The Capex Confession

in #article17 hours ago

The Capex Confession

The market is pricing a religion. And this week, the congregation showed up.

Four of the most powerful corporations on earth — Alphabet, Amazon, Meta, Microsoft — lined up to tell you they will spend somewhere north of $600 billion on AI infrastructure in 2026. Not combined revenue. Not profit. Capital expenditure. Money that leaves the building and doesn't come back for years, decades, maybe never. And the S&P 500, fresh off its best April since 2020, greeted the news by pushing toward record highs. April was the strongest month for equities since 2020, with the index touching new highs as it kicked off May. Everyone has decided to believe.

Here's the tension nobody wants to sit with: markets are not rewarding AI spending equally. They are rewarding AI spending that shows near-term monetization — as seen with Alphabet — and punishing spending without clear incremental returns. Which means the crowd knows, on some level, that not all of this $600 billion will find its way back. They're just betting they can identify the winners before the bill comes due. Good luck. The bond market, at least, is keeping one eye open — the 10-year U.S. Treasury yield sits at 4.39%, offering a quiet editorial on the cost of all this dreaming.


The FOMC held rates steady, as everyone expected. The Fed kept its target range at 3.50%–3.75% for the third consecutive meeting. Jerome Powell's final press conference as chair — his handover to Kevin Warsh looms in May — produced the usual careful formulations. The market has learned to read Powell's silences better than his sentences. What it heard was: nothing changes soon. Rate cut probability curves remain becalmed. The Fed is a spectator in its own economy right now, watching AI infrastructure promises do the stimulative work that monetary easing used to do. The circularity is dizzying.

Apple reported Thursday. AAPL posted fiscal Q2 2026 EPS of $2.01 against an estimate of $1.94 — a beat of 3.61%, up from $1.65 in the year-ago quarter. The stock got a bounce. What's actually interesting isn't the beat — it's the margin story. Gross margins expanding, operating cash flow surging, and a fresh $100 billion buyback authorization alongside a raised quarterly dividend of $0.27. In the first half of 2026, Apple deployed $36 billion toward share repurchases alone. Tim Cook is leaving a very particular inheritance: a machine that converts installed base into recurring cash and returns it to shareholders with metronomic precision. Apple Intelligence, Siri's theoretical successor, is still more concept than product. But nobody held the stock down for it. The buyback is the product now.


What isn't priced anywhere near adequately is the energy shock sitting underneath all of this like a geologic fault. Oil at $102 a barrel. The Strait of Hormuz nearly closed. Ships have almost completely stopped moving through the Strait — a route through which about a fifth of the world's oil and liquefied natural gas passes. The BoE, in its April Monetary Policy Report, held Bank Rate at 3.75% — 8 votes to 1, with one MPC member pushing for a hike to 4% — and issued language that reads like a committee flying partially blind. Higher wholesale gas futures prices will mostly affect UK household utility bills from July, following the expected increase in the Ofgem price cap. The BoE laid out three scenarios in its quarterly report. None of them contains a clean path to 2% inflation by year-end. UK CPI is already at 3.3% and the gas futures curve is still rising. The MPC is trying to look through a supply shock while privately hoping the shock doesn't become structural. That's the same hope the ECB had in 2021.

The ECB, for its part, held its deposit rate at 2% but acknowledged that risks to the eurozone economy had intensified, with the governing council discussing a potential rate rise at length. The language is shifting. There is a scenario — not the base case but increasingly discussable — where both the BoE and the ECB are hiking into a slowdown this summer. Energy-driven stagflation, act two. The first act destroyed bond portfolios and humiliated central bank credibility between 2021 and 2023. The second act arrives into a world with higher starting rates, more geopolitical noise, and equity markets still trading on the assumption that the disinflationary cycle never really ended.


Gold at $4,600 an ounce says something different. Gold doesn't care about AI capex. Gold doesn't get excited about earnings beats. Gold prices have been steady around $4,600 per ounce even as the dollar continues drifting lower against a basket of international currencies. Central bank buying hasn't abated. The dollar is weakening — which is itself a macro editorial, a slow verdict on the fiscal and monetary trajectory of the United States. When the world's reserve currency grinds lower during a risk rally, that's not optimism. That's a hedge.

The OpenAI figure buried in the week's reporting is the one that deserves more attention than it got. OpenAI, privately valued north of $850 billion in secondary markets, reportedly missed its own revenue and user growth targets ahead of a planned IPO. The company around which the entire AI bull thesis orbits — the thing that justified NVDA at $5 trillion market cap, Azure's growth story, AWS's Remaining Performance Obligations metric — is apparently a less clean business than the narrative required. None of the mega-cap CFOs said this on their earnings calls. They didn't need to. The capex kept flowing. The market kept buying.


Here is the animating question for the next six months: what does it look like when $600 billion in annual infrastructure spend doesn't produce the revenue acceleration it was supposed to? We've had the faith phase. At some point comes the accounting.

The BoJ, meanwhile, revised its fiscal year 2026 inflation forecast up to 2.8% from 1.9% in January, while halving its growth forecast to 0.5%. Stagflation in Japan. The word nobody was saying eighteen months ago is now in the actual forecasts of central banks. WTI at $102. Gold at $4,600. The dollar softening. University of Michigan consumer sentiment at record lows. Retail sales remain resilient despite sentiment falling below levels seen in both the 2008 crisis and the 2020 pandemic. Americans are spending against their own judgment. That divergence resolves one way eventually.

The market is choosing to call this a good week. Apple beat. Alphabet beat. The S&P touched all-time highs. April was spectacular. Maybe it is a good week. But the energy shock is real, the central banks are cornered, and the AI capex religion requires a level of monetization that has not yet been demonstrated at scale.

Belief is not a balance sheet. The conversion still hasn't happened.


May 2, 2026

Sort:  

Upvoted! Thank you for supporting witness @jswit.