Corporate filings show the closures, Reuters reported in an exclusive.

The revenue figure is the one that explains the rest. China accounted for 1.5% of global revenue as of 2024.

Microsoft has no current plan to exit. It has simply stopped growing there.

Put 1.5% against the scale of the company and it almost disappears. Microsoft’s cloud business alone crossed $100bn in annual revenue in its most recent quarter.

China is now a rounding error attached to a large geopolitical risk. That is the calculation every sentence below follows from.

The company that told Google it was overreacting

The history makes this sharper than a normal retreat story. In 2010 Google left China over censorship and cyberattacks.

Bill Gates and Steve Ballmer, then chief executive, suggested Google was overreacting. Microsoft stayed, and democracy activists praised Google for going.

Sixteen years later Microsoft is arriving at the same place slowly, without ever announcing it.

That is the difference worth noticing. Google made a decision. Microsoft has made a series of quiet ones, and the filings are where you find them.

2023 was the year it nearly went

The internal argument got as far as the exit question. Reuters reports that Microsoft weighed quitting China in 2023.

Some executives held that the company took on too much geopolitical risk for too little economic return. Read that sentence next to the 1.5% and it is hard to argue with.

Then came the staff question. In 2024 Microsoft offered 1,000 of its top engineers relocation to the US and three other Western countries.

About a third accepted. Two thirds of the people it most wanted to move chose to stay, which is its own commentary on how these offers land.

The state stopped buying

The commercial squeeze is documented rather than inferred. China has pushed domestic software since 2017.

By May 2026 five of six reviewed Chinese government procurement guides did not recommend Microsoft products. The sixth listed Windows 10 China Government Edition, with additional management requirements attached.

A procurement guide is not a ban. It is more effective than one, because nobody has to defend it.

The pressure runs both ways. China has opened a cybersecurity review into Palo Alto Networks, and Washington has moved to push Chinese optics out of American data centres in a transceiver ban that Microsoft pays for directly.

What actually keeps the window open

Here is the part the headline number hides. Microsoft’s remaining China business is largely about Chinese companies leaving China.

Azure serves firms including ByteDance and Shein, which need Western technology to operate overseas. The customer is Chinese. The workload is not.

That is a durable position, and it is a strange one. Microsoft’s China strategy now depends on Chinese companies expanding abroad rather than on selling into China.

Shein illustrates the fragility. Its advisers are pitching a Hong Kong listing at under $30bn, down sharply from its peak.

Microsoft Research Asia tells the same story in miniature. Its labs now sit in Vancouver, Singapore and Tokyo.

The AI boom is what makes the position worth keeping at all. Companies building for overseas markets need Western models, Western tooling and a cloud their foreign customers already trust.

Without that demand, the argument for staying would rest on 1.5% and a long history. Neither of those wins an internal debate.

Where the money went instead

Follow the capital and the retreat looks less like withdrawal than redirection. Microsoft brought its fourth Indian cloud region online this month, against a $17.5bn commitment.

That is one country receiving a commitment larger than the entire revenue contribution of another. India gets regions and megawatts. China gets a maintained presence.

Microsoft’s own line is diplomatic. A spokesperson said the company operates in a regulatory environment that applies to every international supplier, and remains committed to the Chinese market.

Both halves of that sentence are true. It is also the sort of statement a company issues when the numbers have already spoken.

The technology still flows, in one direction

The relationship has not ended, it has inverted. Microsoft has reportedly considered putting China’s DeepSeek inside Copilot to control its AI costs.

So the American company retreating from the Chinese market is weighing whether to run Chinese models in its flagship product.

That is not a contradiction. It is what decoupling actually looks like from inside a business, where the politics and the bill of materials point in opposite directions.

Why this matters outside China

Europe should read this as a precedent rather than a curiosity. A government that wants domestic software can achieve it through procurement guidance alone, in under a decade, without passing a single ban.

That is the mechanism, and it is available to any state. The EU has spent years debating digital sovereignty in terms of rules and funding.

China did it with purchasing decisions, and the result is a 1.5% revenue line at the world’s largest software company.

The counterargument deserves stating. China’s approach comes bundled with censorship, forced localisation and an industrial policy Europe would not accept, and the outcome is a market Western firms increasingly cannot serve.

What would settle it

Two things are checkable. The first is whether the 1.5% falls further in the next annual disclosure, because a number that small can still halve.

The second is the Azure business specifically. If Chinese companies expanding overseas keep buying Western cloud, Microsoft has a real position rather than a legacy one.

The exit question has an obvious answer that nobody at Microsoft will give. It stayed in 2010 to make a point, considered leaving in 2023 for sound reasons, and remains today because a decision it never announced is easier to live with than one it would have to defend.

Get the TNW newsletter

Get the most important tech news in your inbox each week.