AMD and Meta have formalised a multi-year settlement to produce as much as six gigawatts of AMD Intuition GPUs for Meta’s AI infrastructure.
The deployment will start within the second half of 2026, with preliminary shipments supporting the primary gigawatt powered by customized AMD Intuition GPUs primarily based on the MI450 structure and sixth Era AMD EPYC CPUs, codenamed “Venice”.
Entry deeper trade intelligence
Expertise unmatched readability with a single platform that mixes distinctive information, AI, and human experience.
These techniques will run on ROCm software program and utilise the collectively developed AMD Helios rack-scale structure, which originated by collaboration inside the Open Compute Challenge.
The partnership expands on earlier collaborations between the 2 corporations and contains alignment of improvement roadmaps for silicon, techniques and software program.
Meta has already built-in hundreds of thousands of AMD EPYC CPUs and vital numbers of AMD Intuition MI300 and MI350 GPUs throughout its international infrastructure.
Beneath this new settlement, Meta will act as a lead buyer for each sixth Gen AMD EPYC “Venice” CPUs and “Verano”, an upcoming processor designed with optimisations for particular workloads.
As a part of the association, AMD has issued Meta a performance-based warrant permitting Meta to amass as much as 160 million shares of AMD widespread inventory.
The shares will vest as sure milestones associated to GPU shipments are met, beginning with the supply of the preliminary one gigawatt and scaling as much as six gigawatts, contingent on each cargo volumes and AMD attaining particular share value thresholds.
Each corporations stated that this association is meant to assist AI infrastructure at scale as demand for compute energy grows.
AMD chair and CEO Dr Lisa Su stated: “We’re proud to broaden our strategic partnership with Meta as they push the boundaries of AI at unprecedented scale.
“This multi-year, multi-generation collaboration throughout Intuition GPUs, EPYC CPUs and rack-scale AI techniques aligns our roadmaps to ship high-performance, energy-efficient infrastructure optimised for Meta’s workloads, accelerating one of many trade’s largest AI deployments and putting AMD on the centre of the worldwide AI buildout.”
The settlement additionally types a part of Meta’s broader technique to diversify its {hardware} partnerships as a way to construct a versatile know-how stack able to supporting a wide range of AI workloads.
Meta founder and CEO Mark Zuckerberg stated: “We’re excited to kind a long-term partnership with AMD to deploy environment friendly inference compute and ship private superintelligence.
“This is a crucial step for Meta as we diversify our compute. I anticipate AMD to be an essential associate for a few years to return.”
