AMD to Invest $1 Billion in Meta Equity as Helios Systems Run at Meta

Advanced Micro Devices (AMD ) said on October 27, 2026, that its first Helios rack-scale AI systems are now running at Meta, that AMD EPYC processors power Meta’s Muse personal AI agent, and that AMD technologies provide Day-0 support for Meta’s Muse Spark Open and Muse Glimmer models. In the announcement, datelined Santa Clara, California, AMD also said it will invest $1 billion in Meta’s equity offering.

The release describes the work as building on years of co-design between the two companies across AMD EPYC processors, AMD Instinct GPUs, AMD Pensando networking and AMD ROCm software supporting Meta’s services and AI infrastructure.

Lisa Su, AMD’s chair and chief executive officer, described Meta as a strategic partner developed through years of deep technical collaboration. “With Helios now running at Meta, we are taking that partnership to a new scale in the superintelligence era,” she said. Mark Zuckerberg, Meta’s founder and chief executive officer, said in the same release: “AMD has been a key partner in building out the compute that’s running our frontier models.”

Hardware Deployed at Meta

The initial AMD Helios rack-scale systems are installed and being tested by Meta, with a broader deployment expected in the near future, according to the company. AMD said Helios rackscale solutions will power Meta’s leading frontier AI models and services.

AMD EPYC processors support the dedicated computing environments behind Muse. According to the release, Meta has deployed millions of EPYC processors across multiple generations and is a lead customer for AMD’s 6th Gen EPYC CPUs, and the companies are collaborating on future generations, including a next-generation EPYC processor codenamed “Verano.”

AMD Instinct GPUs are currently 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 will deploy a custom AMD Instinct GPU optimized for its workloads, AMD said. The collaboration is also informing future AMD platforms, including AMD Instinct MI500 series GPUs and next-generation rack-scale systems, according to the company, which said the partners have aligned product roadmaps and engineering decisions around current and future Meta workloads at gigawatt scale.

On software, AMD and Meta have worked together to advance ROCm through PyTorch so developers can use AMD GPUs out of the box, the release said. 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 companies have also expanded their software collaboration to large-scale AI training, optimizing recommendation models on AMD Instinct MI350X GPUs and contributing improvements to PyTorch, Triton and ROCm for other developers to use. A Meta-authored research paper submitted on October 15, 2026, presents that effort as a case study of recommendation-model training on the MI350X compared with NVIDIA’s B200. Meta operates a heterogeneous fleet of NVIDIA GPUs, AMD GPUs and proprietary MTIA accelerators to reduce costs and mitigate vendor lock-in, the authors wrote. The paper cites the MI350X’s 1.6x HBM capacity relative to the B200 and reports up to a 1.5x improvement in end-to-end training throughput over a directly ported baseline, results the authors attribute to revising model structures and software recipes rather than porting models directly from NVIDIA GPUs. The authors describe the work as the first hyperscaler report on productionizing AMD GPUs for large-scale training, and the paper is to appear at the 59th IEEE/ACM International Symposium on Microarchitecture (MICRO 2026).

The February 6-Gigawatt Agreement

The October 27 announcement follows the definitive multi-year, multi-generation agreement the companies announced on February 24, 2026, to deploy up to 6 gigawatts of AMD Instinct GPUs. Shipments supporting the first gigawatt deployment are scheduled to begin in the second half of 2026, powered by a custom AMD Instinct MI450-based GPU and 6th Gen AMD EPYC CPUs codenamed “Venice,” running ROCm software and built on the AMD Helios rack-scale architecture. AMD Helios was developed jointly by AMD and Meta through the Open Compute Project and was announced at the 2025 OCP Global Summit.

As part of the February agreement, AMD issued Meta a performance-based warrant for up to 160 million shares of AMD common stock, structured to vest as specific milestones associated with Instinct GPU shipments are achieved. The first tranche vests with the initial 1-gigawatt of shipments, with additional tranches vesting as Meta’s purchases scale to 6 gigawatts. Vesting is further tied to AMD achieving certain stock-price thresholds, and exercise of the warrant is tied to Meta achieving key technical and commercial milestones. Jean Hu, AMD’s executive vice president, chief financial officer and treasurer, said at the time that the company expected the partnership to drive substantial multi-year revenue growth and to be accretive to its non-GAAP earnings per share.

Meta said in its own announcement that the agreement forms part of its Meta Compute initiative and a portfolio-based approach to infrastructure that combines hardware sourced from a range of partners with its in-house Meta Training and Inference Accelerator (MTIA) silicon program.