AMD is putting up to $5 billion behind Anthropic in a new infrastructure agreement that ties the chipmaker more closely to one of the companies racing to build frontier AI systems.
The deal will see Anthropic deploy up to two gigawatts of capacity using AMD’s Instinct MI450-series accelerators. The first gigawatt is scheduled to begin deployment in the first half of 2027, with AMD’s investment linked to deployment milestones that the companies have not fully disclosed.
The companies described the arrangement as two connected pieces: a future equity investment from AMD and a large hardware deployment by Anthropic. They have not said that Anthropic must use AMD’s investment to pay for the systems, and they have not disclosed purchase obligations, cancellation rights or the valuation attached to the investment.
The planned systems will use AMD Helios infrastructure with MI455X GPUs, EPYC “Venice” processors, Pensando networking and ROCm software. AMD said the setup is a full rack-scale platform, not a stand-alone chip order.
“We are thrilled to deepen our partnership with Anthropic and deploy AMD Helios at gigawatt scale,” AMD chair and chief executive Lisa Su said.
For Anthropic, the agreement adds another supplier to a compute network that already includes Nvidia GPUs, Amazon Trainium processors and Google tensor processing units. Anthropic has said it uses more than one million Amazon Trainium2 processors, with Trainium3 deployments planned during 2026.
Anthropic co-founder and chief compute officer Tom Brown said using different hardware allows the company to match workloads to the systems best suited for them.
The companies also plan a multi-year engineering programme that will use Claude to tune workloads for AMD’s Instinct accelerators and support development of AMD’s ROCm software platform. AMD shares rose 2.4% after the agreement was announced.
