The July cull is now a fixture, and Microsoft's workers know it

Microsoft cut around 5,500 people this week, roughly 2.5% of its workforce, timed to the July 1 fiscal-year turn. It is the fourth July in a row the company has done this. The cadence is now the story: a 220,000-person employer has made a predictable annual reduction into a structural feature of the job, and the people absorbing the cost sit in Xbox, sales, and consulting — not in the divisions building the AI infrastructure the company has spent more than $80bn on. ^{1 2}

I want to be careful here, because two things are being said about these cuts and only one of them is fully supported by the evidence.

The first thing being said is that AI is driving the layoffs. Microsoft's spokesperson framed it as aligning "workforce to strategic priorities." ¹ Challenger, Gray & Christmas — the outplacement firm whose monthly tally is the closest thing the US has to a cross-employer layoff cause register — reports that "AI" has been the top self-reported reason for US layoffs three months running, with 87,714 AI-attributed cuts year-to-date, already past the full 2025 total. ²

The second thing being said, less loudly, is that Microsoft's stock is down roughly 19% over the prior month, the worst run since the dot-com bust, and that cost discipline at a company under that kind of repricing pressure looks a lot like cost discipline at any company under that kind of pressure. ³

Both things are happening. That is the honest read. Sales teams are genuinely being reshaped by pipeline automation. Consulting delivery is genuinely being changed by coding assistants. And a company whose AI capex thesis is being repriced by markets has genuine, ordinary, financial reasons to trim headcount at the fiscal-year turn — reasons that would exist if generative AI had never been invented.

What the Challenger number can and cannot tell us. It is a self-report. Employers choose whether to cite AI as the cause, and "AI" is now the most respectable available reason to give. A cut framed as forward-looking automation investment reads differently to markets, to the press, and to the workers left behind than a cut framed as financial distress. The number is real evidence of something. It is not clean evidence of the magnitude of genuine automation-driven displacement, and we should stop treating it as if it were.

That distinction matters because the policy response to "AI is eliminating jobs at scale" is different from the policy response to "companies under cost pressure are using AI as cover for cuts they would have made anyway." Workers lose in both scenarios. But unions, regulators, and workers themselves need to know which one they are arguing about.

The distributional shape is the clearer story. Microsoft has committed north of $80bn to AI infrastructure — data centres, chips, model training, the physical and computational scaffolding of the bet. ² The workers being cut are not in the teams running that scaffolding. They are in Xbox, following an internal divisional reset. They are in sales, where the automation pressure is real. They are in consulting, where AI coding tools are compressing the delivery labour model. ¹

The upside of the AI bet, if it works, accrues to Microsoft's balance sheet and to the teams operating the AI platform. The downside, the immediate, personal, this-month downside, is being borne by a different population entirely. That is not an accident of restructuring. It is a textbook case of distributional incidence: investment concentrated, displacement dispersed. The people who paid the cost of the bet are not the people who will collect if it pays off.

None of this is unique to Microsoft. It is, however, unusually legible at Microsoft, because the numbers are big and the annual cadence makes the pattern visible in a way single-round layoffs at other firms do not.

And the cadence is now the labour signal. Four consecutive Julys of significant cuts at a 220,000-person employer is no longer a series of restructurings. It is a schedule. Workers inside Microsoft, and workers at every large tech firm watching Microsoft, are rationally discounting their tenure security around the fiscal-year turn. Managers know it. Recruiters know it. The people being hired in September know they are being hired into a firm whose July is a known event.

That has consequences beyond the 5,500 who left this week. It changes how people negotiate. It changes what they will and will not push back on in April and May. It changes which internal projects get taken and which get avoided. Managed precarity, once it becomes predictable, does not stop being precarity — it just stops being surprising. And a workforce that is not surprised by its own vulnerability behaves differently from one that is.

I find it hard to look at the four-year pattern, the $80bn capex figure, and the 19% monthly stock drop together and describe what is happening as primarily about AI capability arriving in the workplace. It is about a company managing a very large bet on AI infrastructure, in public markets, with the labour line as the adjustment mechanism. The AI story is true. It is also convenient. Both can be, and are, the case.

What I would watch: whether the July cadence starts spreading to other US$100bn-plus tech employers, whether the Challenger self-report methodology gets refined enough to distinguish automation displacement from restructuring cover, and whether the workers absorbing these cuts organise around the cadence itself. A predictable annual cull is, among other things, a coordination opportunity. Microsoft has, without meaning to, given its workforce a schedule to organise around.

Whether they use it is a different question.

Footnotes

¶