Microsoft Offers Buyouts to 60,000 Employees as AI Spending Climbs

The memo went out on Monday morning to four countries at once. Microsoft offered voluntary exit packages to employees in the United States, Canada, India and Britain at job level 70 and below, a band covering roughly 60,000 people, or 27 percent of its 223,000 staff.

Amy Coleman, the company’s chief people officer, described the program in the memo as a way to give employees more say in a period of change. Those who accept must answer by early December and leave by December 31.

Job level 70 in Microsoft’s internal ladder sits at the senior director tier, above most engineers and below the corporate vice presidents who report to the chief executive. It is where large teams of managers, program leads and sales directors sit, and it is the layer that has grown fastest during four years of hiring for cloud and artificial intelligence work.

The offer lands in the same quarter as a capital spending bill above $50 billion. Microsoft told investors to expect fiscal first-quarter capital expenditure, which covers the three months through September, to exceed that figure, against $34.9 billion a year earlier. Most of the money goes into graphics processors, networking and the buildings that hold them.

Payroll and depreciation are now competing for the same dollars, and depreciation is winning. A data center built this year begins charging against earnings in the following one whether or not the workloads arrive, while a salary line can be reduced in a single quarter with a single memo.

This is not the company’s first attempt at the same problem. In April it opened a voluntary retirement program to United States employees. In July it cut 4,800 positions. Across 2025 it eliminated about 15,000 roles in stages, including 6,000 in May and 9,000 in July, while adding headcount in AI research and silicon design.

A buyout differs from a layoff in ways that matter to the balance sheet and to morale. Severance for an involuntary cut is negotiated, litigated and, in Europe, subject to consultation with works councils. A voluntary program sets the terms in advance and lets the company choose the shape of the exit rather than the courts.

It also protects the parts of the business Microsoft cannot afford to lose. A general reduction in force hits teams in proportion to their size, which means the AI groups would lose people too. A voluntary offer at level 70 and below, weighted toward managers, tends to remove coordination layers rather than the engineers building models.

The four countries were not chosen at random. India holds Microsoft’s largest employee base outside the United States, most of it in engineering and support centers in Hyderabad, Noida and Bangalore, where attrition is already high and replacing staff is cheap. Britain and Canada bring consultation obligations that make involuntary cuts slower and more visible than a program employees can decline.

The take rate is the number that decides how much of this is real. Buyouts in large technology companies typically draw between 10 and 20 percent of those eligible, according to analysts who track severance programs, which would put the outcome between 6,000 and 12,000 departures rather than 60,000.

At 15 percent, the savings would be material but not transformative, on the order of $1.5 billion to $2 billion a year before reinvestment, against capital spending that now runs above $200 billion on an annualized basis. The arithmetic suggests the program is as much about flattening the organization as about finding cash.

Headcount at the largest technology companies has been roughly flat for two years while their capital budgets have doubled. Amazon cut about 30,000 corporate roles in 2025 and has continued to spend on data centers. Meta has trimmed teams while raising tens of billions for infrastructure. Salesforce and Oracle have done variants of the same thing.

Shareholders have generally rewarded the combination. Investors treat capital spending as a claim on future revenue and headcount as a cost to be managed, a split that gives managements every incentive to fund the first by reducing the second.

The risk sits with execution. Microsoft’s cloud and Copilot businesses depend on people who sell to enterprises, and enterprise selling is a relationship business in which a departing account director takes a book of customers with them. Companies that cut too deeply into sales, analysts said, tend to discover the cost two quarters later in the form of slower bookings.

There is a timing consideration as well. Leaving on December 31 means the savings begin in the January quarter, which is the third quarter of Microsoft’s fiscal year, and show up in full in the following one. The announcement is therefore also a signal about the guidance the company will give in January, when it reports results for the quarter now under way.

Wall Street’s read will arrive with the January results. Microsoft shares have traded at a premium to the software index for three years on the strength of Azure growth, and any sign that the growth is being bought with capital rather than earned with products tends to compress that premium quickly. A buyout that lifts margins without slowing bookings would be read as discipline. One that lifts margins because the sales force left would not.

Coleman’s memo framed the decision as one about individual choice, and for the 60,000 people who received it, it is. For the company, it is a decision about what kind of business Microsoft intends to be: one that sells software through people, or one that rents computing by the hour and needs fewer of them.

Related Posts

  • October 27, 2026
  • 7 views
Nvidia Backs an Indian Startup That Sells Less GPU Time

A company whose software is designed to make graphics processors go further has raised money from the largest maker of graphics processors. That is less contradictory than it sounds. Simplismart,…

  • October 27, 2026
  • 7 views
Micron Says Its AI Memory Is Sold Out, and Buyers Will Pay More

Micron Technology has already sold everything it can make of the memory that artificial-intelligence systems need most, and its chief executive says the shortage will outlast most corporate planning cycles.…