The financial terms of the agreement have not been disclosed. AWS Graviton is an ARM-based CPU designed to compete with Nvidia's offerings for handling AI agentic workloads. The Graviton5 chips feature 192 cores and a cache five times larger than the previous generation, reducing latency by up to 33 percent.
The deal matters because it validates custom silicon for AI infrastructure at enterprise scale and signals a broader industry recognition that agentic AI—requiring real-time reasoning, code generation, and multi-step task orchestration—demands CPU-intensive workloads beyond traditional GPU-focused strategies. For Meta, which has spent $48 billion for access to Nvidia GPUs for model training, this CPU investment represents a complementary strategy to handle the operational demands of deployed AI systems. The announcement also underscores Meta's commitment to infrastructure diversification amid intense competition for computational resources among tech giants. Attorneys tracking AI infrastructure consolidation and vendor lock-in issues should monitor whether this model becomes standard practice across the industry.