In a move that turns the AI arms race into a full-on power struggle, AMD and Anthropic today unveiled a sweeping infrastructure partnership that ties AMD's rack-scale Helios systems directly to Anthropic's Claude models, with a strategic equity kicker attached. Anthropic plans to light up as much as 2 gigawatts of AMD Instinct MI450-series GPUs, with the first gigawatt scheduled to come online in the first half of 2027, while AMD has pledged up to $5 billion in equity linked to deployment milestones. Expressing the buildout in gigawatts instead of individual chips drives home how the AI contest has become an infrastructure story and, by some estimates, could translate into tens of billions of dollars in server and services revenue.
Deal details and the engineering tie-up
Under the partnership, Anthropic will roll out AMD Helios rack-scale systems that bundle AMD Instinct MI455X GPUs (part of the MI450 family), EPYC "Venice" CPUs, Pensando networking and ROCm software as it scales up Claude. The companies also outlined a multi-year engineering collaboration. Anthropic's Claude models will help tune AMD's software stack and workloads, and AMD will, in turn, adopt Claude internally for engineering tasks.
"We are thrilled to deepen our partnership with Anthropic and deploy AMD Helios at gigawatt scale," AMD Chair and CEO Dr. Lisa Su said in the announcement. According to AMD, the first gigawatt deployment is slated to begin in the first half of 2027.
Price tag, milestones and market context
Early news coverage that sketched out the size and financing structure of the arrangement reported that the server side of the buildout could be worth "tens of billions" to AMD, with the $5 billion equity commitment structured around deployment milestones. Reuters noted that AMD executives have previously said 1 gigawatt of high-density AI compute can carry a "double-digit billions" price tag, which helps explain how a two-gigawatt commitment lines up with such large hardware values. The same reporting also highlighted discussions with other infrastructure providers and potential financing or backstop arrangements tied to Anthropic's leases.
Where Anthropic is already getting capacity
Anthropic has been scrambling to lock in compute capacity. In May, the company announced an agreement with SpaceX to use the full compute capacity of the Colossus 1 data center, a facility Anthropic described as providing more than 300 megawatts and more than 220,000 NVIDIA GPUs. The company has also signed other multi-gigawatt deals while raising customer rate limits. Those earlier agreements, Anthropic says, are part of a broader strategy to expand Claude's commercial reach and to diversify the locations where its workloads run. For more detail on the SpaceX arrangement, see Anthropic.
What it means for AMD and the AI supply chain
The Anthropic tie-up lands on the heels of AMD's recent run of large-scale AI deals. Last year, AMD announced a multiyear, multi-generation 6-gigawatt agreement with OpenAI that included milestone-linked warrants, signaling a strategy built around locking in long-term, high-volume customers. AMD's playbook of big, milestone-driven commitments plus engineering collaboration is aimed at turning Helios and Instinct into a mainstream alternative to the Nvidia-centric stacks that dominate frontier AI today.
What to watch next
Near term, the fine print will matter. Investors and rivals alike will be parsing the precise milestone language, watching the delivery cadence for Helios systems and tracking whether AMD takes on any financing or lease-back roles to help Anthropic secure data center capacity. Across the industry, cloud providers, hardware vendors and financiers will be watching to see whether this kind of circular deal structure, chips paired with equity or other financial support, becomes the default way to bankroll massive AI buildouts.
