Zuckerberg Admits Meta AI Agent Progress Has Stalled After $14 Billion Bet

Quick Facts

• Zuckerberg told employees at a July 2 town hall that agentic AI progress over the last four months has not accelerated as expected.
• Meta shares fell nearly 5 percent on the news, wiping out most of the prior day’s 9 percent gain.
• The admission follows a restructuring that laid off 8,000 employees and shifted 7,000 others into AI-focused teams.

Meta Chief Executive Mark Zuckerberg told employees Thursday that the company’s AI agent ambitions have fallen short of expectations, according to a recording heard by Reuters. Speaking at an internal town hall, Zuckerberg said, “The kind of trajectory of the agentic development over at least the last four months hasn’t really accelerated in the way that we expected.”

The remarks sent Meta shares down nearly 5 percent, reversing most of a 9 percent gain from the prior session. The timing was blunt: Meta had just posted $26.8 billion in net income for the first quarter of 2026.

Zuckerberg said that when executives planned the restructuring in January and February, they were “super optimistic” about AI coding tools such as Anthropic’s Claude Code, and expected that enthusiasm to fuel faster progress inside Meta’s own products. It did not.

The restructuring he referenced was sweeping. Meta laid off roughly 8,000 workers, about 10 percent of its global workforce, in May. Simultaneously, the company moved approximately 7,000 employees into newly formed AI units, including the Applied AI Engineering group and the Agent Transformation Accelerator. Together, those changes affected close to one-fifth of the company’s total headcount.

Zuckerberg conceded at the town hall that he had made mistakes in the restructuring and said he would “almost certainly make more.” He added that the expected upside of the new structure has not “come to fruition yet,” though he said he expected improvements to appear within the next three to six months.

Meta CTO Andrew Bosworth, also present at the town hall, described employee morale as “maybe not the worst it’s ever been in 20 years here, but it’s probably up there.” More than 1,600 Meta employees have signed a petition protesting a program that monitors their clicks and keystrokes to collect AI training data. Workers assigned to the Applied AI unit, which absorbed around 6,500 engineers and product managers, have described the division to Wired as a “gulag.”

The spending behind this push is substantial. Meta is projected to spend between $125 billion and $145 billion on AI infrastructure in 2026, up from an earlier estimate of $115 billion to $135 billion. Across Amazon, Alphabet, Meta, and Microsoft, collective capital expenditure commitments for 2026 total between $650 billion and $725 billion, the largest single-year infrastructure investment in the industry’s history.

Meta made an aggressive bet on closing its AI gap last year. The company agreed to acquire a 49 percent stake in Scale AI for $14.8 billion and installed former Scale AI CEO Alexandr Wang, 28, as Meta’s first Chief AI Officer. Wang now leads Meta Superintelligence Labs, a team built from researchers recruited from OpenAI, Anthropic, Apple, and elsewhere. The team released its first model, Muse Spark, on April 8, 2026.

Holger Mueller of Constellation Research called Zuckerberg’s remarks unusually candid for a technology founder of his stature. “It’s surprising to see Meta admitting that its agentic AI efforts have come up against a brick wall, but it’s certainly not the only big tech company to have struggled with this,” Mueller told SiliconANGLE.

For software and technology executives watching this closely, Zuckerberg’s admission is a data point that matters. The gap between capital investment in agentic AI and reliable, deployable agentic systems remains wide, even at companies spending at this scale. The question for Meta, and for the industry, is whether the next three to six months deliver what the last four have not.

Read more: Mark Zuckerberg tells staff that AI agents haven’t progressed as quickly as he’d hoped