AI capex hits $725bn

The story here is data centers — and the bill for them just got much bigger. In the space of two days, Microsoft, Meta, Alphabet, and Amazon either reaffirmed or raised 2026 capital spending plans that now add up to roughly $715 billion. That is not a rounding error or a hype-cycle flourish. It is a coordinated decision by the biggest tech companies on earth to spend at a scale that starts to look like national infrastructure. ### Why did this number jump now? Because earnings season forced the companies to stop talking in abstractions. Amazon had already pointed investors to about $200 billion in 2026 capex. Then on April 29, Microsoft said it expects $190 billion this year. Alphabet, also on April 29, raised its full-year capex outlook to $180 billion to $190 billion. Meta lifted its own range to $125 billion to $145 billion. Add the midpoints and you land near $715 billion. Use the tops of the ranges and you get closer to the “$725 billion” headline making the rounds. (cnbc.com) ### What are they actually buying? Mostly compute — servers, networking, memory, and the buildings and power systems needed to run them. This is the physical layer of AI. Training models needs giant clusters. Serving those models to millions of users needs even more capacity. Microsoft was unusually explicit that higher memory prices alone will add about $25 billion to its 2026 capex. That tells you this isn’t just ambition. Input costs are rising too. (cnbc.com) ### Why aren’t investors panicking? Because some of the revenue is finally visible. Alphabet gave the clearest proof point. Sundar Pichai said Google Cloud revenue grew 63% year over year, topped $20 billion for the first time, and backlog jumped to more than $460 billion. On the same call, Alphabet said revenue from products built on its gen-AI models grew nearly 800% ye(cnbc.com)(abc.xyz) ### So who looks strongest right now? Alphabet and Amazon look best positioned on the “show me the money” test. Alphabet has cloud acceleration and a giant backlog. Amazon has AWS, which grew 28% year over year in Q1, plus a previously disclosed $200 billion capex plan that investors have had more time to digest. Microsoft also beat on Azure growth, but its $190 billion spending guide and margin pressure from depreciation made the tradeoff more obvious. (abc.xyz) ### Why did Meta get a tougher reaction? Because Meta’s monetization path is still less direct. Its ads business is strong — Q1 revenue rose 33% to $56.3 billion — but the company also raised 2026 capex guidance to $125 billion to $145 billion. That is a huge number for a company whose AI upside still depends heavily on future products and engagem(abc.xyz)ible payback loop. (prnewswire.com) ### Is this still an AI bubble story? Not in the simple sense. A bubble is usually “spending first, business model later.” What is happening now is more specific. The hyperscalers are building scarce infrastructure because demand is already outrunning supply in cloud AI. Alphabet even said cloud growth was driven by strong demand for AI products (prnewswire.com)r costs keep inflating. (abc.xyz) ### What should employees and smaller investors take from it? Basically — don’t treat all AI exposure as equal. A company promising giant AI upside without current revenue looks very different from one already converting demand into cloud backlog, paid seats, or subscription growth. In this market, visible monetization is getting rewarded more tha(abc.xyz)biggest build.” That makes the winners easier to spot — but it also makes the cost of being wrong enormous.