← Back to feed

The four largest AI infrastructure spenders in the world just reported the same quarter. Nothing is slowing.

Amazon, Alphabet, Microsoft, and Meta combined for $130.65 billion in capital expenditures in Q1 2026 — more than three times what the Manhattan Project cost to develop nuclear weapons, and 71% above what they collectively spent in the same quarter a year earlier.

Google Cloud Breaks $20 Billion

Google Cloud posted $20.03 billion — the first time it has crossed that threshold in a single quarter. Analysts had projected $18.4 billion. The $1.6 billion beat arrived despite Alphabet management explicitly flagging that demand outpaced available supply. They described the quarter as “capacity-constrained,” meaning revenue would have been higher if more compute had been ready to sell.

Alphabet’s total revenue for Q1 landed at $109.9 billion with EPS of $5.11, against $90.23 billion and $2.81 EPS in Q1 2025 — a 22% top-line gain year-over-year. Google Advertising contributed $77.2 billion.

The Capacity Constraint Is the Story

The capacity constraint note is the most consequential sentence in Alphabet’s results. It tells you two things simultaneously: Google’s enterprise AI demand is real, and Google’s own infrastructure build-out is now the binding constraint on how fast cloud revenue can grow — not market penetration.

Morgan Stanley analyst Brian Nowak wrote after the results that he believes Google’s TPU business is not priced into the stock and could become a significant driver heading into 2027. Google announced the TPU 8t (training-optimised) and TPU 8i (inference-optimised) at Cloud Next the week prior.

AWS and Azure

Amazon Web Services beat cloud growth expectations driven by enterprise AI spending. AWS shares dipped 3.7% in extended trading despite the beat — a sign the market had already priced in strong results. AWS forecasts had centred around 25% growth.

Azure was expected near 40% growth. Microsoft entered results as the weakest performer of the four on a stock basis, down approximately 12% year-to-date, with analysts noting that only 3.3% of its 450 million enterprise users are paying $30 per month for Copilot — meaning AI monetisation of the installed base has barely started.

No Ceiling

Every management team on the four earnings calls signalled elevated capital expenditure through the rest of 2026. The $130.65 billion Q1 figure represents a run-rate of over half a trillion dollars annually across just four companies.

The infrastructure thesis that has driven AI valuations for two years is not hypothetical. The money is being spent. The cloud revenue is arriving. And at least one of the four largest buyers of compute says it left revenue on the table because it could not build fast enough.