Meta signs multi-year, multi-billion dollar deal to use AWS Graviton5 CPU chips
Updated
Updated · Reuters · Apr 24
Meta signs multi-year, multi-billion dollar deal to use AWS Graviton5 CPU chips
12 articles · Updated · Reuters · Apr 24
Meta will use tens of millions of Graviton5 CPU cores from Amazon Web Services, each chip containing 192 cores, in a deal spanning several years and valued at billions of dollars.
The agreement aims to scale Meta's AI infrastructure and diversify its compute sources, with CPUs playing a growing role in AI workloads after model training, complementing GPUs from firms like Nvidia.
AWS has developed its own CPUs since 2018, now on the fifth generation, and competes with Nvidia and AMD. Meta previously signed major chip deals with Nvidia, AMD, and Google Cloud.
With Meta choosing AWS's custom chips, is Nvidia's long-held reign over the AI market finally at risk?
AWS just poached Meta from Google. Can Google's cheaper AI platform win back big tech's business?
Meta is using AWS chips while building its own. Who is the real winner in this deal?
As AI's energy demand soars, what is the hidden environmental cost of this tech arms race?
Will the next global tech boom be decided not by silicon, but by access to electricity grids?
After the 2025 'silicon shock,' are tech giants truly prepared for the next supply chain crisis?
How Meta's Massive Graviton 5 Deployment Signals ARM's Rise to 90% of AI Server CPUs
Overview
In April 2026, Meta announced a major deal with AWS to deploy tens of millions of Graviton 5 ARM-based processor cores, marking a strategic shift to meet its massive AI computational demands. This move deepens Meta's partnership with AWS and reduces reliance on other cloud providers. Graviton 5's energy efficiency and high performance enable cost savings and accelerate Meta's AI development, especially for complex agentic AI workloads. Alongside this, Meta pursues a multi-vendor strategy, collaborating with Nvidia and developing custom CPUs to optimize performance and supply chain resilience. This deal also reflects a broader industry shift toward ARM processors, projected to power most AI servers in the near future.