Meta Signs Major AI Chip Deals With Google and AMD

by Justin Erickson

Meta is expanding its AI infrastructure through separate multiyear, multibillion dollar agreements with both Google and AMD. Meta will rent Google’s Tensor Processing Units (TPUs) via Google Cloud to run some of its large-scale AI workloads rather than relying only on in-house components. Meta will also deploy up to 6 gigawatts of AMD Instinct GPUs over several years, which “…equals the annual electricity usage of approximately 5 million U.S. households…” – built on AMD’s Helios rack-scale architecture and paired with 6th Gen EPYC CPUs. AMD said the agreement is worth “significant double-digit billions of dollars per gigawatt” in potential revenue and issued Meta a performance-based warrant for up to 160 million AMD shares – driving AMD’s stock up roughly 8.8%. Taken together with Meta’s previously disclosed deal with Nvidia, this highlights their new “portfolio-based” approach to AI hardware, spreading risk and investment across multiple suppliers.