Meta has signed an agreement with Amazon Web Services (AWS) to deploy AWS Graviton processors at scale to support next-generation AI workloads.
The company will initially use tens of millions of Graviton cores and can increase deployment as AI demand grows.
Access deeper industry intelligence
Experience unparalleled clarity with a single platform that combines unique data, AI, and human expertise.
learn more
This move expands the long-standing partnership between Meta and AWS. Amazon says this reflects the growing importance of CPU-intensive infrastructure for agent AI, which requires real-time inference, code generation, complex search processes, and task orchestration.
Meta relies on the Graviton5 chip to power agent AI operations. This requires an infrastructure that can handle billions of interactions and coordinate multi-step tasks.
By adopting Graviton5, Facebook owners aim to meet their code generation, inference, and model training demands on a platform designed for high computational power.
Santosh Janardhan, Head of Metainfrastructure, said: “AWS has been a trusted cloud partner for many years, and extending to Graviton will enable us to run CPU-intensive workloads behind agent AI with the performance and efficiency we need at our scale.”
The Graviton5 processor has 192 cores and five times larger cache than its predecessor, and AWS says it reduces core communication latency by up to 33%.
The chip runs on the AWS Nitro System, allowing direct access to the hardware while maintaining standard AWS services such as Elastic Network Adapter and Elastic Block Store.
Graviton5 instances also use Elastic Fabric Adapters. This enables low-latency, high-bandwidth communication between distributed AI workloads. This is a complex and large-scale coordination technical requirement.
The chip was developed entirely in-house by AWS and is said to deliver up to 25% better performance than previous models, with a focus on energy efficiency as well. They are manufactured using 3-nanometer chip technology, a process that is expected to improve computing efficiency and reduce energy usage.
“Meta’s expanded partnership to deploy tens of millions of Graviton cores shows what happens when you combine purpose-built silicon with the full AWS AI stack to power next-generation agent AI,” said Amazon Vice President and Distinguished Engineer Nafea Bshara.
The partnership builds on Meta’s use of existing AWS services, including Amazon Bedrock, and forms part of a broader strategy to equip its infrastructure for emerging AI scenarios and agent systems.
