Microsoft lifts 2026 capex to $190B
- Microsoft told investors on April 29 that calendar 2026 capex will reach about $190 billion as it races to add Azure and AI capacity. (microsoft.com) - The eye-catching detail is that roughly $25 billion of that total comes from higher component pricing, while quarterly capex still hit $31.9 billion. (microsoft.com) - This matters because all three hyperscalers are still outrunning cloud estimates — and still spending harder, despite investor nerves about AI returns. (cnbc.com)
Microsoft just made the AI buildout look even more expensive. On April 29, the company said calendar-year 2026 capital expenditures should land a(microsoft.com) its own, but the real signal is simpler — demand is still strong enough that Microsoft is willing to spend through rising hardware costs instead of slowing down. (microsoft.com) ### What actually changed? The new piece of news is the forecast. Microsoft’s fiscal third-quarter results were strong — revenue came in at $82(cnbc.com) at constant currency. But the line that grabbed attention was capex: roughly $190 billion for calendar 2026, far above what many analysts had modeled. (cnbc.com) ### Why is the number so high? Part of it is plain old inflation in AI hardware. Amy Hood said about $25 billion of the 2026 capex plan reflects higher component pricing. (microsoft.com)ty — it is paying more for the same underlying pieces too. That matters because it means the AI boom is colliding with supply-chain pricing power, not just software demand. (microsoft.com) ### Is this just Microsoft? No — that is the bigger story. Alphabet also lifted its 2026 capex plan to as much as $190 billion, and Amazon, G(cnbc.com) was the fastest grower in CNBC’s roundup, but the shared pattern was the important part: cloud demand tied to AI is arriving fast enough that all three are still opening the spending taps. (cnbc.com) ### So are the cloud businesses paying for it? Partly, yes. Microsoft’s AI business passed a $37 billion annual revenue run rate, up 1(microsoft.com)g. But investors are still doing the obvious math — revenue is growing fast, yet infrastructure bills are exploding too. The question is not whether AI demand exists. It is whether margins can hold while the industry pours cash into chips, memory, and data centers. (news.microsoft.com) ### Why are investors une(cnbc.com)e four hyperscalers were already on track for close to $700 billion of combined AI-related capex in 2026, which means cash generation matters more than ever. If these companies keep spending ahead of realized returns, the market will start asking harder questions about payback periods instead of celebrating every cloud beat. (cnbc.com) ### Where does edge inference fit in? This is the part people are starting to ga(news.microsoft.com)r — especially when latency, bandwidth, or inference cost matters more than access to the biggest model. That does not kill the hyperscaler thesis. But it does strengthen the case for a split world where the biggest models train in giant data centers while some inference moves outward to cheaper or more specialized setups. This is an inference from the spending trend, not a policy shift Microsoft announced. (microsoft.com)aybe, but that is not the base case management is selling. The company is framing the spend as necessary to meet Azure and AI demand, not as speculative excess. The cleaner read is that Microsoft would rather risk looking aggressive now than risk being capacity-constrained later — especially while OpenAI-linked demand and enterprise AI adoption are still pushing usage up. (microsoft.com) ### Bottom line? The headline is not just that Microsoft raised capex. It is that AI infrastructure has become a scale race where even a company this p(microsoft.com)ft looks early. If it softens, $190 billion will look like the moment investors stopped treating AI capex as background noise.
