TradeFlockUSA
Tech

Microsoft's Azure backlog hits a record as enterprises lock in multi-year cloud contracts

CIOs are trading flexibility for capacity — a tell that the AI buildout is no longer optional capex.

James WhitakerTechnology Editor
Earth viewed from orbit at night with city lights

Earth viewed from orbit at night with city lights

REDMOND, Wash. — Microsoft's cloud business is starting to look less like a utility and more like a shipyard with a waitlist. People familiar with the company's enterprise pipeline said Azure's remaining performance obligations climbed again in the quarter, driven by multi-year deals that bundle ordinary software seats with scarce GPU capacity.

The shift is a reversal of the early-cloud playbook, when vendors competed to make switching cheap and commitments short. Today the constraint is not demand. It is transformers, land, and the ability to stand up clusters before a rival does. CIOs who spent 2023 and 2024 experimenting with copilots are now being told, politely, that the interesting SKUs are allocated.

That dynamic is a gift to Microsoft's sales force and a headache for finance chiefs who liked the idea of cloud as opex with an exit ramp. Several Fortune 100 technology officers described being offered better unit pricing in exchange for four- and five-year minimums — terms that would have been laughed out of a 2019 procurement review.

Competitors are not idle. Amazon and Google are writing similar contracts, and a handful of neoclouds are pitching themselves as overflow valves. But Microsoft's installed base of Office and security products gives it a bundling advantage that is difficult to unwind in a single RFP. The backlog, in other words, is not just compute. It is distribution.

Investors should watch two numbers that will not show up in a keynote: how much of the backlog is GPU-tied versus general-purpose compute, and how often customers are allowed to re-slide those commitments if model architectures change. The first tells you about scarcity. The second tells you about lock-in.

James Whitaker

Technology Editor

Reports on semiconductors, cloud infrastructure, and the industrial politics of AI.