Meta signs deal for millions of Amazon AI CPUs, shifting cloud dynamics
On Friday, Amazon revealed that Meta has entered into an agreement to utilize millions of AWS Graviton chips, which are ARM-based CPUs rather than GPUs, to meet its expanding AI requirements. This partnership signifies a transformation in AI infrastructure, as CPUs are adept at managing computationally demanding tasks such as real-time reasoning and code generation, unlike GPUs that are utilized for training large models. AWS has tailored its latest Graviton for AI computation. This development marks a return of Meta's investments to AWS following a $10 billion, six-year contract with Google Cloud signed last August. The announcement coincided with the conclusion of the Google Cloud Next conference. While Amazon produces its own AI GPU, Trainium, Anthropic recently secured a significant supply of these chips through a $100 billion, 10-year agreement, with Amazon also investing an extra $5 billion in Anthropic. The Meta collaboration enables Amazon to promote its proprietary CPUs in competition with Nvidia's new ARM-based Vera CPU, designed for AI workloads. Amazon CEO Andy Jassy has recently voiced criticism of Nvidia and Intel in his shareholder correspondence, highlighting the superior price-performance ratios for AI. This puts additional pressure on Amazon's internal chip development team, which TechCrunch had the opportunity to visit last month.
Key facts
- Meta signed a deal to use millions of AWS Graviton chips for AI workloads.
- AWS Graviton is an ARM-based CPU, not a GPU.
- AI agents create compute-intensive workloads like real-time reasoning and code writing.
- AWS designed its latest Graviton specifically for AI compute.
- Meta previously signed a $10 billion, six-year deal with Google Cloud in August.
- AWS announced the deal as Google Cloud Next conference wrapped up.
- Anthropic agreed to spend $100 billion over 10 years on AWS Trainium chips.
- Amazon invested an additional $5 billion in Anthropic, bringing total to $13 billion.
Entities
Institutions
- Meta
- Amazon
- AWS
- Google Cloud
- Microsoft Azure
- Anthropic
- Nvidia
- Intel
- TechCrunch
Sources
- Quartz —
- TechCrunch AI —