Big Tech lifts capex to $725B

AI infrastructure is turning into the main event in Big Tech earnings. Not apps. Not gadgets. Data centers, GPUs, networking gear, and the power systems needed to run them. What changed this week is that Microsoft, Alphabet, Meta, and Amazon all gave fresh numbers that make the spending wave look even bigger than investors already feared — or hoped. (microsoft.com) ### Why is capex suddenly the story? Because the AI race is no longer mostly about models. It is about who can physically supply compute at scale. Microsoft said on April 29 that it expects to invest roughly $190 billion in capital expenditures in calendar 2026. Alphabet, the same day, raised its 2026 capex range to $180 billion to $190 billion. Meta had already entered the year pla(microsoft.com)ploding as it pours money into AI. Add those together and you get a spending stack that is approaching the $700 billion mark across the biggest platforms. (microsoft.com) ### What did Microsoft actually say? Microsoft gave the cleanest number — and the most revealing one. Amy Hood said the company expects roughly $190 billion of 2026 capex, and about $25 billion of that comes from higher component pricing. That matters because it tells you this is not just “we chose to spend more.” Part of the jump is simple inflation in the AI supply chain — especially the(microsoft.com)n half its prior-quarter spend was on short-lived assets like GPUs and CPUs, which means this is not a slow, one-time campus build. It is a fast replacement cycle too. (microsoft.com) ### Is demand real enough to justify that? More than it was a year ago — yes. Microsoft said its AI business is now running at a $37 billion annual revenue pace, up 123% year over year. Alphabet said Google Cloud revenue jumped 63% and topped $20 billion for the first time, with backlog surging past $460 billion. Amazon said AWS grew 28% to $37.6 billion — its fastest growth in 15 quarters(microsoft.com)experimentation into heavier production use. (microsoft.com) ### So why are investors still uneasy? Because revenue growth and capital intensity are rising together, and that is a weird mix. Normally, cloud businesses get more efficient as they scale. AI is doing the opposite for now. Amazon’s trailing-12-month free cash flow fell to $1.2 billion from $25.9 billion, mainly because purchases of property and equipment jumped by $59.3 billion, d(microsoft.com)on capex in just one quarter. The business is growing fast enough to fund the buildout — but the buildout is eating a lot of the cash. (ir.aboutamazon.com) ### What is the bottleneck now? Basically, hardware availability and hardware pricing. Microsoft’s extra $25 billion tied to component costs is the clearest clue. In normal cloud expansion, the hard part is deciding where to build. In AI expansion, the hard part is getting enough accelerators, memory, networking, and power gear at all — th(ir.aboutamazon.com)compute become?” (microsoft.com) ### Does this help customers? In one sense, yes — more capex means more AI capacity, fewer shortages, and more room for enterprises to deploy models. But the catch is that customers also need these tools to produce measurable returns. If AI copilots, agents, and model APIs do not become repeat-use software budgets, then this spending starts to look like overbuilding. Right now, the strongest evidence is in cloud growth and backlog, not yet in a neat, settled margin story. (abc.xyz) ### Why does 2027 already matter? Because none of these companies sound close to done. Alphabet signaled spending will rise again in 2027. Microsoft is still capacity-constrained in parts of Azure AI. Meta keeps expanding data-center plans. Once the industry commits to this scale, it is hard to pause without risking share loss. That is why this week matters — it made clear that 2026 is not the peak. It may just be the new floor. (fool.com) ### Bottom line Big Tech is no longer dabbling in AI infrastructure. It is rebuilding the physical internet around it. The bullish case is that demand is finally large enough to absorb the spend. The bearish case is that everyone is racing at once, with hardware costs still rising. The next test is simple — whether all this capex turns into recurring, high-margin revenue fast enough to keep the buildout from looking reckless.