Meta and AMD Ink Multi-Year Deal to Deploy Up to 6 GW AI Chips
Meta and AMD Ink Multi-Year Deal to Deploy Up to 6 GW AI Chips
Meta teams with AMD on a multi-year plan to deploy up to 6 GW of AI chips, combining Instinct GPUs and EPYC CPUs for Meta's next-gen data centers, starting in H2 2026.
Meta Platforms and AMD announced a multi-year partnership to deploy up to 6 gigawatts of AI chips, powering Meta's next generation data centers. The deal marks one of the largest AI infrastructure commitments publicly disclosed and signals Meta's push to diversify its compute sources as competition in AI hardware intensifies.
Under the agreement, Meta will deploy several generations of AMD Instinct graphics processors along with EPYC central processors. The shipments are planned to start with the first gigawatt of capacity in the second half of 2026. The systems will run on AMD’s Helios rack-scale architecture and will use a custom GPU based on the MI450 design tailored for Meta’s workloads.
AMD CEO Lisa Su described the collaboration as a "multi-year, multi-generation" effort aligned with delivering high-performance, energy-efficient infrastructure. Meta CEO Mark Zuckerberg said the partnership supports long-term AI ambitions, including the development of increasingly powerful models and a move toward personal superintelligence.
The deal also expands Meta’s use of AMD’s EPYC CPUs, with Meta slated to be a lead customer for the Venice processor and the Verano follow-on. By diversifying beyond a single chip supplier, Meta aims to build the scale needed for inference at unprecedented levels while keeping an eye on cost and energy efficiency as AI models grow larger.
