military posture problem is already real and immediate, not theoretical. Modern military hardware (warships, aircraft, missile systems, electronic warfare platforms, hardened communications) is copper-intensive at every layer. S&P Global projects defense-driven copper demand to triple by 2040 from current levels. That figure predates the current acceleration in European and U.S. defense budgets. If anything, it understates what is now locked in.

Engine Three: Electrification

The most-discussed driver is also the least-appreciated in its sheer cumulative weight. S&P Global projects global electricity demand growing by nearly 50% by 2040, requiring the equivalent of adding roughly 330 Hoover Dams of generation capacity every year from now until then. EVs, grid modernization, utility-scale solar and wind (which require substantially more copper per unit of output than conventional generation), and residential heat pump adoption all compound. The IEA’s stated-policy scenario (the conservative baseline, not the climate-ambition scenario) already implies a 30% copper supply deficit by 2035.

04 | The Chokepoint: The processing layer nobody talks about

Here is the part of this thesis that most copper analyses miss entirely or if mentioned, briefly. The supply gap discussed above presumes that mined copper concentrate can actually reach end users. In practice, it usually cannot. Not without transisting Chinese processing infrastructure - smelting and refining.

China controls approximately 40% of global copper smelting capacity - the step that converts copper concentrate (roughly 30% copper content) into refined metal. It imports 66% of global copper concentrate production. The IEA projects China’s share of global refining capacity rising to 50% by 2040, without deliberate Western-world intervention. While copper mining happens more globally, such as in Chile, Peru, the DRC, or Indonesia, the lowest cost path from ore to refined metal runs through a Chinese smelter. Effectively, China is the toll collector in the copper value chain.

This is not an abstract geopolitical concern. When China imposed export restrictions on gallium and germanium in 2023, and rare earth processing restrictions in 2024, the downstream effects were immediate and painful for Western manufacturers. Copper is an order of magnitude larger than any of these markets. The precedent has been set. The capability exists. The US naming copper a critical mineral in 2025 was the right diagnosis. The treatment has not been prescribed.

05 | The Historical Analog: 1970s

Oil in 1970 is a structural parallel. In 1970, oil was a well understood commodity with established futures markets, clear price discovery, functioning arbitrage, and decades of economic research. What was not understood, not by governments, not by capital markets, not by most of the analysts who covered the sector, was that it was about to become the organizing principle of great-power competition. OPEC’s 1973 embargo didn’t come out of nowhere. The geology was already there. The geopolitical alignment was already forming. The demand was already locked in. What was missing was the catalyst that forced everyone to see what the data had been saying for years.

The decade that followed was disorienting for anyone who had been thinking about oil as a commodity. The impact was tremendous. Oil prices spiked from roughly $3/barrel to $35/barrel. The US created the Strategic Petroleum Reserve. The Carter Doctrine redefined military posture in the Persian Gulf. The Iranian Revolution remade the Middle East’s political map. These outcomes were not unforeseeable. They were foreseen, in writing, by a small number of analysts who took the structural data seriously and drew the logical conclusions. Yet, most investors did not.

Copper today sits in the same position oil occupied in 1970. It is well-understood as a commodity. It is not yet understood as a geopolitical asset. The S&P Global study is the Yergin equivalent - rigorous, sourced, co-authored by the person who wrote the definitive work on the prior commodity geopolitical transition, and being processed mostly as background reading rather than as the investment signal it represents.

The copper transition will take longer to play out than oil’s did, because the constraint lives in the smelting layer rather than in production. It’s more capital-intensive, less visible to the public, and harder to build around quickly. Slower doesn’t mean less certain. It means the window to position is open longer. And it means the repricing, when it finally arrives, will be sharper because the underpricing will have compounded for years.

The core mispricing.

The Halo framework, where value lies in processing resources rather than information, describes copper smelters and refiners with more precision than almost any other asset class. They are capital intensive. They require decade-long permitting timelines. They cannot be replaced quickly. They are not subject to software margin compression or AI disruption. They process the physical input that AI infrastructure requires. The irony is clean: the more AI scales, the more durable the Halo status of copper processors becomes.

Markets are treating copper as a cyclical commodity trade - buy miners when prices rise, sell when they fall. That is the wrong framework for this cycle, for the same reason it was wrong to trade Texas oil stocks as commodity cyclicals in 1974. Freeport-McMoRan (FCX) is the only fully integrated copper company in the US, operating the Miami smelter in Arizona, the last major copper smelter in the Western Hemisphere outside South America. Aurubis (ARUB.DE), Europe’s largest copper smelter, is the clearest direct expression of ex-China Western processing capacity. Boliden (BOL.ST) operates integrated mines and smelters across Scandinavia. These are not commodity cyclicals. They are strategic infrastructure assets trading at commodity multiples. That gap closes, either gradually through policy, or violently through crisis.

