The remarks make explicit what Meta had previously framed only as ‘efficiency'. The May 20 layoffs of approximately 8,000 employees, roughly 10% of the company's 78,865-person workforce, are now positioned by the CEO himself as a direct trade-off against the AI infrastructure spending Meta unveiled at Wednesday's Q1 earnings.

Meta raised its full-year 2026 capex guidance to $125-$145 billion, up from $115-$135 billion, sending shares down 9% after hours. Additional layoffs are planned for the second half of 2026.

The AI productivity question Zuckerberg sidestepped

Zuckerberg's most carefully constructed line of the session was about whether AI productivity itself is driving the cuts.

“Getting everyone internally to use AI tools and getting to do the work more efficiently is not the thing that's driving layoffs,” he said, a denial that holds the line on the explicit AI-replaces-workers narrative his peers at Microsoft and Salesforce have flirted with.

He immediately added the qualifier: “We'll see how all this stuff trends.” He said the company would “be able to share more soon.”

The hedge matters because, on Meta's own previous statements, AI tools are reshaping engineering productivity at the company. Zuckerberg has publicly claimed that output per engineer has risen 30% since early 2025 driven by AI coding tools, and that power users have seen an 80% year-on-year increase.

Either those productivity gains are not yet meaningful enough to drive workforce decisions, or the framing for employees in a town hall on the eve of layoffs is being calibrated more carefully than the framing for analysts and investors. Both can be true.

Zuckerberg also offered an unusually candid admission about the longer horizon. “I wish that I can tell you that I have a crystal ball plan for the next, like, three years of how all this stuff is going to play out. I don't. I don't think anyone does,” he said.

The statement, made to a workforce that has now been told that further cuts are possible and that the company's optimal long-term size is unknown, is a striking departure from the corporate certainty that typically accompanies large restructurings.

The ‘two cost centres' framing is the cleanest version yet of an argument that has been made implicitly across the tech sector for the past 18 months.

Meta and Microsoft announced 23,000 combined job reductions in the same week, with the same underlying logic: companies are not cutting because they cannot afford their workforces, but because they have decided to redirect that capital to AI infrastructure instead.

Meta's 2025 net income exceeded $22 billion in the fourth quarter alone. The company is profitable. It is choosing to spend the money on machines.

Zuckerberg's remarks on Thursday turn the implicit into the explicit. The trade-off is not between revenue and expense; it is between two categories of expense. Compute infrastructure is the category Meta has decided to grow.

Personnel is the category it has decided to shrink. Both are paid for from the same pool of capital. Both, ultimately, generate value through the products Meta sells to advertisers. The bet is that compute generates more of that value per dollar than headcount.

The wider industry context makes the trade-off legible. Oracle eliminated up to 30,000 employees in March to fund its $156 billion AI infrastructure commitment. Atlassian cut 1,600 jobs and replaced its CTO with two AI-focused executives. Amazon cut 16,000 corporate jobs in January.

Block eliminated 4,000 roles. The tech sector has shed more than 95,000 jobs across 247 layoff events in 2026, and the explanations have converged. Companies that previously framed cuts as performance-driven, or as adaptations to slowing growth, are now stating the AI capital reallocation directly. Meta's town hall is the latest and clearest example of a CEO saying it on the record to his own employees.

The one missing piece

What Zuckerberg did not say, and what neither he nor Janelle Gale nor Susan Li would commit to, is whether the layoffs scheduled to begin May 20 represent the bottom of the cycle or a midpoint.

The CFO's acknowledgment that Meta does not know its optimal long-term size, the chief people officer's refusal to rule out further cuts, and the CEO's explicit linkage between capex and headcount together leave Meta's remaining 71,000 employees with a structural uncertainty: their continued employment is contingent on a capex line item that the company has been raising every quarter.

Meta's capex of $125–$145 billion in 2026 is roughly double its 2025 spend, and nearly all of it goes to data centres, GPUs, custom silicon, and the infrastructure supporting Llama models and Meta Superintelligence Labs. The Hyperion data centre in Louisiana alone is a $10 billion, five-gigawatt facility.

The Prometheus supercluster in Ohio is a one-gigawatt project coming online this year. Each new infrastructure commitment, on Zuckerberg's own logic, comes from the same capital pool that pays for headcount. The arithmetic of the town hall is not difficult to extend forward, even without a crystal ball.

Whether the trade-off works, whether the AI infrastructure being built generates returns commensurate with the human capital being eliminated, will be the defining question for Meta and for the broader Big Tech cohort over the next several years. The market's 9% sell-off after Wednesday's earnings suggests at least some investors have begun pricing in the risk that it does not.