Amazon Web Services (AWS) today signed a multi-year, multi-billion-dollar AI infrastructure deal with Meta, the parent company behind Instagram, Facebook, WhatsApp, and others. The core part of the deal is that Meta will be using millions of AWS Graviton5 cores to power its agentic AI and compute requirements. The first deployment will start with tens of millions of Graviton cores and will be expanded further based on the requirement.
AWS began its Graviton journey in 2018 with the first-generation Arm-based Graviton processor, built by Annapurna Labs, which Amazon had acquired earlier. The first chip powered A1 EC2 instances and was mainly aimed at scale-out and cost-sensitive workloads.
More recently, AWS announced Graviton5, its next-generation CPU for newer EC2 instances, as part of its broader push to make its custom silicon a major advantage against rivals such as Microsoft Azure and Google Cloud. Since 2024, both Microsoft Azure and Google Cloud have also announced their own custom ARM CPUs to better compete against the AWS Graviton series.
In fact, AWS recently claimed that Graviton now accounts for roughly half of the new EC2 CPU capacity added over the past two years, showing how quickly its Arm-based custom silicon has moved from a small cost-optimization option to a mainstream compute platform inside AWS.
The latest generation Graviton5 chip features 192 cores and a five times larger cache than the previous generation to improve inter-core communication by 33%. Since Graviton is built on the AWS Nitro System, Meta can run its own virtual machines without performance compromises. Also, Graviton5 instances support the Elastic Fabric Adapter (EFA) for low-latency, high-bandwidth communication between instances, which would support Meta’s agentic AI workloads spread across different processors.
Santosh Janardhan, Head of Infrastructure, Meta, said the following:
“As we scale the infrastructure behind Meta’s AI ambitions, diversifying our compute sources is a strategic imperative. AWS has been a trusted cloud partner for years, and expanding to Graviton allows us to run the CPU-intensive workloads behind agentic AI with the performance and efficiency we need at our scale.”
Earlier this week, AWS signed a massive AI infrastructure deal with Anthropic to develop 5 GW of capacity for training and deploying Claude. Out of this 5 GW capacity, some new Trainium2 capacity will be coming online in the first half of 2026, while another 1 GW total of Trainium2 and Trainium3 capacity will be coming online by the end of 2026.
0 Comments
Load the comments and join the conversation!
Read the comments, ask the editors questions, show respect and join the conversation.