Startups & Funding

In another wild turn for AI chips, Meta signs deal for millions of Amazon AI CPUs

Meta ditches GPU-only strategy, betting on Amazon's ARM chips for AI agents.

Deep Dive

In a major strategic shift, Meta has signed a deal to use millions of Amazon's AWS Graviton chips to power its growing AI infrastructure, Amazon announced Friday. The Graviton is an ARM-based CPU, not a GPU, signaling a key pivot in AI computing: while GPUs remain essential for training large models, the rise of AI agents—which handle compute-intensive tasks like real-time reasoning, code generation, and multi-step coordination—is driving demand for specialized CPUs. AWS's latest Graviton was designed specifically for these agentic workloads, offering better price-performance for inference and orchestration.

This deal brings a significant portion of Meta's cloud spending back to AWS, after Meta signed a $10 billion, six-year deal with Google Cloud last August. Amazon timed the announcement to coincide with the close of Google Cloud Next, a subtle jab at its rival. The move also intensifies the chip competition: Amazon's Graviton goes head-to-head with Nvidia's new Vera CPU, which is also ARM-based and targets AI agents. Meanwhile, Amazon's own AI GPU, Trainium, is largely committed to Anthropic, which agreed to spend $100 billion over 10 years on AWS. With CEO Andy Jassy publicly vowing to win on price-performance, the pressure is on Amazon's chip team to deliver.

Key Points
  • Meta will use millions of AWS Graviton CPUs (ARM-based) for AI agent workloads, not GPUs.
  • The deal shifts Meta's cloud spending back to AWS after a $10B Google Cloud deal in August.
  • Amazon's Graviton competes with Nvidia's Vera CPU, both designed for agentic AI tasks.

Why It Matters

AI agents are reshaping chip demand, and Amazon's CPU win over Meta signals a new battleground beyond GPUs.