Mark Zuckerberg told Meta staff at an internal town hall on Thursday that AI agent development has not accelerated in the way executives had expected, months after the company laid off roughly 8,000 employees and reassigned another 7,000 into AI-focused groups. Meta is on pace to spend as much as $145 billion on AI infrastructure this year, and Zuckerberg conceded the upside of the restructuring has not yet materialized. He told employees he expects improvements over the next three to six months.
The 8,000 cuts amounted to roughly 10% of Meta's corporate workforce. The 7,000 reassignments moved engineers and other staff into several AI groups, including one known internally as Agent Transformation. Zuckerberg framed both moves earlier this year as a bet on speed — that Meta needed to rewire its org chart around AI before competitors did.
At Thursday's meeting, Zuckerberg addressed the layoffs directly, saying they were not as clean as they should have been. He acknowledged that some of the reassignments and cuts landed on employees whose roles could have been preserved under a more deliberate restructuring.
Key facts
- 01Zuckerberg told Meta staff at a Thursday town hall that AI agent progress has not accelerated as executives expected.
- 02Meta laid off roughly 8,000 employees earlier this year, about 10% of its corporate workforce.
- 03Another 7,000 employees were reassigned into AI groups, including one internally called Agent Transformation.
- 04Meta is expected to spend as much as $145 billion on AI infrastructure this year.
- 05Zuckerberg said he expects improvements from Meta's AI investments in the next 3 to 6 months.
The stated rationale for the cuts was speed. Top officials at Meta, Zuckerberg said, were worried the company would not adapt fast enough to the changing shape of the tech industry as AI models and agents matured.
That worry drove the pace, and by Zuckerberg's own account this week, the pace outran the results.
The AI-focused company structure has not, in his words, come to fruition yet. That is a striking admission from a CEO who spent much of the last year telling investors that Meta's AI reorganization would compound returns across the ad stack, the Llama model line, and the company's consumer products.
The $145 billion infrastructure budget is the largest single-year AI capex commitment Meta has ever made, and among the largest in the industry. Most of that spend flows into data centers, Nvidia GPUs, custom silicon, and networking gear to support both training runs and agent inference at scale.
Meta's AI unit has also drawn internal criticism in recent months, with several reports describing the environment inside the reorganized division as punishing for the engineers assigned there. That backdrop makes Zuckerberg's acknowledgment more pointed: the org paid a morale cost for a payoff that has not yet arrived.
The three-to-six-month window Zuckerberg offered is short by AI industry standards, where model generation cycles often run twelve months or longer and agent reliability improvements have historically arrived in small increments rather than step changes. Whether Meta can compress that timeline depends on internal model progress that neither Zuckerberg nor the company has publicly benchmarked against competing agent systems from Anthropic, OpenAI, or Google.
The gap between AI capex and AI revenue is now the defining tension at every hyperscaler, and Meta is stating it out loud. Spending $145 billion in a single year to accelerate a technology that your own CEO says has not accelerated as expected is a bet that the curve bends soon — and if it doesn't bend inside the three-to-six-month window Zuckerberg gave his staff, the pressure to justify both the layoffs and the infrastructure bill will land squarely on Meta's next earnings cycle.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
