AMD & Anthropic strike gigawatt-scale AI partnership

AMD and Anthropic have formed a strategic partnership to deploy up to 2 gigawatts of AMD Instinct MI450 Series GPUs. AMD also plans to invest up to USD $5 billion in Anthropic.

The deployment will use AMD Helios rack-scale systems, with the first gigawatt set to begin rolling out in the first half of 2027. Anthropic said the build-out extends its existing use of AMD Instinct MI355X GPUs and will add a larger footprint based on newer hardware.

The agreement links one of the biggest buyers of AI computing infrastructure with a chipmaker seeking a larger share of the training and inference market. It also includes a significant financial component through AMD's proposed equity investment, tying the companies together beyond a supply arrangement.

Anthropic plans to deploy AMD Helios systems using AMD Instinct MI455X GPUs, part of the MI450 Series. The systems will also include AMD EPYC "Venice" CPUs, Pensando networking technology, and ROCm software.

The companies are also beginning a multi-year engineering collaboration centred on Anthropic's Claude AI model. Under the arrangement, the teams will use Claude to optimise workloads for AMD Instinct GPUs and speed development of ROCm, AMD's software stack for AI and compute workloads.

AMD will also adopt Claude across its engineering and product development teams. That suggests the partnership is intended to shape both companies' internal development processes as well as the infrastructure used to train and run Anthropic's models.

Demand for compute infrastructure has become a central issue for AI model developers as they compete for access to chips, networking, and data centre capacity. For Anthropic, securing supply across a broader range of hardware vendors could reduce dependence on a narrower group of providers while expanding the resources available for Claude.

For AMD, the deal delivers a high-profile deployment for its Instinct line at a scale measured in gigawatts, a metric more often used for utility or hyperscale data centre demand than for individual AI projects. It also places AMD's hardware, networking, and software together in a large installation, rather than supplying only discrete components.

Dr Lisa Su, Chair and Chief Executive Officer of AMD, described the arrangement as a deeper tie between the two companies.

"We are thrilled to deepen our partnership with Anthropic and deploy AMD Helios at gigawatt scale," said Dr Lisa Su, Chair and Chief Executive Officer of AMD. "This collaboration brings together Anthropic's leadership in frontier AI with the full strength of AMD high-performance computing. Together, we will accelerate AI adoption at scale and establish Helios as a major platform for the next generation of AI infrastructure."

Anthropic framed the partnership around access to computing resources for Claude and the need to match workloads to different hardware.

"Access to compute is central to keeping Claude at the frontier and meeting demand from our customers," said Tom Brown, Co-Founder and Chief Compute Officer of Anthropic. "By partnering with AMD across the stack, we are securing the capacity we need and optimizing it for training and serving Claude. Running across a diversified range of hardware lets us map the right workloads to the right hardware."

Broader push

The arrangement comes as chipmakers and AI developers pursue closer ties that combine hardware supply, software integration, and capital. More than a standard customer-vendor relationship, the AMD-Anthropic deal spans infrastructure procurement, engineering work, and investment, reflecting how competition in AI increasingly depends on long-term access to large pools of compute.

AMD has been expanding its AI product line to challenge the dominance of rival suppliers in data centre accelerators. Winning larger deployments from model developers matters not only for sales volumes but also for proving that its chips and software can operate at the scale required for training and inference workloads.

Anthropic, meanwhile, has been building the infrastructure needed to support demand for Claude from developers and corporate users. Extending its use of AMD hardware indicates that alternative AI chip platforms are taking on a larger role in that effort.

The first phase of the rollout is set at one gigawatt, with the overall plan covering up to two gigawatts of MI450 Series GPU deployment.