AMD announces partnership with Meta to provide 6 Gigawatts of GPUs

A few months ago, AMD signed an agreement with OpenAI to supply GPUs for upcoming datacentre projects. Now, the company has also signed an agreement with Meta to deploy up to 6 Gigawatts of AMD Instinct graphics cards to power AI datacentres.

The initial deployment will use a custom Instinct GPU based on AMD’s MI450 architecture, paired with 6th Gen EPYC “Venice” CPUs and running ROCm software. The systems will be built on AMD’s Helios rack‑scale architecture, which AMD and Meta co‑developed through the Open Compute Project. Meta plans to use the platform to accelerate training and inference workloads across its global infrastructure.

In return for this, AMD will issue Meta with up to 160 million shares, depending on performance targets being met. If 6 Gigawatts worth of GPUs are used up, then it appears Meta will get a roughly 10% stake in AMD in return.

Beyond GPUs, the companies are also expanding their EPYC CPU collaboration. Meta has already deployed millions of EPYC processors and significant volumes of Instinct MI300 and MI350 GPUs, and will now become a lead customer for both the 6th Gen EPYC “Venice” and the next‑generation “Verano” CPUs.

AMD and Meta say the partnership will support AI infrastructure at global scale, enabling faster development and deployment of AI models and services used by billions of people.

KitGuru Says: AMD now has two very important AI-related customers on its roster in OpenAI and Meta.

Become a Patron!