AMD Helios Racks Now Running at Meta as Partnership Reaches Gigawatt Scale
AMD says its Helios rackscale solutions are now running at Meta, with EPYC processors powering Meta's Muse AI agent and AMD technology providing Day-0 support for Meta models. The chipmaker also said it will invest $1 billion in Meta's equity offering.
- AMD says initial Helios rackscale solutions are installed and being tested at Meta, with broader deployment expected in the near future.
- AMD EPYC processors power the environments behind Meta's Muse AI agent, and Meta has deployed millions of EPYC chips.
- Meta will deploy a custom AMD Instinct GPU optimized for its workloads.
- AMD provides Day-0 support for Muse Spark Open and Muse Glimmer.
- AMD will invest $1 billion in Meta's equity offering.
Helios installed, broader deployment expected
According to AMD, initial AMD Helios rackscale solutions are installed and being tested by Meta, on the path to a broader deployment in the near future. AMD describes the partnership as spanning CPUs, GPUs, networking, systems and software, and says it builds on years of co-design between the two companies across AMD EPYC processors, AMD Instinct GPUs, AMD Pensando networking and the AMD ROCm software stack.
Lisa Su, Chair and CEO of AMD, said that with Helios now running at Meta, the companies are taking the partnership "to a new scale in the superintelligence era." She added that the two are building and deploying infrastructure at gigawatt scale. Mark Zuckerberg, founder and CEO of Meta, said AMD "has been a key partner in building out the compute that's running our frontier models."
CPUs and GPUs behind Muse and Meta's workloads
AMD EPYC processors support the dedicated computing environments behind Meta's Muse personal AI agent. AMD says Meta has deployed millions of EPYC processors across multiple generations and is a lead customer for 6th Gen EPYC CPUs. The companies are also working together on future generations, including "Verano."
On the GPU side, AMD Instinct GPUs are used in production in Meta's infrastructure for training, ranking and recommendation engines. Meta was an early definition customer for the AMD Instinct MI450 architecture, and AMD says it will deploy a custom AMD Instinct GPU optimized for its workloads. The collaboration is also informing future AMD platforms, including the AMD Instinct MI500 series GPUs and next-generation rack-scale systems.
- EPYC: powers the environments behind the Muse agent
- Instinct: used for training, ranking and recommendation engines
- Helios: rackscale systems installed and in testing
Software support and the $1 billion investment
AMD and Meta have worked together to advance ROCm through PyTorch, so that developers can use AMD GPUs right out of the box, according to AMD. AMD delivered Day-0 support for Meta's Muse Spark Open models on AMD Instinct GPUs, and for Muse Glimmer on AMD Ryzen AI PCs and AMD Radeon GPUs.
The software work has also extended to large-scale AI training. The companies optimized recommendation models on AMD Instinct MI350X GPUs and contributed improvements to PyTorch, Triton and ROCm for other developers to use.
AMD says the companies have aligned product roadmaps and engineering decisions around current and future Meta workloads. In addition, AMD will invest $1 billion in Meta's equity offering. The release gives no further detail on the terms of that investment or on the timing or size of the wider Helios rollout beyond "the near future."
What it means
For AMD, the announcement positions its full stack, from CPUs and GPUs to networking and open software, as a supplier of Meta's AI infrastructure, which AMD says must serve billions of people. For developers, the stated work on PyTorch, Triton and ROCm is meant to make AMD hardware usable with common tools. Much of the forward-looking content, including broader Helios deployment and future products such as the MI500 series, is described by AMD as expectation rather than completed fact.
