Microsoft, AWS plan $390B combined capex
- Microsoft told investors on April 29 it expects roughly $190 billion of 2026 capex, while Amazon has already committed about $200 billion, mostly for AWS. - The tell is pace: Microsoft spent $31.9 billion in one quarter, and AWS just posted 28% growth — its fastest in 15 quarters. - Add Alphabet and Meta, and 2026 hyperscaler capex now points near $700 billion — enough to reshape power, cooling, and server supply.
The story here is datacenters — and the fact that Microsoft and Amazon are now spending on them at a scale that barely looked real a year ago. Microsoft used its April 29 earnings call to point investors to roughly $190 billion of 2026 capital spending. Amazon had already set a roughly $200 billion plan for 2026, with most of that aimed at AWS. Put those together and you get a $390 billion buildout centered on AI compute, cloud capacity, and the physical guts of both. (microsoft.com) ### Why are these numbers such a big deal? Because capex is the concrete version of an AI strategy. Hype is cheap. A datacenter full of GPUs, networking gear, backup power, chillers, and land is not. Bloomberg framed the broader 2026 spending plans from Amazon, Microsoft, Meta, and Alphabet at about $650 billion, and industry trackers(microsoft.com)is not normal corporate spending — it is an industrial build cycle. (bloomberg.com) ### What did Microsoft actually say? Microsoft’s official April 29 release showed another strong cloud quarter, and the more revealing detail came in management’s guidance: roughly $190 billion of 2026 capital expenditures, including about $25 billion tied to higher component costs. In th(bloomberg.com)into short-lived assets like GPUs and CPUs — which tells you this is not just land and buildings, but rapidly deployed compute. (microsoft.com) ### What about Amazon and AWS? Amazon set the tone earlier, in February, when it laid out a $200 billion 2026 capex plan. Andy Jassy said the spending would be predominantly in AWS, with some for core cloud workloads but most for AI. That matters because AWS is not spending into a slowdown. In Amazon’s April 29 quarter, AWS revenue ro(microsoft.com)t the capacity is getting absorbed. (datacenterdynamics.com) ### So what are they buying? Basically everything that turns electricity into tokens. Servers. GPUs and custom AI chips. High-speed networking. Liquid cooling. Backup power systems. New buildings and regional expansions. Amazon has been especially loud about chips, saying its custom silicon busine(datacenterdynamics.com)ns of billions to the bill. The bottleneck is no longer just software talent — it is physical capacity. (ir.aboutamazon.com) ### Why do investors care beyond Microsoft and Amazon? Because this spending spills into landlords and equipment vendors. If hyperscalers keep ordering capacity, the beneficiaries are not just chipmakers. Colocation operators, wholesale datacenter REITs, power-management companies, and therm(ir.aboutamazon.com)picks and shovels into the buildout, not just AI stories. This last step is partly inference, but it follows directly from where the capex is going. (bloomberg.com) ### What is the catch? Returns have to show up. Amazon’s $200 billion plan rattled investors when it was announced, and Microsoft’s number also landed well above prior expectations. These companies are betting that AI demand stays strong enough to fill the new capacity fast, because idle (bloomberg.com)ing the obvious question — how long can spending outrun visible profits? (datacenterdynamics.com) ### Bottom line This is not a side bet on AI anymore. Microsoft and Amazon are treating compute like core infrastructure, and 2026 is shaping up as the year that spending moved from huge to historic. (fool.com)