06 | What moves first, and in what order

The hypothetical near-term catalyst sequence.

1. Broadcom confirms the AI capex trajectory on Wednesday. The Street models 26% EPS growth. What matters more than the beat is the tone on hyperscaler infrastructure commitments through year-end. If Broadcom management confirms 2026 AI infrastructure spending is running ahead of initial guidance, which Google and Meta have already done independently, that’s direct evidence of the demand vector I’ve described. It also reinforces the Dudley point about inflationary demand for materials. Watch for any color on data center power and connectivity intensity; every high-density campus is up to 50,000 tons of copper.

2. Halo rotation accelerates as valuation extremes force capital reallocation. US equities, specifically information-processing equities, are at or near historical extremes. The CAPE at the 99th percentile. Household equity allocation at 99th percentile. Hedge funds net-selling at the fastest pace since March 2025. The rotation that follows overvalued information-processing assets goes somewhere. It goes to Heavy Asset, Low Obsolescence. The category that includes copper processors. That rotation hasn’t started in earnest. When it does, the re-rating will be significant.

3. Section 232 copper tariffs get filed. The IEEPA pathway is now judicially closed. Section 232, which survived court challenges on steel and aluminum, is the remaining durable mechanism. With copper already designated critical, the investigation can be initiated by executive action alone and follows a defined statutory timeline. When announced, it immediately expands the CME-LME price premium, accelerates incentives for domestic and allied processing investment, and functions as a policy starting gun for ex-China infrastructure buildout.

4. LME copper prices acelerate and the narrative shifts. The ICSG’s surplus-to-deficit flip in 2026 will become visible in LME inventory data as concentrate bottlenecks tighten and Chinese smelter output rolls over. That current prices will looks conservative if disruptions at Grasberg and El Teniente persist. The price level matters less than the narrative reclassification it triggers: from “cyclical bounce” to “structural repricing event.” That is the moment generalist capital is forced to engage seriously with the thesis.

5. The first major Western-government copper processing commitment is announced. Perhaps a US IRA-style incentive, EU Critical Raw Materials Act funding, or an Australian strategic minerals agreement that explicitly targets copper smelting or refining capacity. When that happens, Aurubis and Freeport stop trading on commodity multiples and start trading on infrastructure multiples. The gap between those two frameworks is large.

Potential companies in the mix: FCX; ARUB.DE; BOL.ST; SCCO; IVN.TO; COPX ETF; HG1 (Futures)

Not investment advice. Tickers listed for illustrative context only. FCX is the sole fully integrated US copper producer. ARUB.DE (Aurubis) is Europe’s largest copper smelter and the most direct expression of ex-China Western processing capacity.

07 | What would break this

This thesis is not without its risks. Some potential risks outlined here:

• If China enters a hard economic landing. China represents approximately 58% of global copper consumption. A genuine hard landing compresses near-term demand materially. This risk is non-trivial given the property sector’s unresolved dysfunction and persistent deflationary pressure. The structural thesis survives intact because the AI, defense, and electrification demand vectors are non-Chinese and still scaling, but the timing shifts and the near-term price path gets ugly. This is the most credible medium-term risk to the trade.

• AI compute efficiency. Effectively Moore’s law for AI. If model architecture improvements reduce the compute required per useful inference by orders of magnitude, the marginal data center campus doesn’t need to get built. This is a more nuanced risk than it appears: even if inference becomes dramatically more efficient, training compute continues to scale (Jevons’ paradox in compute has been consistent), and the installed base of existing and committed campuses already represents significant copper demand regardless of future efficiency gains. I don’t view this as a thesis-breaker, but it’s worth monitoring.

• Western permitting reform actually works. The supply-side bear case for the deficit thesis is the one scenario where governments successfully streamline mining and smelting permits at a pace that allows new supply to arrive before the structural gap opens fully. I will note that this has been promised in every prior commodity cycle and delivered in none of them. The Rosemont copper mine in Arizona (fully permitted, economically viable) sat in litigation for over a decade. The track record is what it is. If permitting reform did succeed, the deficit thesis doesn’t disappear, but the severity and timing shift materially.

