AMD announced a major infrastructure agreement with Anthropic on July 23, committing up to $5 billion in strategic equity investment alongside a massive hardware deployment spanning tens of billions of dollars.
Under the deal, Anthropic will deploy up to two gigawatts of AI compute capacity using AMD’s Instinct MI450-series accelerators, including MI455X GPUs, EPYC Venice processors, and Pensando networking hardware. The first gigawatt rollout begins in the first half of 2027.
How the Deal Works
AMD’s $5 billion equity investment is tied to Anthropic meeting specific deployment milestones. The companies have not disclosed what happens if those milestones are delayed, nor the valuation attached to the investment.
Anthropic will use AMD Helios systems, a rack-scale platform that combines accelerators, CPUs, networking, and ROCm software into a single integrated package. This builds on Anthropic’s existing use of AMD’s MI355X accelerators.
“We are thrilled to deepen our partnership with Anthropic and deploy AMD Helios at gigawatt scale,” said AMD CEO Lisa Su. “This collaboration brings together Anthropic’s leadership in frontier AI with the full strength of AMD high-performance computing.”
What Two Gigawatts Actually Means
Two gigawatts equals 2,000 megawatts of power capacity. AMD executives have stated that developing one gigawatt of AI computing infrastructure can cost tens of billions of dollars depending on equipment and facilities involved.
To put the scale in perspective, Anthropic previously secured more than 300 megawatts through SpaceX’s Colossus 1 facility in Memphis, which houses over 220,000 Nvidia GPUs. The AMD deployment would represent more than six times that power capacity, though the facilities use different hardware so direct computing performance comparisons are not straightforward.
Some AMD systems will live in Anthropic’s own data centers, while others will be hosted by cloud providers and specialist AI infrastructure companies.
AMD’s Growing Customer List
The Anthropic deal follows two similar arrangements AMD has made with other major AI companies:
- OpenAI (October 2025): Systems supporting up to six gigawatts, plus warrants allowing OpenAI to acquire roughly 10% of AMD.
- Meta (February 2026): Up to six gigawatts of GPU capacity with performance-based warrants.
- Anthropic (July 2026): Up to two gigawatts with direct equity investment.
The structures differ. AMD is making a direct cash investment in Anthropic, while the OpenAI and Meta deals give those companies rights to acquire AMD shares when specified conditions are met.
Anthropic’s Multi-Supplier Strategy
Anthropic continues to diversify its compute infrastructure across multiple chip vendors. The company’s current supplier network includes:
- Amazon: Up to five gigawatts of capacity, using more than one million Trainium2 processors with Trainium3 deployments planned for 2026.
- Google/Broadcom: Multiple gigawatts of next-generation TPU capacity starting in 2027.
- Nvidia: Existing deployments including Colossus 1.
- SpaceX: Additional capacity at Colossus facilities.
- AMD: Up to two gigawatts starting 2027.
“Running across a diversified range of hardware lets us map the right workloads to the right hardware,” said Anthropic co-founder Tom Brown. “Access to compute is central to keeping Claude at the frontier and meeting demand from our customers.”
Engineering Partnership
Beyond hardware supply, the deal includes a multi-year engineering program. Anthropic will use Claude to optimize workloads on AMD Instinct accelerators and help develop AMD’s ROCm software platform. AMD plans to deploy Claude across its own engineering and product-development teams.
AMD also reportedly discussed providing financial guarantees for Anthropic’s future data-center leases, though the companies have not confirmed these proposed terms.
Frequently Asked Questions
How much is AMD investing in Anthropic?
AMD committed up to $5 billion in strategic equity investment, tied to Anthropic meeting deployment milestones for the AMD Helios infrastructure.
When will Anthropic start using AMD chips?
The first gigawatt of AMD-based infrastructure deployment begins in the first half of 2027.
What chips will Anthropic use from AMD?
Anthropic will deploy AMD Helios systems featuring MI455X GPUs from the Instinct MI450 series, EPYC Venice processors, Pensando networking, and ROCm software.
How does this compare to Anthropic’s Nvidia usage?
Anthropic already has 300+ megawatts through SpaceX’s Colossus 1 facility using 220,000+ Nvidia GPUs. The AMD deployment would add six times that power capacity, though with different hardware architecture.
What does this mean for the AI chip market?
AMD is establishing itself as a credible alternative to Nvidia for large-scale AI deployments. With gigawatt-scale commitments from OpenAI, Meta, and now Anthropic, AMD has secured over 14 gigawatts of prospective AI compute demand.
