Meta cuts 8,000, signs AWS pact
- Meta told employees on April 23 it will cut about 10% of its workforce, or roughly 8,000 jobs, and cancel hiring for 6,000 open roles as it shifts spending toward artificial intelligence. - A day later, Amazon Web Services said Meta signed a multi-year agreement to deploy tens of millions of Graviton processor cores, making Meta one of AWS’s largest Graviton customers. - The back-to-back moves show Meta trimming labor while expanding rented AI compute after years of building its own stack. (aboutamazon.com)
Meta is cutting about 8,000 jobs and shelving 6,000 planned hires while signing a new Amazon Web Services chip deal to keep expanding its artificial intelligence systems. (cnbc.com) (aboutamazon.com) The layoffs amount to about 10% of Meta’s workforce, according to an internal memo reported by Bloomberg and CNBC. The cuts are set to begin May 20, 2026. (bloomberg.com) (cnbc.com) Meta also told staff it would stop filling 6,000 open roles that had been planned before the restructuring. CNBC said the company tied the move to its rising artificial intelligence spending. (cnbc.com) (techcrunch.com) On April 24, Amazon Web Services said Meta signed a multi-year agreement to deploy AWS Graviton processors at scale. The rollout starts with tens of millions of Graviton cores and can expand as Meta’s AI demand grows. (aboutamazon.com) AWS said Meta is now one of the largest Graviton customers in the world. GeekWire reported the agreement centers on Graviton5 processor cores running in Amazon data centers. (aboutamazon.com) (geekwire.com) Graviton chips are Amazon’s in-house central processors, built on Arm designs, and they handle general computing work rather than the graphics-heavy training jobs usually associated with Nvidia systems. AWS said Meta will use them for “CPU-intensive” agentic AI workloads. (aboutamazon.com) (networkworld.com) That matters because large AI products need more than graphics processors. They also need huge fleets of cheaper general-purpose chips to run orchestration, memory, retrieval, and serving tasks around the models. (networkworld.com) (aboutamazon.com) Meta has spent years building custom infrastructure, including its own MTIA accelerators, while also buying Nvidia hardware and using cloud services. The AWS agreement adds another supplier and gives Meta more rented capacity outside its own data centers. (networkworld.com) (aboutamazon.com) The company has been here before on headcount. Meta cut 21,000 jobs in 2022 and 2023 during Mark Zuckerberg’s “year of efficiency,” and this round shows the cost pressure from the AI buildout has not eased. (bloomberg.com) (nymag.com) The sequence is blunt: fewer people, more processors. Meta is shrinking payroll in one memo and expanding machine capacity in the next. (cnbc.com) (aboutamazon.com)
