Business

Meta Signs Billion-Dollar Deal With Amazon

By Erica Coleman · April 24, 2026

Meta has signed a multiyear deal with Amazon Web Services to use billions of Amazon’s custom-built Graviton5 chips to power its artificial intelligence operations — a partnership an AWS executive told Reuters is worth billions of dollars and one of the largest AI infrastructure commitments announced this year.

The deal, announced Friday, starts with tens of millions of Graviton cores deployed across Meta’s systems, with the option to expand as the company’s AI needs grow. Graviton5 chips deliver up to 25% better performance than the previous generation, Amazon said, and are purpose-built for the kind of high-volume, real-time workloads that define Meta’s AI operations — search, recommendation systems, code generation, and the multi-step reasoning tasks that power its growing suite of AI agents across Facebook, Instagram, and WhatsApp.

The partnership reflects a broader shift in how the technology industry is building AI infrastructure. Graphics processing units — the chips that Nvidia dominates — remain essential for training large AI models from scratch. But as companies move from training models to actually deploying them at scale for billions of daily interactions, a different kind of chip is needed: one optimized for fast, efficient, continuous processing rather than the burst-intensive demands of model training. That is exactly what Amazon’s Graviton line is designed for.

For Amazon, the deal is a major commercial validation of a chips strategy it has been building for years largely in the shadow of Nvidia. AWS chief executive Andy Jassy disclosed this week that Amazon’s chips business — spanning Graviton, Trainium, and Nitro — now runs at an annual revenue rate of more than $20 billion, growing at triple-digit percentages year over year. He said demand has been so intense that two large AWS customers had already asked to purchase all of Amazon’s available Graviton capacity in 2026 — requests Amazon declined.

Meta becoming one of the largest Graviton customers in the world signals that the company is accelerating its push to reduce dependence on Nvidia, which has faced supply constraints, pricing pressure, and increasing scrutiny over its dominant market position. Nvidia’s chips currently power the majority of large-scale AI training across the industry.

The deal also has a competitive dimension that extends beyond chips. By deepening its relationship with AWS, Meta is building closer ties to the same cloud infrastructure platform that Anthropic — one of its most direct competitors in AI — relies on for much of its computing. Anthropic signed a 10-year, $100 billion commitment to run its workloads on AWS earlier this month, with Amazon agreeing to invest an additional $5 billion into the company.

For everyday users of Meta’s platforms, the practical effect of this deal will be invisible — better AI recommendations, faster responses from Meta AI, and smoother performance of the agentic features the company has been rolling out across its apps. The infrastructure shift is meant to make those experiences more reliable and less expensive to run at the scale of three billion daily active users.