None of these scenarios change the fundamental direction. The physics of the supply gap are too large and too slow-moving to be reversed by any single policy intervention or demand moderation. What these risks affect is the velocity: how quickly the market is forced to reprice the asset class, and how smooth or violent that repricing is.

To sum it up

The market is pricing copper processors as cyclical commodity businesses. They are not. They are strategic infrastructure

• Irreplaceable

• Permitting-moated

• Increasingly indispensable to AI buildout, defense rearmament, and electrification simultaneously.

The data already demands long-term processing agreements, critical mineral subsidies, Section 232 protection. When governments formalize this, these assets re-rate from commodity multiples to strategic infrastructure multiples. That re-rating is the trade that would materially uplift pricing. The geology is set. The geopolitics are forming. The only variable left is when the market catches up to what the physical world already knows. Copper smelters are the most Halo asset in the world right now,

The defining question of the next decade in global commodity markets is not who controls the most copper in the ground. It’s who controls the ability to convert copper in the ground into copper that can build AI data centers, naval destroyers, and electrified grids, without needing China’s permission. That question is beginning to be asked at the policy level. It has not yet been asked in equity markets.

We are in 1970. The oil crisis hasn’t happened yet. The geology is already there. The geopolitics are already forming. The demand is already locked in. And someone just handed you the framework to see it whether they meant to or not.

Educational opinions only, not investment advice. No perspectives here on where the markets will go. Just food for thought

The S&P 500 sits 1% from its all-time high. Beneath that surface calm, three structural dislocations have opened up that the market is mispricing - not trivially, but in ways that reflect genuine collective blind spots. Each is actionable.

1. Long crude, short rate-cut expectations. Two carrier groups converge on Iran on a military timeline that closes by mid-March. Oil to $90 - $115 in escalation scenarios. Core PCE is already 50 basis points above target. Markets are pricing two Fed cuts in 2026. That is wrong regardless of Iran, and dramatically wrong if oil moves.

2. Buy gold and real assets. The Treasury safe-haven is broken. The IMF has confirmed what many suspected: stocks and bonds in the US now sell off together. The 60/40 portfolio’s foundational assumption, that Treasuries provide ballast in a crisis, is empirically false. Capital needs a new safe haven. Gold, at $5,080, is filling that role and is not nearly done doing so.

3. Short the dollar, buy European defense stocks. Dollar bearishness at the most extreme since 2012. European defense is a legislated, government-mandated, decade-long spending super-cycle, not a theme or a trade. The Stoxx Europe Defence index (EU defense stock index) outperforms the S&P 500 by 20%+ over the next two quarters. This is the cleanest large-cap macro trade in global markets.

Where We Are

I want to begin with what I think is an underappreciated observation about this moment. Not since the early years of this decade have I seen a configuration in which so many of the bedrock assumptions of modern portfolio construction are being challenged simultaneously, and yet markets appear largely unbothered by this fact.

The S&P 500 is 1% from its all-time high. Q4 earnings growth was 14% year-over-year. By the surface metrics, the bull case is intact. But let us go one level deeper, as I find it always productive to do. Beneath the headline index, sector dispersion is the widest in years. Energy is up 22.75% year-to-date while financials are down 4.16%. Seven sectors are positive and four are negative. That disparity does not reflect a healthy, broad-based rally. It reflects two entirely different theories about the world running simultaneously in the same market.

At the same time, the IMF has formally stated that the historical negative correlation between stocks and bonds has reversed: they now move together during periods of stress. The Bank of America fund manager survey shows the most bearish dollar positioning since at least 2012. The Congressional Budget Office projects federal debt reaching 120% of GDP by 2036, and the Supreme Court’s ruling this week striking down most of Trump’s IEEPA tariffs removed an estimated $1.5 - $2 trillion in projected deficit offsets. Iran and the United States are closer to military confrontation than at any point since 2003.

And yet: the VIX sits at 19. Gold is at $5,080. The 10-year Treasury yields less than 5%. The market is not ignoring these risks. It is choosing, collectively, to treat them as manageable. Experience suggests that collective choices of this kind are worth scrutinizing carefully.

This week brings three catalysts that force a partial adjudication.

• Tech lynchpin earnings: Nvidia reports Wednesday against consensus estimates of $65.7 billion in revenue. Salesforce reports the same evening into what analysts are calling the “SaaSpocalypse”, a 42% stock decline driven by fears that AI agents will make enterprise software obsolete.

• Inflation print: Friday brings the PCE inflation print, the Fed’s preferred gauge, with core already at 3.0%.

