Microsoft, Google lead $751B capex

This is a data-center story disguised as an earnings story. The headline number — roughly $750 billion of 2026 capex across the top hyperscalers — matters beca(know.creditsights.com)power, networking, cooling, and long-dated cloud contracts. The reason people care now is simple: after the latest earnings (know.creditsights.com)spend about $750 billion in 2026, up from earlier estimates near $620 billion. (know.creditsights.com)analysts rewriting their models after company guidance got more aggressive. CreditSights said Alphabet, Amazon, and Meta all came in above its already-elevated expectations, and that pushed its 2026 capex estimate for the top five hyperscalers to about $750 billion, up roughly $300 billion year over year. (know.creditsights.com) ### Who is actually spending? Alphabet is the easiest confirmed piece of the puzzle. On its April 29, 2026 earnings call, the company said full-year capex would be $180 billion(know.creditsights.com) AI infrastructure demand is still running hot. Amazon’s same-day call showed AWS growing 28% year over year to a $150 billion annual run rate, which helps explain why analysts keep modeling Amazon at the top end of hyperscaler spend. Meta’s April 29 investor materials are part of the same read-through. (abc.xyz)? Because they are the cleanest “demand is outrunning supply” stories. Alphabet said Google Cloud revenue jumped 63% and backlog nearly doubled sequentially to more than $460 billion. That is not what a company says when it is trying to cool spending. It is what a company says when it thinks every extra server rack can be monetized. Microsoft’s cloud business is showing the same basic pattern — huge AI demand, big infrastructure needs, and a willingness to keep leaning in. (abc.xyz) #(abc.xyz)cle in the top-five capex pile, and its ratio is the most extreme — about 86% of sales in 2026 by that estimate. That lines up with the market’s focus on Oracle’s role in giant AI infrastructure commitments, including the widely discussed OpenAI relationship and Stargate-related buildout. The catch is that Oracle has to fund a much bigger expansion from a smaller base than Microsoft, Amazon, or Alphabet. (know.creditsights.com) ### Is (abc.xyz)ound the Oracle-OpenAI contract has centered on a five-year figure near $300 billion starting in 2027. That helps explain why Oracle’s capex profile looks so stretched, but investors should be careful not to stack that whole contract on top of the $750 billion like it is extra. It is part of the reason analysts are lifting the infrastructure base case. (intuitionlabs.ai) ### What breaks first? Usually not chips — power and construction. Once capex ge(know.creditsights.com) cool the cluster, and connect it to fiber on time?” That is why long-term supplier lock-ins matter more now. The scarce thing is increasingly the whole data-center stack. That also explains why colocation landlords and infrastructure intermediaries keep getting pulled into the trade. This last point is an inference from the spending mix and company commentary, not a direct company claim. (know.creditsights.com)rscalers are acting like AI compute is a utility business in the making. When companies are willing to spend at levels CreditSights maps to 47% of Microsoft sales and 46% of Alphabet sales, they are not testing demand. They are trying to secure position before supply tightens further. (know.creditsights.com) ### Bottom line? The story is not just “Big Tech spends a lot.” It is that the AI race has crossed into physical infrastructure at national-scale dollar amounts — and Microsoft, Google, Amazon, Meta, and Oracle are now shaping the supply chain around that bet. (know.creditsights.com)