Meta has signed a deal to adopt millions of Graviton CPUs developed by Amazon Web Services, taking into account rising AI demand, TechCrunch reported on April 24.

Graviton is an Arm-based CPU. GPUs remain essential for training large models, but CPU demand is also growing as AI agents handle compute-intensive workloads such as real-time inference, writing code, search and coordination of multistep tasks after model training. AWS's latest Graviton is designed to match such AI-related computing demand.

Meta signed a $10 billion, six-year deal with Google Cloud in August last year. Amazon announced the deal with Meta as soon as the Google Cloud Next event ended.

Amazon also has its own AI GPU, Trainium. Anthropic secured a significant volume of Trainium this month after agreeing to spend $100 billion on AWS over 10 years.

The deal is significant because Amazon has secured a major customer for its in-house CPU. Nvidia is also targeting the AI agent workload market by launching its Arm-based CPU Vera. Nvidia sells chips directly, while AWS provides access only through its cloud services.

Amazon Chief Executive Andy Jassy said in a shareholder letter earlier this month that companies want better price-performance for AI and that it intends to target the market with its in-house chips.