Meta Forms New AI Units, Plans to Lay Off 8,000 on Wednesday

At 4 a.m. local time on Wednesday, 8,000 Meta employees will receive layoff notices via email.

The company's latest move comes in two phases: last week, it internally reassigned 7,000 remaining employees into four new AI units; this Wednesday, it will cut roughly 10% of its total workforce. Combined, the moves affect about 20% of staff—a significant scale in the tech industry.

Bloomberg and Reuters reviewed an internal memo from Meta CPO Janelle Gale, who wrote:

"many leaders will announce org changes... operate with a flatter structure with smaller teams"

In short, management layers will be reduced and teams will be downsized.

What the Four New AI Units Look Like

According to internal documents, the units are named:

  • Applied AI Engineering (AAI)
  • Agent Transformation Accelerator (ATA) XFN
  • Central Analytics
  • Enterprise Solutions

The memo states these groups will "develop AI agents that can autonomously complete tasks currently performed by humans." In other words, they aim to build AI that can replace employees—many of whom still sit in nearby cubicles.

The teams are organized under an "AI native design" principle. Gale did not elaborate on the term, but combined with "flatter structure + small teams + fewer managers," the meaning is clear: remove a layer of managers, let agents replace human labor, and retain those who can actually write code and manage agents.

Zuckerberg Is Also Calculating the Numbers

Meta had 77,986 employees at the end of March. The 8,000 layoffs, 7,000 reassignments, and closure of 6,000 unfilled positions together affect roughly 20% of the workforce.

On the capital expenditure side: Meta previously guided 2026 spending at $115–135 billion, but the latest figure has risen to $125–145 billion. The entire increase is being poured into AI and data centers.

In other words, Meta is cutting people while pouring money into computing power. Zuckerberg made this clear on the Q1 earnings call, saying the organization needs to be "compact enough for AI to accelerate it." The implication is that the company itself must adapt to agents as a new type of "employee," while old hierarchies would only slow things down.

Engineers Signed a Petition

Notably, over 1,000 employees signed an internal petition opposing Meta's mouse-tracking software installed on company computers.

This detail adds a subtle tension to the restructuring. On one hand, the company uses employee computer activity as training data (as previously reported); on the other, it moves those same employees into new AI agent groups while laying off others. Employees are becoming both a source of training data and targets for replacement.

Why This Matters

In previous years, big tech layoffs were mostly about cutting redundancy and controlling costs. Meta's approach is different. The names of the new units—Applied AI Engineering, Agent Transformation Accelerator—make the direction clear: this is not simply about saving money by cutting jobs, but about using AI agents to directly absorb the work of certain positions, leaving the remaining staff to either build agents or manage them.

Mustafa Suleyman's prediction two weeks ago that "white-collar jobs will be eaten within 18 months" is now being demonstrated by Meta in real time. The difference is that Suleyman drew a timeline for all companies, while Meta is providing a specific list of layoffs on Wednesday.

After 4 a.m. next Wednesday, 8,000 people will receive emails, and 7,000 will change departments. For the first time, AI replacing human labor is happening inside a major tech company on a week-by-week countdown.

Sources: Meta Moves 7,000 Workers Into AI Roles Ahead of Job Cuts (Bloomberg); Meta lays out plans for May 20 layoffs, restructuring (Yahoo Finance / Reuters); Days before 8,000 employees will be laid off, Meta has also decided to reassign 7,000 employees to focus on AI (Sherwood News); CocoLoop