• Iran: Secretary of State Rubio meets Netanyahu on February 28 specifically to coordinate on Iran.

I am not making predictions about what happens this week. I am making observations about the structural mispricings that these events will either expose or temporarily obscure. Three of them stand out.

The First Observation

01 | The Fed Is More Trapped Than It Knows and an Oil Shock Will Prove It

BUY: Energy equities, crude options | SELL: 2-yr Treasuries, rate-cut optionality

I want to start with the rate market, because I think it contains the single most obvious mispricing in any asset class right now. A mispricing that exists independent of what happens in the Middle East, though Iran makes it considerably worse.

Core PCE, the Federal Reserve’s preferred inflation gauge, stands at 3.0% year-over-year and accelerating. It rose 0.4% in December alone, the sharpest monthly print since February 2025. Nominal GDP growth exceeds 7-8% by Atlanta Fed tracking. The January FOMC minutes were explicit: most officials view the risk of inflation remaining above target as “meaningful,” and several stated that rate hikes could be appropriate if disinflation stalls. The Fed voted 10-2 to hold rates at the January meeting, and Chair Powell publicly celebrated the existence of dissents - a notable signal. In this environment, the market is pricing two full cuts and 50/50 odds on a third in 2026. I find this extraordinary.

The historical precedent here is instructive. As Ed Yardeni has noted, cutting rates when nominal GDP is running above 7% and core inflation is above 3% would be, as he puts it, “incredibly unusual.” The last time the Fed cut in comparable circumstances, the late 1970s under Chairman Burns, the result was a decade of inflationary pain that required Paul Volcker to inflict a severe recession to reverse. I am not predicting a replay of that episode. I am noting that the risk is asymmetric: the cost of cutting too early is considerably higher than the cost of waiting.

Now layer in the Iran situation. The facts are as follows. Two US carrier strike groups (the USS Abraham Lincoln and the USS Gerald Ford) are converging on the Arabian Gulf. Senior Pentagon officials briefed the White House that all forces required for potential operations will be in place by mid-March. Trump stated publicly that the world would know how this ends within “ten days”, a statement made four days ago. Indirect talks in Geneva produced “guiding principles” but no deal, and the US Vice President explicitly said Iran had not acknowledged American red lines on enrichment or missile capability.

We assign probabilities as follows. A negotiated deal (35% probability): Iran agrees to enrichment limits, oil stays range-bound, energy stocks give back some of their YTD gains. This is the constructive case and, in our judgment, not the modal outcome. Limited military strikes (45% probability): The US conducts airstrikes on nuclear and IRGC infrastructure. Iran does not close the Strait of Hormuz. Oil spikes 18-25% before partially retracing. This is the base case. Strait of Hormuz disruption (20% probability): Iran closes or threatens the Strait, through which approximately 20% of global oil supply transits. Oil breaches $115 per barrel. Global risk-off.

The interaction of Iran with the Fed is the key insight. An oil shock at a moment when core PCE is already 3.0% is not a temporary inconvenience. It is the thing that transforms a “higher for longer” Fed into a Fed that cannot cut at all, or must hike. The two cuts currently priced in for 2026 vanish. If core goods inflation (already elevated from the tariff pass-through of 2025) re-accelerates alongside energy, the Fed faces a genuine stagflationary dilemma of the kind it has not confronted since the Carter administration.

The Strait disruption tail, though only 20% probable, deserves particular attention. A 20% probability of a scenario that sends oil to $115, forces a Fed hike, and triggers global risk-off is dramatically underpriced when 2-year Treasuries are trading as if the next move is a cut. The asymmetry in the fixed income market is the most glaring: you lose perhaps 20-30 basis points if Iran deals and the Fed eventually cuts; you lose considerably more if either the geopolitical or the inflation trajectory breaks in the wrong direction.

The Second Observation

02 | The Treasury’s Role as Global Safe Haven Is Ending. Gold Is the Beneficiary, and the Market Is Slow to Accept This.

BUY: Gold, TIPS, real assets | REDUCE: Long-duration nominal Treasuries, 60/40 allocation

I have been thinking for some time about the foundational assumption that underlies most institutional portfolio construction: the assumption, embedded in the 60/40 portfolio, that when equity markets fall, Treasury bonds rise. It has been the basis of the risk-parity trade, the liability-driven investing model, and trillions of dollars in balanced fund allocation for three decades. And the IMF has now formally confirmed what many practitioners suspected but few wanted to say plainly: that assumption is no longer operative.

