← Back to feed

Meta has set a date. The first wave of layoffs — approximately 8,000 employees, roughly 10% of the company’s global workforce — is scheduled to begin May 20, 2026. Further cuts are planned for later in the year.

Three sources familiar with the plans confirmed the timeline. The scope and date have been reported by the New York Post, Financial Express, and Global Banking & Finance Review, citing independent sourcing.

A Capital Reallocation, Not a Restructuring

Meta has been explicit with investors and internally about the logic: money that once funded broad headcount is being redirected into AI infrastructure. The company committed $65 billion in capex for 2026, the majority of which is earmarked for GPU clusters, custom silicon, and data centre buildout.

The arithmetic is stark. Meta has already locked in:

• $21B to CoreWeave through 2032 for AI cloud compute
• 1GW+ MTIA deal with Broadcom through 2029 for custom inference chips
• 400-petaflop MTIA 400 deployment targeting NVIDIA for GenAI inference at scale

At $200,000–$300,000 per employee in fully-loaded annual cost, 8,000 jobs represents roughly $2B in annual savings — enough to fund several hundred thousand additional H100-equivalent GPU months, or years of contract with a hyperscaler.

Why Now

Meta’s timing tracks with its model development cycle. Llama 4 Maverick and Scout are in production. A frontier Llama 5 generation is in training or pre-training. Large model runs require large compute, and that compute needs to be booked, built, or contracted well ahead of when it is actually consumed.

The headcount reduction also pre-empts any perception of inefficiency as Meta increasingly automates internal tooling with its own models. Mark Zuckerberg flagged in early 2026 that AI coding agents would handle a meaningful share of software engineering tasks by the end of the year. Maintaining the same headcount while deploying internal agents would be difficult to justify to investors.

Precedent and Pressure

Meta is not alone. The pattern — shrink engineering headcount while increasing compute and model investment — has been visible across the industry since late 2025. What makes Meta’s move notable is scale. Eight thousand in a single announced wave, at a company that positioned itself as a leading AI-first employer, sets a benchmark other companies will reference.

The additional rounds expected later in 2026 suggest this is structural, not a one-time correction. Analysts will watch which teams are cut: whether it is operations and support (expected) or whether it touches core product engineering (which would signal deeper AI-for-engineering conviction than previously signalled).

The first pink slips go out in 32 days.