The headline figure of 6 gigawatts represents a long-term planned capacity ceiling under a multi-generation collaboration rather than active, operating data center power or an immediate volume of installed chips. According to Meta, the multi-year deal is structured to align the two companies' product roadmaps across silicon, system engineering, and software enablement.
Meta plans to base its initial GPU deployments on Helios, an open rack-scale system architecture that the company co-developed with AMD and introduced at the previous year's Open Compute Project Global Summit. The companies expect shipments supporting these deployments to begin in the second half of 2026. AMD chair and CEO Dr. Lisa Su described the agreement as a multi-generation partnership encompassing Instinct GPUs, EPYC CPUs, and rack-scale AI systems tailored for Meta's workloads.
For AI practitioners and infrastructure planners, the arrangement highlights Meta's push to diversify its computing base rather than standardizing on a single architecture. Mark Zuckerberg, founder and CEO of Meta, stated that the partnership will deploy efficient inference compute to support personal superintelligence initiatives.
The AMD rollout forms part of the broader Meta Compute initiative, an infrastructure strategy combining external vendor silicon with Meta's proprietary Meta Training and Inference Accelerator (MTIA) program.