Ai Business

AMD to Invest Up to $5 Billion in Anthropic in The Next Circular Deal

co-written by eustella AI22. July 2026, 16:44

US chipmaker AMD and AI company Anthropic announced a strategic partnership on July 22, 2026. Anthropic will deploy up to two gigawatts of AMD Instinct GPUs from the MI450 series in AMD’s “Helios” rackscale systems, with the first gigawatt slated to come online in the first half of 2027. In parallel, AMD intends to invest up to $5 billion in Anthropic — contingent on meeting certain milestones.

For AMD, the agreement is a significant step in its competition with market leader Nvidia: in return, Anthropic has committed to purchasing AI server chips worth tens of billions of dollars from AMD over the term of the deal.

What Exactly Was Agreed

Anthropic will run the Helios rackscale systems with AMD Instinct MI455X GPUs — part of the MI450 series — combined with AMD EPYC “Venice” CPUs, AMD Pensando networking and the ROCm software stack. According to AMD, the systems are purpose-built for next-generation AI training and inference. The new agreement builds on Anthropic’s existing use of AMD Instinct MI355X GPUs.

One gigawatt of compute capacity roughly corresponds to the power output of a single nuclear reactor. The first gigawatt is scheduled to be deployed in the first half of 2027.

Engineering Collaboration: Claude to Accelerate AMD’s Software

Alongside the hardware deal, the two companies are launching a multiyear technical collaboration. Anthropic’s Claude AI model is meant to help optimize workloads for AMD Instinct GPUs and accelerate development of the ROCm software. Conversely, AMD plans to adopt Claude broadly across its engineering and product-development teams.

Voices on the Partnership

AMD chair and CEO Lisa Su described the collaboration as bringing together Anthropic’s AI expertise and AMD’s high-performance computing, and spoke of establishing Helios as a central platform for the next generation of AI infrastructure.

Tom Brown, co-founder and chief compute officer at Anthropic, stressed the importance of access to compute: it is central to keeping Claude at the frontier and meeting customer demand, and partnering across the stack secures the necessary capacity. According to the company, Anthropic deliberately pursues a diversification strategy to match workloads with the best-suited hardware.

Context: Circular Deals and the Chip Shortage

The AMD deal follows a pattern that has become common around AI companies: a chipmaker invests in an AI company — which in turn commits to buying large volumes of chips from that same maker. The capital effectively flows in a loop, with the supplier helping to finance part of the demand for its own products.

There are numerous examples of such “circular” arrangements. In the current deal, AMD invests up to $5 billion in Anthropic, which in turn takes delivery of AMD chips worth tens of billions of dollars. Similarly structured is the partnership announced in November 2025, in which Nvidia (up to $10 billion) and Microsoft (up to $5 billion) invested in Anthropic, while Anthropic committed to $30 billion in Azure spending and to capacity on Nvidia systems.

The pattern is most pronounced at OpenAI: Nvidia announced in the fall of 2025 that it would invest up to $100 billion in OpenAI, which in return plans to fill its data centers with millions of Nvidia chips — a figure that was reduced to roughly $30 billion in early 2026. Weeks later, OpenAI struck a deal for six gigawatts of AMD Instinct GPUs; as part of it, AMD granted OpenAI warrants for up to 160 million of its own shares — around ten percent of the company.

Critics compare such constructs to the “vendor financing” practices of the dot-com era, warning that the AI boom may be partly propped up by these reciprocal deals rather than by genuine end demand.

The backdrop is an industry-wide semiconductor shortage that has driven chip stocks sharply higher in recent months. AMD’s shares are up more than 140 percent so far this year. AI companies are currently racing to lock in compute and power capacity early.

Notable is the shift in stance by Anthropic CEO Dario Amodei, who as recently as late 2025 had criticized competitors for aggressively securing compute ahead of demand. Since then, Anthropic too has struck a series of multibillion-dollar compute deals.

Anthropic’s Broader Compute Portfolio

The AMD deal is one in a whole series of infrastructure agreements. Anthropic trains and runs Claude on a deliberately broad hardware base — alongside AMD, it uses chips from Nvidia, Amazon and Google. According to reporting by the Financial Times, the company has secured access to more than ten gigawatts of new capacity in this year alone. The key agreements at a glance:

Amazon / AWS (Project Rainier, Trainium): AWS is Anthropic’s primary cloud and training partner. Amazon has invested roughly $8 billion in Anthropic since 2024. Together, the two operate “Project Rainier,” one of the world’s largest compute clusters, with more than one million Trainium2 chips. In April 2026 they expanded the partnership: Anthropic secures up to five gigawatts of Trainium capacity and commits to spending more than $100 billion on AWS technology over ten years. Around one gigawatt of Trainium2 and Trainium3 capacity is expected to be online by the end of 2026.

Google and Broadcom (TPUs): In October 2025, Anthropic announced access to up to one million Google TPUs — a deal worth tens of billions of dollars, expected to bring more than one gigawatt of capacity online in 2026. In April 2026 an expansion with Google and Broadcom followed, covering several additional gigawatts of TPU capacity; according to a Broadcom filing with the securities regulator, it amounts to roughly 3.5 gigawatts, due to come online starting in 2027.

Microsoft and Nvidia (Azure, Grace Blackwell / Vera Rubin): In November 2025, Microsoft, Nvidia and Anthropic formed a joint partnership. Anthropic committed to purchasing $30 billion of Azure compute and to contracting up to one gigawatt of additional capacity. In return, Nvidia plans to invest up to $10 billion and Microsoft up to $5 billion in Anthropic. Claude is also to be optimized for Nvidia’s upcoming Grace Blackwell and Vera Rubin systems. With the move, Claude is available across all three major cloud platforms — Amazon, Google and Microsoft.

SpaceX (data-center capacity): In May 2026, according to the Financial Times, Anthropic agreed to pay Elon Musk’s space company $1.25 billion per month to lease data-center space.

The deals reflect sharply rising demand for Anthropic’s products: according to the company, its annualized revenue run rate stood at more than $30 billion in 2026, up from around $9 billion at the end of 2025.