A significant multi-year agreement between AMD and Meta aims to enhance Meta's AI infrastructure, featuring the deployment of 6 gigawatts of custom high-performance GPUs. Shipments of these GPUs, based on the MI450 architecture, are scheduled to commence in the latter half of 2026.
Dr. Lisa Su, the CEO of AMD, emphasized the strategic nature of this partnership, which aligns the roadmaps of both companies across future generations of AMD's products, including EPYC CPUs and rack-scale AI systems. The collaboration aims to improve the performance and energy efficiency of Meta's extensive AI workloads, ultimately supporting one of the industry's largest AI deployments.
Additionally, the agreement includes Meta's integration of AMD's 6th Gen EPYC CPUs, codenamed “Venice,” which is anticipated to enhance Meta's AI capabilities through improved workload density. AMD has also provided a performance warrant for up to 160 million shares based on GPU shipment targets and stock price performance, underscoring the financial implications of this collaboration.
The partnership will also see the development of the AMD Helios rack-scale architecture, as part of the Open Compute Project, aimed at creating scalable, energy-efficient AI infrastructures. This collaboration is expected to contribute to significant revenue growth for AMD while equipping Meta with essential tools for its ongoing AI innovation efforts.