Meta Taps AWS Graviton Chips to Handle Next-Gen AI Workloads

Meta is making a massive bet on Amazon’s custom silicon. The social media giant announced a deal to bring tens of millions of AWS Graviton processing cores into its compute portfolio, positioning itself as one of Graviton’s biggest customers worldwide.

These cores are the fundamental processing units inside CPUs that handle the heavy lifting for the next wave of agentic AI—autonomous systems that can reason through problems, plan actions, and execute complex tasks without constant human guidance.

The partnership extends Meta’s existing relationship with AWS, but this time the focus is on hardware diversification. By adding Graviton chips to the mix, Meta gains flexibility to match the right architecture to the right job, whether that’s powering inference at scale or running CPU-intensive data pipelines.

AWS Graviton5 processors, in particular, deliver faster data throughput and higher memory bandwidth—critical for AI systems that need to continuously reason and act in real time. Nafea Bshara, VP and Distinguished Engineer at Amazon, noted that the expanded collaboration is about more than chips: it’s about giving builders the infrastructure foundation to deploy AI that scales to billions of users.

Santosh Janardhan, Meta’s Head of Infrastructure, framed the move as a strategic imperative. “Diversifying compute sources is key as we scale AI,” he said. “AWS has been a trusted partner for years, and Graviton lets us run CPU-intensive agentic workloads with the performance and efficiency we need.”

The initial deployment will start with tens of millions of cores, with room to expand as Meta’s AI ambitions grow.

Source: Facebook