Four companies committed $700 billion to capital expenditure in 2026. Microsoft: $190 billion. Alphabet: $180-190 billion. Amazon: ~$177 billion. Meta: $125-145 billion. The market looked at this number and saw insatiable AI demand. The S&P closed April at 7,209 — best month since November 2020.
Two of the four told you what the number actually means.
"The increase was mostly due to higher component costs, particularly memory pricing."
— Meta, Q1 2026 earnings call, on raising capex to $125-145B
"Soaring memory costs."
— Microsoft, on full-year capex of $190B
NAND flash is up 246% from pre-Hormuz levels. DRAM is up 172%. These aren't AI demand premiums. They're supply disruption premiums. Qatar's Ras Laffan — source of 30% of global helium used in semiconductor fabrication — has been offline since the blockade began. Japan's photoresist plants, which produce 95% of the world's EUV-grade material, are running short on naphtha feedstock that transits the Strait. Six of twelve Japanese crackers have cut output. The inputs to make memory chips are scarce. The chips cost more. The capex numbers go up.
The market doesn't separate these.
Nobody knows the exact split. That's the point. When NAND triples and DRAM nearly triples, a $700 billion capex figure doesn't mean what it meant twelve months ago. A dollar of capex buys fewer chips. The same build-out costs more. The market sees a bigger number and hears "more demand." The earnings calls say "higher component costs."
Google was the only one the market rewarded — up 4% after hours. Cloud revenue grew 63% to $20 billion, backlog doubled to $462 billion, and crucially, Google's capex raise was proportional to its revenue growth. Revenue justified the spend. For Meta, it didn't. Stock fell 6%. For Amazon, FCF collapsed 95% to $1.2 billion. The market is starting to differentiate — but it's differentiating between companies, not between demand and cost within the capex number itself.
The month that proved it
Today the Bureau of Economic Analysis reported Q1 GDP at 2.0% — below the 2.3% consensus. Core PCE hit 3.2%, sixty percent above the Fed's target. Headline PCE reached 3.5%, driven by energy. Brent crude touched $126 a barrel intraday — a four-year high — after CENTCOM briefed the president on three military strike options against Iran.
The S&P gained more than 10% in April. Best month since November 2020. Consumer sentiment sits at 49.8 — a 74-year low. Gas prices at $4.18 — highest since July 2022.
The narrative says the economy dodged stagflation because GDP came in at 2.0% instead of the Atlanta Fed's 1.2% nowcast. The data says GDP missed consensus, inflation is hot, oil is at war-time highs, and the consumer is the most pessimistic in seven decades. The relief isn't that the data was good. It's that it wasn't catastrophic.
Eli Lilly reported earnings this morning: revenue up 56%, Mounjaro up 125%, stock up 10%. The market's best month was built on AI capex and weight-loss drugs. Both are real businesses. But one of those capex numbers is inflated by the same supply disruption that's causing the inflation the Fed can't agree how to handle.
Seven hundred billion dollars. Some of it is demand. Some of it is the Strait of Hormuz. The market counted all of it as demand. It closed the month at an all-time high.