Microsoft $80B AI Infrastructure Spend: Azure, Data Centers, and the OpenAI Commitment

Microsoft's fiscal 2025 capex came in north of $80B — up 44% from $55.7B in 2024, with more than half flowing into AI-enabled data centers. That's the short answer. The longer answer is more interesting.

Almost none of that increase is offices or traditional IT. It is data centers, Nvidia GPUs, and the power to run them — the physical plumbing of AI, much of it built to feed Azure and the OpenAI commitment. The question worth asking isn't whether $80B is a big number. It obviously is. It's what Microsoft gets for spending it, and whether Azure AI revenue compounds faster than the depreciation it creates.

Microsoft 2025 capex AI: the headline numbers

Microsoft's fiscal 2025 capex landed above $80B, up roughly 44% from the $55.7B it spent in fiscal 2024, with management stating that more than half is AI-enabled data center infrastructure: servers, Nvidia GPUs, and the buildings and power to run them. That 2024 figure was itself up about 58% over fiscal 2023's $35.4B. In two years Microsoft's annual capital spend more than doubled, and guidance pointed to further growth into fiscal 2026.

For context, Microsoft generated about $245B in revenue in fiscal 2024 and was tracking above $280B in fiscal 2025. Spending $80B+ of capex against that base means roughly one in every 3.5 dollars of revenue is being plowed back into physical infrastructure — a ratio that looks more like a utility than a software company. That shift, across every hyperscaler, is the whole story of this cycle.

Microsoft 2025 capex AI by quarter

Capex didn't arrive evenly. It ramped through the fiscal year as data centers came online and GPU orders cleared. Here is the approximate quarterly shape against the prior year, based on Microsoft's reported and guided figures. Microsoft's fiscal year ends June 30, so FY2025 spans Q3 2024 through Q2 2025 calendar.

Quarterly figures are approximate, reflecting reported actuals and management commentary including finance-lease obligations that Microsoft counts toward total spend. The direction is unambiguous: the front half of the fiscal year nearly doubled year over year. Track the live numbers on the Big Tech Earnings dashboard.

Where the Microsoft AI capex actually goes

"$80B on AI" is a slogan, not a budget. The money splits into a few buckets, and the proportions tell you what Microsoft believes about the next five years.

The Nvidia dependence is the part most analysts flag as a risk. Unlike Google, which designs TPUs in-house, Microsoft buys the bulk of its accelerators from Nvidia and pays a margin that runs well above 70% gross. Its Maia 100 in-house chip exists but handles a small share of total compute. That means a larger slice of Microsoft's $80B+ flows straight to Nvidia — a structural cost the company is trying to chip away at but hasn't solved.

The OpenAI commitment inside the Microsoft 2025 capex

A large share of this spend exists for one reason: OpenAI. Microsoft has invested roughly $13B in OpenAI and serves as its primary cloud provider, which means a meaningful portion of Azure's 2025 AI buildout is capacity OpenAI consumes for training and inference. When you see Microsoft's capex spike, you are partly watching it pre-build the data centers that run ChatGPT and GPT-5.

In 2025 the relationship was restructured. OpenAI converted into a for-profit structure and gained the ability to use other cloud providers — notably striking deals with Oracle and others as part of the $500B Stargate infrastructure push. Microsoft gave up exclusivity but retained large compute purchase rights, a revenue share, and IP access through 2030. The net effect: Microsoft no longer carries OpenAI's entire compute burden alone, which actually de-risks some of its capex even as the headline number rises.

The strategic logic is that Azure becomes the default home for frontier AI workloads — OpenAI's and everyone else's. You can see how the market prices that bet on the AI Valuations dashboard.

How Microsoft 2025 capex compares to other hyperscalers

Microsoft isn't spending in a vacuum. All four US hyperscalers ramped simultaneously, and the combined number crossed $320B for the first time. The comparison matters because compute is partly a relative game — falling behind on capacity means losing AI workloads to whoever has GPUs available.

Amazon leads on raw dollars, but Microsoft's position is distinctive because of the OpenAI relationship — no other hyperscaler has a frontier lab consuming its capacity at that scale. Meta's capex nearly doubled year over year — the steepest acceleration of the group — while Oracle's tripled off a small base as it landed AI hosting contracts including Stargate. The takeaway: this is no longer a question of whether to spend. Every player concluded being short on compute is more dangerous than overbuilding.

What Microsoft is actually building with it

Strip away the accounting and the $80B+ funds four concrete things, each tied to a revenue line.

Azure is the proof point. Microsoft reported Azure growth in the 30-35% range through fiscal 2025, with management attributing several points of that growth directly to AI services, and a commercial bookings backlog measured in the hundreds of billions. That backlog is the justification for the capex — it's contracted demand the company can't serve without more data centers. When the spend funds a backlog, it's investment. When it funds speculative capacity, it's a gamble. Nadella has been explicit that Microsoft would rather risk overbuilding than miss the demand.

What it means for founders and investors

The hyperscaler capex wave reshapes the ground startups operate on. Three implications stand out.

You can see the same dynamic across the sector on the AI Spending dashboard and in how the market prices it on AI Valuations. The capex race is the clearest signal yet that the incumbents believe AI is a generational platform shift — they don't spend $80B a year hedging.