← Back to feed

Microsoft’s FY26 Q4 results make the AI infrastructure race look different from Alphabet’s. Google just showed what happens when quarterly capex overtakes operating cash flow. Microsoft showed the opposite pattern: it spent aggressively and still produced nearly $20B in free cash flow.

That does not make the buildout cheap. It makes Microsoft the hyperscaler with the cleanest near-term funding cover.

The Quarter

Metric	FY26 Q4
Quarterly revenue	$90B
Annual revenue	$331B
Profit growth	31%
Azure annual sales	Above $100B
Azure quarterly growth	43%
Cash capex	$35.8B
Total capex including leases	$41B
Free cash flow after capex	$19.6B

The important ratio is cash capex against operating cash flow. Microsoft used roughly 65% of quarterly operating cash flow for cash capital expenditure and still generated $19.6B in free cash flow. Alphabet’s Q2 comparison is sharper: $44.9B in cash capex against approximately $26B in operating cash flow, pushing free cash flow negative.

The gap is not just accounting. Microsoft can keep funding AI campuses, GPU leases, custom silicon, and OpenAI-aligned Azure capacity from the operating engine of Office, Windows, enterprise cloud, and security. That gives it more room to absorb a multi-year infrastructure cycle without asking investors to tolerate an immediate cash-flow inversion.

Azure Is Now the Proof Point

Azure passing $100B in annual sales changes the cloud narrative. For years, Microsoft’s AI story leaned on OpenAI distribution, Copilot bundling, and enterprise software attachment. Q4 shows the infrastructure base is no longer theoretical. Azure is now large enough that a 43% growth quarter adds revenue at a scale only AWS and Google Cloud can match.

The market question has also shifted. Investors are no longer asking whether AI demand exists. They are asking whether AI demand can repay capex fast enough. Microsoft’s answer is that the demand is already inside Azure’s run-rate, not sitting entirely in a future backlog.

That matters because AI capex has become less like ordinary server refresh and more like industrial construction. The spending lands before the revenue. Facilities, power, cooling, networking, and chip allocations require capital years before utilisation is cleanly visible in reported segments.

The Lease Number Matters

The $35.8B cash capex figure understates the actual infrastructure commitment. Microsoft reported $41B in total capex when lease-financed infrastructure is included. That lease layer is where hyperscaler AI accounting gets harder to read. A cloud provider can preserve cash flow while still taking on long-duration capacity obligations through leases and partner-financed builds.

That is not a trick. It is how data-center expansion works at this scale. But it means free cash flow alone no longer captures the full infrastructure burden.

Microsoft is not immune to the same circular financing risks visible across the sector: hyperscalers, chip vendors, AI labs, and compute providers increasingly finance one another’s demand. It is simply entering that phase with more operating cash flow than most rivals.

Why This Is Different From Another Big Tech Earnings Beat

The AI infrastructure race is splitting into two camps.

One camp has capability ambition but weak cash cover: model labs and neo-clouds that need external capital, prepaid contracts, equity raises, or debt structures to keep building. The other camp has operating cash engines large enough to fund the buildout internally. Microsoft is the strongest example of the second camp.

That strength buys time, not certainty. If AI workloads fail to convert into durable high-margin revenue, even Microsoft will eventually face a return-on-capital problem. But for now, Q4 shows why the company can keep spending while others explain how they will finance the next tranche.

The practical readout is blunt: Azure’s $100B business now gives Microsoft enough financial cover to treat AI infrastructure as a balance-sheet strategy, not a venture-backed sprint.