Meta's 6,500-Person AI Unit Erupts in Public Meltdown as Fourth Reorganization Looms

An employee-only presentation at Meta exploded this week when an unidentified worker seized the livestream to deliver an expletive-laden rant about being the company's bitch, then demanded presenters message a senior AI executive to tell him he's a piece of shit. One presenter covered their face with their hands, according to a witness who spoke to Wired. The meeting leaders muted the interruption and pressed on, though employees flooded the stream chat calling the start spicy.

The incident, which played out before thousands of viewers on a company-wide call, landed like a grenade inside Meta's Applied AI team. That unit was formed just three months ago in March 2026 to support researchers at Meta Superintelligence Labs, but sources describe it as already nearing a breaking point. TechCrunch reports the 6,500-person team is on the verge of revolt, with morale cratered by relentless layoffs and a strategy that shifts faster than employees can execute it.

Mark Zuckerberg has overhauled Meta's AI structure repeatedly, and the pace is accelerating. The New York Times reports Zuckerberg is again reshaping the company's AI efforts, while Reuters flags a planned fourth restructuring in just six months. Each reboot tears up reporting lines, scatters existing projects, and forces staff to re-prove their value under new leadership, according to multiple accounts.

The whiplash has left the Applied AI team without stable direction. One employee told Wired that workers feel like they are building and rebuilding the same infrastructure endlessly, with each new org chart rendering the last quarter's work obsolete. Puck News, in a piece titled Why Did Meta Reorganize Its AI Unit Again?, notes that the constant shuffling has made it impossible for engineers to ship or for managers to build coherent roadmaps. The flat management structure, which Leaddev reports has been taken to an extreme inside the new unit, strips away traditional hierarchy without replacing it with clear decision-making authority.

Meta's AI investment binge sits at the center of the unrest. The company has executed wave after wave of layoffs in recent years, cuts that TechCrunch says have only accelerated as billions get funneled into AI infrastructure. Workers who remain describe a soul-crushing environment where job security vanished and workloads swelled to fill empty seats.

The Applied AI team was supposed to be a prestige assignment, a chance to work on the technology Zuckerberg has declared central to Meta's future. Instead, it has become a gulag, in the words of one engineer who spoke to TechCrunch. Morale inside Meta's broader AI operation was already brittle, Wired notes, after a string of departures from Meta Superintelligence Labs earlier this year sent researchers to rivals including OpenAI. The public meltdown this week suggests the pressure has reached a level where private complaints are spilling into open revolt.

The chaos inside Applied AI threatens to derail the very products Zuckerberg is betting the company on. Meta needs the unit to ship models, build infrastructure, and keep pace with OpenAI and Google in a race where months of delay can mean permanent disadvantage. A team too paralyzed by dysfunction to execute coherently cannot deliver that speed.

The timing is especially fraught because Meta's AI strategy depends on blending research from Superintelligence Labs with production engineering from Applied AI. If the two groups cannot collaborate, the research sits unpublished and the products stall. Reuters' report of yet another restructuring suggests Zuckerberg sees the problem but believes the solution is another org chart, not a change in how the work gets done. That diagnosis worries employees who have lived through the previous three.

OpenAI, Google, and Anthropic are all recruiting aggressively from Meta's AI talent pool, and the public dysfunction gives them fresh material. Wired previously reported researchers leaving Meta Superintelligence Labs for OpenAI, and the current atmosphere inside Applied AI will likely accelerate that outflow. For a field where a handful of researchers can swing a company's entire model program, even modest attrition hurts.

The broader risk is to Meta's reputation as an AI destination. The company has spent heavily to position itself as a serious rival to OpenAI, including open-sourcing its Llama models and building enormous compute clusters. Internal meltdowns undermine that narrative, particularly when they suggest the infrastructure is sound but the organization around it is broken. Investors and partners who might have seen Meta as a stable horse to back may reconsider if the human engine keeps seizing up.

Zuckerberg faces a choice between another top-down restructure, which Reuters says is already planned, and a genuine attempt to fix the culture that keeps producing dysfunction. History at Meta suggests he will choose the reorganization. The pattern, Puck notes, is to shuffle leaders, rename teams, and declare a fresh start, a move that addresses optics without touching root causes.

Employees are not optimistic. The worker who hijacked the livestream chose public humiliation over any internal channel, suggesting they believed formal routes were blocked or pointless. If that sentiment spreads, Meta's AI ambitions will face a talent drain that no amount of compute spending can fix. The next few weeks, as the fourth restructuring lands, will show whether Zuckerberg recognizes that organizations are made of people, not just org charts.