AMD Partners with Meta to Invest $1 Billion

AMD announced on October 27, 2026 that it will deepen its strategic partnership with Meta. The AMD Helios rack level solution is currently running at Meta, with AMD EPYC processors providing support for Meta’s Muse AI agent, and AMD technology also providing Day-0 support for Meta’s Muse Spark Open and Muse Glimmer models. This collaboration is built upon years of joint design and collaboration between the two companies, covering AMD EPYC processors, AMD Instinct GPUs, AMD Pensando networks, and AMD ROCm software to support Meta’s services and its growing AI infrastructure.

Helios is already running on Meta

AMD Chairman and CEO Su Zifeng stated that Meta is a strategic partner of AMD, and the two parties have established this partnership through years of deep technical cooperation, from EPYC processors supporting Meta infrastructure to Instinct GPUs and open software supporting its AI workloads. She said that with Helios now running on Meta, the two sides are taking their collaboration to a new level in the era of super intelligence, jointly building and deploying infrastructure at gigawatt scale, and shaping future generations of silicon wafers, software, and systems to support Meta’s most ambitious super intelligence models and services.

Meta founder and CEO Mark Zuckerberg stated that AMD has always been a key partner in building and running its cutting-edge model computing power. He said that Meta is investing heavily in computing power to bring personal superintelligence to everyone, with AMD playing an important role in it.

The announcement states that AMD technology supports Meta infrastructure in terms of computing, networking, systems, and software. The first batch of AMD Helios rack level systems have been installed and are currently being tested by Meta. They will be deployed more widely in the future. AMD EPYC processors support the dedicated computing environment behind Meta Muse personal AI agents. Meta has deployed millions of EPYC processors in multiple generations of products and is a major customer for the sixth generation EPYC CPUs. Both parties are still collaborating on future generations of products, including “Verano”.

GPU and software collaboration expansion

AMD Instinct GPUs are currently in production in Meta infrastructure for training, sorting, and recommendation engines. Meta is still an early defining customer of the AMD Instinct MI450 architecture and will deploy a customized AMD Instinct GPU optimized for its workload. This collaboration is also impacting AMD’s future platforms, including AMD Instinct MI500 series GPUs and next-generation rack level systems.

AMD and Meta are jointly promoting ROCm through PyTorch to ensure that developers can directly use AMD GPUs. AMD also provides Day-0 support for Meta’s Muse Spark Open model on AMD Instinct GPU, and Day-0 support for Muse Glimmer on AMD Ryzen AI PC and AMD Radeon GPU. Both parties will also expand their software collaboration to large-scale AI training, optimize recommendation models on AMD Instinct MI350X GPU, and contribute improvements to PyTorch, Triton, and ROCm for use by other developers.

As Meta expands its AI capabilities, both parties are collaborating on silicon wafers, software, and systems to meet the demand for running AI infrastructure for billions of people. The two companies have aligned their product roadmap and engineering decisions around Meta’s current and future workloads, and have developed a multi-year collaboration to build and deploy integrated AI systems at a gigawatt scale. In addition, AMD will invest $1 billion to participate in Meta’s equity offering.