The Numbers Were Good. The Spending Plans Scared People.
Microsoft posted $82.9 billion in Q3 revenue. Azure grew 40% year over year. Their AI annual run rate crossed $37 billion. Microsoft 365 Copilot now has 20 million paid seats. By any normal measure, that's a blowout quarter.
Meta grew revenue 33%, their fastest pace since 2021. Their business AI tools handled 10 million conversations a week as of late March, up from just 1 million at the start of the year. Also a blowout quarter.
Then both companies raised their capex guidance, and stocks sold off anyway.
Meta raised its 2026 capex forecast to as high as $145 billion, up from the $115–135B range they had announced just days earlier. The stock dropped nearly 7% after hours. Investors called it "asymmetric risk." What they meant was: we believe you're spending this money, we just don't know if you're going to get it back.
$700 Billion This Year. $1 Trillion Next Year.
When you add up all the hyperscalers, Wall Street now estimates total AI capital expenditures will hit somewhere between $800 and $900 billion in 2026. Analysts at Evercore and Bank of America put 2027 above $1 trillion. That's not a typo.
Nobody knows where this ends. The hyperscalers are betting that whoever builds the most infrastructure wins the AI era. That might be true. It also might be the largest coordinated capital misallocation in history. Probably won't know for a few years.
Meanwhile Mistral raised $830 million in debt to build an NVIDIA-powered data center near Paris. Even the scrappy European challenger is doing nine-figure infrastructure bets now.
What I Keep Coming Back To
Microsoft's AI run rate is $37 billion annually. That's real revenue, not theoretical. Azure growing 40% off an already large base is genuinely impressive. The return-on-investment case exists.
The tension is that the spending is growing faster than the revenue. Meta's $145B capex will be nearly double their entire revenue from just a few years ago. At some point the story has to change from "we're building" to "we're earning." We're not there yet, and that's what the market is pricing in.