Since 2020, according to IMF analysis, stocks and bonds in the US, Germany, Japan, and the UK have tended to sell off together during periods of market stress. The driver, the IMF argues, is the persistence of supply shocks and inflationary fiscal expansion, which causes both assets to reprice downward simultaneously rather than providing the balancing effect that defined the prior era. In other words, the diversification benefit of owning Treasuries alongside equities, the very reason most institutional portfolios are structured the way they are, has materially degraded.

This is not a subtle, academic finding. It has profound practical implications. If Treasuries no longer provide ballast during equity drawdowns, investors face a meaningful question: what does? Gold, throughout recorded financial history, has been the answer of last resort, not because it earns a coupon or grows earnings, but precisely because its value is not contingent on the creditworthiness of any government or the policy decisions of any central bank.

“Without fiscal and monetary discipline, investors will search for other safe havens.” The IMF said this. They are correct. And gold, at $5,080, suggests the search is well underway.

The case for gold does not rest on Iran or the inflation data alone, though both support it. It rests on four converging structural forces that are simultaneously undermining the dollar’s safe-haven status and the Treasury’s crisis-hedge role. First, the fiscal trajectory. The CBO projects federal debt at 120% of GDP by 2036, with net interest payments alone reaching nearly 5% of GDP. The tariff ruling this week removed $1.5 - $2 trillion in projected deficit offsets. The St. Louis Fed has noted that CBO projections tend to understate the debt burden in low-inflation environments, which is where we are trending. The numbers are, as I have noted elsewhere, genuinely difficult to comprehend. A trillion dollars is a dollar a second for 31,700 years. The US is adding to its debt at a pace that defies casual numeracy.

Second, the independence question. Trump’s persistent and public pressure on Federal Reserve Chair Powell, including direct demands for rate cuts contradicted by the inflation data, has raised, in the minds of foreign capital allocators, a question that is uncomfortable to ask but impossible to ignore: is the Federal Reserve’s institutional independence durable? Bank of America’s fund manager survey found the most bearish dollar positioning since 2012, and the political dimension of dollar skepticism is explicitly cited among the reasons. Robin Brooks, the former Goldman Sachs strategist, has argued persuasively that the decade-long pattern of the dollar rising on strong US economic data is breaking down, as markets begin to price in the possibility that nominal yields will be managed below their market-clearing level.

Third, the 60/40 reallocation itself. If institutions collectively reconsider their bond allocations as a result of the stock-bond correlation reversal, the structural demand for Treasuries, which has for decades been supported by every major balanced portfolio in the world, diminishes. Some of that capital needs somewhere to go. A meaningful fraction of it will find its way to gold, to real assets, and to alternatives that provide genuine crisis hedging. This is a slow-moving but powerful and durable flow.

Fourth, and most importantly: the alternatives are unattractive. The euro and the yen each face their own structural challenges. Bitcoin, whatever one believes about its long-term potential, remains too volatile and too institutionally constrained to serve as a reserve safe haven for major allocators. That leaves gold: not as a speculative bet, but as the asset of last resort with the deepest history and the broadest global acceptance. I have sometimes heard gold dismissed as a “pet rock” by sophisticated investors who find its lack of yield disqualifying. I understand the sentiment, but I note that in a world where Treasuries are losing their safe-haven premium and the dollar is losing its reserve currency premium, the relevant comparison is not gold versus a 4.5% Treasury yield - it is gold versus a Treasury that falls alongside your equities when you need it most.

At $5,080 per ounce, gold appears expensive relative to its own recent history. But price alone is not the relevant metric. The question is whether the structural forces driving gold appreciation have run their course (I believe they have not). The fiscal deterioration is ongoing. The stock-bond correlation reversal is structural, not cyclical. The dollar’s institutional credibility is under sustained pressure. And the 60/40 reallocation trade is still in its early innings. Gold at $5,500 - $6,000 per oz. within two quarters is the most underappreciated large-cap liquid macro outcome in current markets.

Key Economic Prints This Week

• PCE Inflation (Friday): The most consequential data print. Core at +0.4% month-over-month in December, the sharpest acceleration since early 2025. A repeat number extinguishes any residual 2026 rate-cut probability. Watch for services vs. goods breakdown and revisions to prior months.

• Consumer Confidence (Tuesday): Walmart’s guidance miss flagged lower-income consumer strain explicitly. CFO John David Rainey cited “a pretty fluid, dynamic macro backdrop.” Confirmation here tightens the consumption divergence narrative and adds pressure to the Fed’s growth-vs.-inflation trade-off.

