Synopsis
Meta Platforms has entered a multi-year agreement with Amazon Web Services to utilise tens of millions of AWS Graviton CPU cores for its artificial intelligence infrastructure. This strategic move will bolster Meta's capacity for CPU-intensive agentic AI workloads, enhancing performance and efficiency as the company scales its AI ambitions.Listen to this article in summarized format
The deal will make Meta one of the largest users of Graviton chips, according to an Amazon blog post published on Friday.
Graviton is Amazon’s own line of server processors used in its cloud computing platform. AWS says Graviton5, its latest generation processor, delivers up to 25% better compute performance than Graviton4.
Meta said the chips will be used for CPU-heavy workloads linked to agentic AI, which plan and carry out multi-step tasks. The company said the first deployment will involve tens of millions of cores, with room to expand as its requirements grow.
“As we scale the infrastructure behind Meta’s AI ambitions, diversifying our compute sources is a strategic imperative,” Santosh Janardhan, head of infrastructure at Meta, said in a statement. “Expanding to Graviton allows us to run the CPU-intensive workloads behind agentic AI with the performance and efficiency we need at our scale.”
AWS has been developing its in-house Graviton processors since 2018 and is now on its fifth generation. The chips are manufactured by Taiwan Semiconductor Manufacturing Company. With this, Meta is adding a very large pool of cloud computing capacity to handle AI-related processing.
“This isn’t just about chips; it’s about giving customers the infrastructure foundation, as well as data and inference services, to build AI that understands, anticipates, and scales efficiently to billions of people worldwide,” said Nafea Bshara, vice president and distinguished engineer at Amazon.
The company has previously entered into large chip supply agreements with Nvidia and Advanced Micro Devices, and has also worked with Arm Holdings on CPU development.
