Microsoft's Azure tax is the real target of its cloud antitrust probe

When the Federal Trade Commission opened its broad antitrust investigation into Microsoft in late 2024, it was easy to read the move as a last flourish by its departing chair, Lina Khan, the era's celebrated antitrust enforcer. She vacated the post in January 2025. The investigation did not follow her out the door. Under her successor, Andrew Ferguson, a Republican chosen by a president who had campaigned against "Big Tech's vendetta", the probe has grown rather than withered: in February 2026 the agency sent civil investigative demands to at least half a dozen of Microsoft's rivals, and by June it had formally widened its scope to cloud computing, software licensing, product bundling and Copilot, the AI assistant. Microsoft's cloud is no longer a fight about one activist's ambitions. The question is what the machine she started will eventually do to the company's most valuable asset.

The real instrument of Microsoft's cloud power is less glamorous than an AI subpoena. It is what critics call the "Azure tax": a licensing change Microsoft made in 2019 that requires customers to buy separate, more expensive rights to run Windows and SQL Server on rival clouds — Amazon Web Services, Google and Alibaba — than on its own. Competitors put the premium at as much as four times the price of running the same software on Azure. AWS has put the point still more sharply, telling Britain's competition watchdog that half of Azure's enterprise workloads would move to rivals if licensing were fair. The claim is self-serving. It also names the mechanism that matters: the world's most popular business software costs materially more to load onto a competitor's hardware, which is a structural reason its natural home is Azure.

The scale explains the seriousness. In the quarter to June 2026, Azure generated $29.4bn of revenue, about a third of Microsoft's $90bn total, and grew 42% even after the company tightened its definition of the business. Azure passed $100bn of annual revenue in fiscal 2026; the wider "Microsoft Cloud" bundle that the licensing tax helps keep whole brought in $59.3bn in a single quarter. This is why Microsoft chose early September to start disclosing Azure's revenue in dollars for the first time, collapsing three divisions into two. The number has stopped being a curiosity and become the heart of the investment case.

And yet, an investor might note, Microsoft's shares barely moved when the probe escalated. There are reasons for the shrug. Breakups belong to the 1990s; the FTC spent 19 months fighting Microsoft's $69bn purchase of Activision Blizzard and dropped the case in 2025; and probes do not always become complaints. The likelier end is not a judge but a consent order that rewrites licensing terms — an echo of the deals Microsoft signed with a group of small European cloud providers to let them host its software on non-Azure infrastructure at comparable prices, deals that pointedly excluded AWS and Google.

That unglamorous outcome is the actual risk, and it is more consequential than a courtroom loss would be. A remedy that caps the Azure tax would not unseat Microsoft, whose cloud growth rests on genuinely superior integration and enormous capital. But it would shave the premium that makes the platform sticky, weaken the bundling that lets Office's dominance feed the cloud's, and hand procurement officers a weapon they have never had in licensing negotiations. The market prices little of this in, on the reasonable grounds that the machine runs slowly and may yet stall. Institutions outlive the people who start them, however; that is the lesson of the past year. Azure's freshly disclosed numbers make it possible to watch, in plain dollar terms, whether its remarkable growth is a product of choice or of lock-in — which is the very question the probe, and now the silence around it, was designed to force into the open.