• Durable Goods Orders (Wednesday): First business investment read in the post-tariff-ruling environment. A bounce signals capex is returning; a miss suggests corporate uncertainty persists despite the legal clarity.

The Third Observation

03 | European Defense Is Not a Theme. It Is the Largest Government-Mandated Spending Cycle in a Generation, yet Markets Are Still Treating It Like a Trade.

SELL: USD | BUY: EUR, Rheinmetall, BAE Systems, Thales, European defense exposure

I want to make a distinction that I think is being lost in most discussions of European defense stocks. There is a difference between a cyclical trade, something that works because sentiment has shifted or a particular data point has surprised vs. a structural investment, in which the underlying economics are compelled by forces that are durable and largely irreversible. European defense is the latter, and the market is still, to a meaningful degree, treating it as the former.

The facts. Germany’s 2026 defense budget is €108 billion, a 25% increase over 2025, backed by legislation that exempts defense spending from Germany’s constitutional debt brake. The German government has authorized up to €400 billion in additional defense borrowing over five years. The EU’s “ReArm Europe / Readiness 2030” plan adds €150 billion in EU-level defense loans to member states. At the NATO Hague Summit, member states committed to 5% of GDP in defense spending by 2035, a target that, if achieved, would require the EU’s NATO members to approximately double their current annual expenditure. The Goldman Sachs Europe Defense basket is up 18% year-to-date. Barclays estimates European defense spending will surge to nearly $700 billion annually over the next decade.

Who knows what NATO spending in 2035 will actually be. Politicians make commitments that go unfulfilled. Geopolitical situations evolve. But note that the spending trajectory does not require 5% of GDP to make the investment case compelling. Even at half of the committed increase, the structural tailwind for European defense manufacturers is unlike anything the sector has experienced in the post-Cold War era. Germany’s defense chiefs and Britain’s defense chiefs issued a joint public plea this month for sustained popular support for rearmament, a remarkable document that speaks to both the seriousness and the political durability of the commitment. “Rearmament is not warmongering; it is the responsible action of nations determined to protect their people,” they wrote. This is not the language of a government exploring a spending option. It is the language of a government that has made a decision.

“After decades of underinvestment since the end of the Cold War, Europe faces the challenge of rebuilding military readiness - a process that will span well over a decade.” The supply constraint is real: you cannot build a tank factory in six months.

Janus Henderson Investors, European Defense Sector Note, February 2026

The supply constraint is the key insight that most equity analysis misses. European defense manufacturers (Rheinmetall, BAE Systems, Leonardo, Thales, Saab) have order books that are already full for four to five years. They are not competing for contracts; governments are competing for their production capacity. That dynamic, demand constrained by supply rather than supply constrained by demand, creates a durability of pricing power and margin expansion that the equity market, which habitually discounts cyclical businesses with high cyclical risk premiums, is undervaluing. The investment in new production capacity itself will take years to come online. The premium earned by existing capacity will persist throughout that period and beyond.

Now combine this with the dollar picture. The Bank of America fund manager survey shows the most extreme bearish positioning against the dollar since at least 2012, a level of conviction that tends to reflect not temporary sentiment but the accumulation of structural concerns that have crossed a threshold of investor tolerance. This leads to a four-sided dollar headwind, the product of years of diverging institutional credibility:

• political (NATO threats, Fed pressure)

• mechanical (the breaking of the strong-data-equals-strong-dollar pattern)

• fiscal (120% of GDP debt trajectory)

• competitive (European equity inflows at all-time records)

The Stoxx 600 (EU stock index) trades at a P/E of 18.3x versus 27.7x for the S&P 500. European stocks are on track for their highest-ever monthly inflows in February, following two consecutive record weekly flows of approximately $10 billion each. The valuation gap is the widest in two decades. The flow data suggests it is beginning to close. For US dollar-based investors in European defense stocks, the combination of earnings growth, multiple re-rating, and currency tailwind is a triple advantage that is, in my experience, genuinely rare in large-cap public equities.

There are clear risks. Defense production has very long delivery cycles; order books full for five years do not immediately translate into revenue. Geopolitical situations that drive urgency can also resolve, temporarily reducing political will. Individual European defense companies carry execution and procurement risk. And the valuation gap between European and US equities, while wide, has remained wide before. None of these risks are trivial. But the question for a serious investor is not whether the risks exist. It is whether the expected return compensates for them. In this case, I believe it does, by a considerable margin.

