Meta buys AI compute power from AMD in $100bn deal

Meta and AMD have agreed a partnership for the social media company to power its AI infrastructure with up to 6GW of AMD Instinct GPUs, the silicon computing technology used to support modern AI models. The agreement extends beyond hardware to include systems and software, enabling vertical integration across the entire infrastructure.

AMD has issued Meta a performance-based warrant for up to 160 million shares, structured to vest as specific milestones associated with GPU shipments. The deal has been valued at more than $100 billion, which could mean that Meta owns up to 10% of AMD’s stock, as reported by the Wall Street Journal.

The first hardware shipments will begin in the second half of this year and will be built on the Helios rack-scale architecture, which was developed jointly by AMD and Meta through the Open Compute Project.

Personal superintelligence for all

Meta has stated that it needs massive, scalable compute power that can handle the growing demands of the next generation of AI and enable ‘personal superintelligence for all’. This agreement is intended to futureproof what Meta refers to as, “Our leadership in AI,” as part of the Meta Training and Inference Accelerator (MTIA) silicon program.

Mark Zuckerberg, Founder and CEO of Meta said, “We’re excited to… deploy efficient inference compute and deliver personal superintelligence. This is an important step for Meta as we diversify our compute.”

“We are proud to expand our strategic partnership with Meta as they push the boundaries of AI at unprecedented scale,” said Dr. Lisa Su, chair and CEO at AMD. “This multi-year, multi-generation collaboration… aligns our roadmaps to deliver high-performance, energy-efficient infrastructure optimized for Meta’s workloads, accelerating one of the industry’s largest AI deployments.”