Four hyperscalers spent $130 billion on AI infrastructure in a single quarter — more than three times the inflation-adjusted cost of the Manhattan Project — and project up to $665 billion for the full year. Investors rewarded Alphabet, punished Meta, and noticed that nearly half of Google’s record profit came not from search or cloud, but from its Anthropic stake.
Microsoft, Alphabet, Meta and Amazon collectively reported first-quarter results after Wednesday’s bell, and the message to Wall Street was the same one Big Tech has been delivering for two years: AI is a black hole, and we are not done pouring money into it. The four companies told analysts they will spend up to $665 billion in 2026 on AI infrastructure, nearly 75 percent above the $381 billion they spent in 2025, according to Fortune.
The first-quarter capex figure alone — $130.65 billion across the four firms — was, as Fortune put it, ‘more than three times the inflation-adjusted cost of the Manhattan Project, in a single quarter.’ Investors split the verdict: Alphabet rallied roughly 10 percent on the print, Microsoft slipped 4 percent, and Meta tumbled 9 percent as it raised its full-year capex outlook to $125-145 billion, well above consensus, Bloomberg reported.
Alphabet’s headline number — an 81-percent jump in profit to $62.6 billion — drew the most scrutiny. A Fortune analysis published Thursday found that approximately $28.7 billion of that profit, very nearly half, did not come from search advertising, YouTube, Google Cloud or any operational line at all. It came from a fair-value mark on Alphabet’s equity stake in Anthropic. Amazon’s stake in the same company also juiced its quarter. The arithmetic is uncomfortable: two of the most-watched profit beats of the AI era were powered, in significant part, by writing up the value of a private competitor.
Operationally, AWS was the brightest spot of the cycle. Amazon Web Services posted 28-percent revenue growth to $37.6 billion, its fastest pace in fifteen quarters and roughly $1 billion above estimates, according to Motley Fool. Microsoft’s Azure and Alphabet’s Google Cloud each accelerated as well. The story coming out of every cloud earnings call was identical: backlog is a record, capacity is sold out, and the binding constraint is no longer demand — it is power, GPUs and physical real estate.
Skeptics seized on the gap. CoinDesk and Artificial Intelligence News both flagged the same risk: the AI infrastructure thesis demands that revenue eventually catches up to the spend. Right now, capex growth is outpacing AI-attributable revenue growth at three of the four hyperscalers. The Anthropic-stake disclosure cuts the same way — strip out fair-value gains on private holdings, and Alphabet’s ‘real’ operating leverage is impressive but not historic.
For investors and policymakers, two questions now sit in front of the market. First, can the demand-supply imbalance in compute persist long enough to justify the build? Second, what happens to GDP, energy markets and Treasury yields if it cannot? On Thursday’s print, the market voted yes on the build and reserved judgment on the thesis. The bigger reckoning — the one Fortune called ‘the $665 billion question’ — will arrive in earnings calls yet to come.