The bet: European defense stocks outperform the S&P 500 by at least 20 percentage points between now and the end of Q2. EUR/USD moves toward 1.12-1.15 by mid-year. For US-based portfolios, European defense ETFs and individual names with heavy European revenue exposure are the primary expression of this thesis.

The Larger Point

On Second-Level Thinking and Why These Three Observations Are Related

I have always believed that the most valuable investment insights are not those that identify a fact the market doesn’t know, but those that draw a different conclusion from facts that everyone can see. All three observations above fit that description.

Everyone knows about Iran. The carrier groups are being tracked in real time by Twitter accounts with satellite feeds. What the market appears not to have drawn the correct conclusion from is the interaction of Iranian escalation risk with the specific position the Fed occupies. Unable to cut without risking its credibility, unable to hike without risking the economy, and operating with inflation already above target. An oil shock into that configuration is not a standard geopolitical risk event. It is a monetary policy crisis in waiting.

Everyone can see that gold is rising. What the market appears not to have fully internalized is why. Gold does not rise because it is useful, because it pays a coupon, or because its fundamentals have improved. Gold rises when the institutions that are supposed to provide safety stop providing it reliably. The 60/40 portfolio breakdown that the IMF has now formally quantified is a fundamental shift in those institutions. The IMF argues that fiscal discipline is necessary to restore the safe-haven status of sovereign bonds. We are moving in the opposite direction on fiscal discipline. That is the structural backdrop against which to evaluate gold, and it does not resolve quickly.

And everyone has noticed that European defense stocks have had a strong year. What is underappreciated is the distinction between a strong year and the beginning of a decade-long structural re-rating. The railroad boom of the 1860s looked like a speculative froth until you realized that the transcontinental railroad was not an option but an inevitability. That the companies building it, whatever their short-term stock price volatility, were building something that would define economic geography for a century. European defense rearmament has that character. It is not optional. The political will is present, the legislative framework is in place, and the geopolitical drivers are not receding. The question is not whether this spending happens. The question is who captures the value of it.

There is also a currency overlay that ties all three together. A weakening dollar makes energy more expensive globally, supporting oil prices. It makes real assets like gold more attractive as a dollar alternative. And it provides a direct tailwind for USD-based investors in European assets. The three trades are not independent; they are expressions of a coherent view about the direction of American institutional credibility and the rebalancing of global capital allocation that follows from its erosion.

I want to close with the honest acknowledgment that I could be wrong on any or all of this. Iran could deal. PCE could fall more rapidly than expected. Gold’s structural bid could prove less durable than the institutional case suggests. European defense procurement could be delayed by political friction. Investment requires forming probabilistic views about an uncertain future and acting on those views with appropriate humility about the range of outcomes. What I am confident in is that the three mispricings identified above represent genuine gaps between market pricing and the most likely path of the underlying economics. That is as much as anyone can ask for.

Implications

On Iran and energy: This is the most time-sensitive of the three. The military timeline closes by mid-March; Secretary Rubio meets Netanyahu on February 28. Express through energy equities and crude options rather than futures, given roll costs and volatility. The short 2-year Treasury position is the cleaner, more liquid hedge against rate-cut mispricing . It works in every Iran scenario except a deal, and even in a deal, the inflation data alone argues against the currently priced cutting cycle. Size at 10-15% of book, given binary event risk. This is an insurance position, not a core holding.

On gold and the safe-haven transition: This is a 12-24 month core position, not a trade. The structural forces (fiscal deterioration, 60/40 breakdown, dollar credibility erosion) are slow-moving but durable. Gold at $5,080 is not cheap in absolute terms, but cheapness is the wrong frame; the relevant question is whether the forces driving gold appreciation have exhausted themselves. TIPS provide a fixed-income complement for investors who need some bond exposure but want real rather than nominal duration.

On European defense: This is the longest-duration and the highest-conviction of the three. The spending commitments are legislated, multi-decade, and geopolitically irreversible. The valuation discount to US peers is extreme. The currency tailwind is building. The supply constraint creates durable pricing power. For US equity-centric portfolios, European defense ETFs and individual names with heavy European revenue exposure are the primary expression. Rheinmetall, BAE Systems, and Thales are the primary beneficiaries; each offers differentiated exposure across land systems, aerospace, and electronics respectively.

The broader message is this: we are in a period of genuine regime change in global asset allocation. The assumptions that have governed institutional portfolio construction for thirty years (dollar safety, Treasury ballast, US equity premium, European irrelevance) are all under challenge simultaneously. That creates opportunity for those willing to update their priors. It creates risk for those who assume the prior regime will reassert itself by default.

