AMD, Anthropic Strike AI Chips Deal with up to 2 GW of MI450 GPU Deployments
Anthropic plans to deploy as much as 2 gigawatts of AMD’s upcoming Instinct MI450-series GPUs, giving the chipmaker one of its biggest public customer wins yet in the race to challenge Nvidia in AI infrastructure.
AMD said in a press release on Wednesday that the first 1 gigawatt of deployments is expected to begin in the first half of 2027 through its Helios rack-scale systems. The package will use MI455X GPUs, AMD’s “Venice” EPYC processors, Pensando networking gear and the ROCm software stack. AMD also said it has committed to make a strategic equity investment of as much as $5 billion in Anthropic.
The agreement ties a large hardware commitment to a broader engineering relationship. The companies said they will work together over multiple years to use Anthropic’s Claude models to optimize workloads for AMD’s Instinct chips and speed development of ROCm, while AMD plans to expand internal use of Claude across engineering and product teams.
The announcement adds to a run of AI infrastructure deals AMD has used to build out its position with hyperscalers and model developers. On July 20, 2026, Microsoft said it would deploy AMD’s Helios platform at scale on Azure for frontier-model inference and other AI services.
AMD said in its first-quarter earnings release that revenue rose to $10.3 billion and that its data-center business had become the main driver of growth, with customer forecasts for MI450 systems running ahead of its initial expectations.
For Anthropic, the deal marks another step in securing computing capacity as demand for Claude grows. The company said on May 28 that it raised $65 billion in Series H funding at a $965 billion post-money valuation. The Wall Street Journal and the Financial Times separately reported that the chip purchase could amount to tens of billions of dollars over time.
