According to an internal memo from Chief People Officer Janelle Gale, the restructuring breaks down into three parts:
- ~8,000 layoffs scheduled for May 20
- ~7,000 mandatory reassignments into four new AI-focused units
- ~6,000 open positions closed
That's nearly 15,000 roles affected across a 78,000-person company. Maher Saba, Meta's VP of Reality Labs and head of Applied AI Engineering, wrote in a separate memo obtained by Reuters that AI is the company's highest priority and that "the transfers aren't optional." Employees have started calling it being "drafted."
The company Zuckerberg wants to build looks very different from the one that exists today. On Meta's Q4 2025 earnings call, he said that "projects that used to require big teams now be accomplished by a single very talented person." Gale's memo describes a leaner, flatter company where smaller teams own more and move faster. In some of the new AI units, a single manager oversees up to 50 engineers. The typical ratio in tech is somewhere around 1 to 12.
Meta expects to spend $125 to $145 billion in capex this year, nearly double 2025, with the vast majority going to AI infrastructure. Gale framed the layoffs in her memo as a way to offset those investments.
Internally, Meta claims a 30% increase in output per engineer since early 2025, with power users of its AI coding tools seeing gains as high as 80%. Independent research paints a much more modest picture:
- METR, a nonprofit that evaluates AI capabilities, ran experiments showing that coding tools initially slowed experienced developers by about 19%. A follow-up study showed improvement, but gains for new participants were small and statistically uncertain.
- DX, a developer analytics firm, found that increased AI tool adoption corresponded to just a 7.76% bump in actual engineering output. Typical gains across most organizations land between 5% and 15%.
Meta is making the loudest version of a bet that's becoming common across tech. Coinbase, Block and Airbnb have all made similar moves this year, flattening management layers and shrinking teams on the same premise: AI fills the gap where headcount used to.
Meta is making the most aggressive corporate wager that AI can replace human headcount at scale, but the productivity numbers behind it are two to five times what independent research has measured anywhere else. If Meta's internal tools really are that far ahead, this restructuring becomes the template every large company copies within two years. If the numbers are aspirational, 15,000 people just got reorganized around a premise that hasn't been proven yet. Every other tech CEO is watching. Very few seem to be questioning the math.