Educational opinions only, not investment advice. No perspectives here on where the markets will go. Just food for thought

1: Buying more.

2: Diversifying global exposure — starting with the E.U.

Context

We’ve all seen the headlines, so touching here as brief context.

On April 2nd, Trump announced sweeping global tariffs aimed at “reducing reliance on foreign goods”—a move he dubbed “Liberation Day”. The policy shocked markets due to tariffs being both (i) higher; and (ii) sooner than analysts expected.

• Higher: 22.5% new effective tariff rate—a sharp jump from 2% previously, and well above Wall Street estimates (~10% expected, up to 15% worse case). Ultimately the announced tariffs were significantly worse than the worst case.

• Sooner: More surprisingly, the implementation window was 3 to 7 days, compared to expectation of 90 days, giving little room for diplomacy or adjustment, and thus significantly lowering the likelihood of countries negotiating more favorable deals.

Markets reacted.. adversely (to put it mildly). U.S. equities had ~$10 trillion in value wiped out (Bloomberg). The S&P has dropped ~10%. Major banks, including JP Morgan and Goldman Sachs, are signaling increased recession risk. For historical context, it’s worth revisiting the McKinley Tariff (1890) and Smoot-Hawley Act (1930)—as history often rhymes. This pandemonium isn’t isolated to the U.S., it’s global (see below)

Poor labeling, but the ordering is SPY (US) in blue; EU in yellow; Singapore / Pacific in teal; Asia excl. Japan in orange; and Japan in purple. Source: Google Finance, as of Apr. 8, 2025.

Reaction 1: Buying More

Why I’m increasing my net exposure

Markets may fall further—recessions have historically experienced declines of ~30-35%, with some as steep as -48% (1970s oil crisis), and -57% (‘08). It’s also quite possible that markets rebound—renewed trade diplomacy, Fed lowers rates or provides dovish guidance, or fiscal stimulus.

Either way, I’m personally always thinking about my investments on a 40+ year horizon. When quality assets are available at a 10% discount from just a week ago, that’s a prime buying opportunity. So I’m buying.

I’m increasing my net exposure by +20% over the coming days. And if prices drop further, I’ll continue to buy up (further benefitting my dollar cost average). If, instead markets rebound, I’ll benefit from a timely entry.

As a long-term investor, it’s a rational moment to buy up.

“…be greedy only when others are fearful” -Warren Buffett

Reaction 2: Diversifying, via E.U. Exposure

Amid trade tensions, the thinking behind it

I’m rotating 25% of my S&P 500 (SPY) exposure into VGK, an ETF focused on developed E.U. markets. Consider it a sector rotation analogue at a global level.

While I remain structurally bullish on the U.S., the commercial benefits of globalization (trade flows, cross-border collaboration, soft power and diplomacy) are likely to shift elsewhere as the U.S. adopts a more protectionist stance.

Globalization still matters. Fundamentally due to the benefit of economic specialization, and underpinned by free trade, globalization remains essential to economic growth. If the U.S. steps back, someone else will step forward.

To be clear — a global trade war hurts everybody. It’s not to say that one region’s loss is another region’s gain. Rather, under an evolving world order, who will step up? There are numerous contenders, and we can take a quick zoomed out view by looking at GDP over time by region

GDP by Region

E.U., China, and the U.S. are the largest major contenders.

source: IMF (Apr. 2025)

Other players (ASEAN-5, United Kingdom, Canada, Middle East) are all important (sufficiently developed and/or growing players on the global stage), but for now, I’m excluding them from consideration due to lack of sufficient scale, concentration risks, or domestic economic concerns.

While both China and the E.U. are affected by the tariff shock (largely via risk to their exports), the U.S.-China relations are especially adversarial. In contrasts, the E.U. has taken a more diplomatic tone, presenting a stabilizing force.

Europe offers scale, industrial strength (esp. buoyed by Germany), stable institutions. The region benefits from strong trade ties across continents, and stands to gain from a world still reliant on global supply chains and interdependence. Further, it is bolstered by a renewed focus on economic competitiveness. Thus, I believe it is positioned well to step up to the mantle in promoting free trade and international cooperation.

There are certainly risks to the E.U., particularly in a more fragmented and volatile world. With globalization and free trade precedents unwinding, thoughtful global diversification is prudent. The E.U. currently looks compelling.

Educational opinions only, not investment advice. No perspectives here on where the markets will go. Just food for thought