Meta is spending as much as $145 billion on AI infrastructure this year, but its CEO just told employees the technology isn't delivering.
Mark Zuckerberg acknowledged at an internal town hall Thursday that AI agents - automated systems designed to execute tasks for users - have not progressed as quickly as he expected, according to a recording heard by Reuters. The trajectory of agentic development "hasn't really accelerated in the way that we expected" over the last four months, he said, adding that the company's bets on its new structure "haven't come to fruition yet."
The admission came alongside a broader mea culpa. Meta's sweeping reorganization, which eliminated about 10% of the global workforce and moved roughly 7,000 employees to AI-focused teams in May, was not as "clean" as it could have been, Zuckerberg said.
Executives miscalculated the timing of the changes, he added, and have since been seeking to moderate some of the organizational shifts without reversing course. When Meta began planning the restructuring in January and February, Zuckerberg said his top lieutenants were worried the company wasn't moving fast enough. At the time, executives were "super optimistic" about tools like Claude Code, the coding assistant from AI startup Anthropic - a remark that suggests Meta saw an outside product as a benchmark for its own AI ambitions.
The internal doubts help explain why Meta is pouring capital into infrastructure at a pace that outstrips most of its peers. The company's projected $145 billion AI spend this year accounts for roughly one-fifth of the more than $700 billion Big Tech is collectively deploying on the technology.
Zuckerberg told employees he expects Meta to begin seeing more meaningful returns from those investments within the next three to six months. A Meta spokesperson declined to comment on the town hall.
Separately, Chief Technology Officer Andrew Bosworth said during the same meeting that a review of a data security incident involving Meta's controversial mouse-tracking software found no employee data was included in AI training.
