Meta has signed a supply agreement with Amazon for their 'AWS Graviton' AI chip, which will serve as the foundation for building next-generation AI agents.
On April 24, 2026, Amazon announced that it had signed an agreement to supply AWS Graviton to Meta. This agreement will make Meta one of the world's largest AWS Graviton customers.
Meta signs agreement with AWS to power agentic AI on Amazon's Graviton chips
In another wild turn for AI chips, Meta signs deal for millions of Amazon AI CPUs | TechCrunch
https://techcrunch.com/2026/04/24/in-another-wild-turn-for-ai-chips-meta-signs-deal-for-millions-of-amazon-ai-cpus/
Meta has announced that it has entered into an AWS Graviton supply agreement, expanding its long-standing partnership with Amazon to build next-generation AI agents. As part of this agreement, tens of millions of AWS Graviton cores will be supplied initially.
Regarding this agreement, Amazon explained, 'It reflects a shift in how AI infrastructure is built. While GPUs remain essential for training large-scale models, the rise of AI agents has led to a surge in demand for CPU-intensive workloads such as real-time inference, code generation, search, and multi-step task orchestration. Graviton 5 is specifically designed for these workloads, and Meta gains the processing power to run these workloads at scale and efficiently.' It should be noted that Amazon's AWS Graviton uses ARM-based CPUs, not GPUs.
AWS Graviton supports a variety of Meta's internal workloads and also plays a role in supporting Meta's AI development. These tasks require an infrastructure that can handle billions of interactions while coordinating complex, multi-stage agent workflows, making AWS Graviton a perfect fit.
The AWS Graviton 5, which Amazon will supply to Meta, features 192 cores and a cache five times larger than its predecessor, successfully reducing inter-core communication latency by up to 33%. This translates to faster data processing and wider bandwidth, critical requirements for AI agents that need to perform continuous inference and execute multi-step tasks.
AWS Graviton is built on the AWS Nitro System , using dedicated hardware and software to deliver high performance, high availability, and high security. The AWS Nitro System provides bare metal instances with direct hardware access, as well as familiar Elastic Network Adapters (ENA) and Amazon Elastic Block Store (Amazon EBS) devices, allowing Meta to run its own virtual machines without compromising performance.
Furthermore, Graviton5 instances also support Elastic Fabric Adapters (EFA), enabling low-latency, high-bandwidth communication between instances. This is essential for Meta's AI agent workloads, which require distributing large-scale tasks across multiple processors and having them work together.
Nafea Bhushara, Vice President and Principal Engineer at Amazon, said, 'This isn't just about chips. It's about providing our customers with the infrastructure, data, and inference services to build AI that can understand, predict, and efficiently scale with billions of people around the world. Our expanding partnership with Meta and the deployment of tens of millions of Graviton cores demonstrate how combining custom-designed silicon with the entire AWS AI stack can enable the next generation of AI agents.'
Santosh Janardan, Head of Infrastructure at Meta, commented, 'Diversifying computing resources is strategically essential as we expand the infrastructure that supports Meta's AI business. AWS has been a trusted cloud partner for many years, and the expansion to AWS Graviton will allow us to run the CPU-intensive workloads behind our AI agents with the performance and efficiency we need at our scale.'
More recently, Anthropic, the developer of Claude, announced an expansion of its strategic partnership with Amazon and an investment to secure AI chips, including AWS Graviton.
Anthropic announces investment of over $100 billion to secure Amazon's AI chips, and Amazon also invests up to $25 billion in Anthropic - GIGAZINE
Related Posts:
