AMD Commits Up to $5 Billion to Anthropic in 2-Gigawatt Chip Deal

AMD will invest up to $5 billion in Anthropic under a partnership that places up to 2 gigawatts of the chipmaker's Instinct MI450 Series GPUs in the Claude developer's data centers, the companies announced Wednesday. Anthropic will run the processors inside AMD Helios rack-scale systems, with deployment of the first gigawatt starting in the first half of 2027. The Wall Street Journal first reported the deal.

AMD's announcement attaches no date to the money, stating only that the company "has committed to make a strategic equity investment of up to $5 billion in Anthropic in the future." The commitment depends on certain milestones being met, according to Mobile World Live.

Inside the Helios deployment

Anthropic's racks pair Instinct MI455X GPUs with AMD's EPYC "Venice" CPUs, Pensando networking and ROCm software. Each Helios rack carries 72 of those GPUs, 31 TB of HBM4 memory and roughly 2.9 exaflops of FP4 performance, according to Techzine. Bloomberg put a single gigawatt at enough to power up to 750,000 US homes at any given time.

The agreement reaches past hardware into both companies' engineering. Claude will be used to optimize workloads for Instinct GPUs and to accelerate ROCm development, and AMD will adopt the model across its engineering and product development teams. Where the racks will physically land was not disclosed.

AMD's OpenAI warrant ran the other direction

The structure inverts AMD's arrangement with OpenAI. CNBC reported that AMD issued OpenAI a warrant last October for up to 160 million shares of AMD common stock, close to 10% of the company, with vesting tied to deployment volume and AMD's share price. That arrangement paid the customer in the supplier's own equity, while this one sends cash toward the buyer and leaves AMD "both supplier and shareholder," as Tom's Hardware put it.

Bloomberg noted that arrangements of this kind between chipmakers and AI developers have drawn criticism as circular, with the industry arguing they are necessary to fund the buildout.

A GitHub config file named Anthropic before AMD did

SemiAnalysis reported Sunday that a YAML configuration file in an AMD senior director's public GitHub repository listed Anthropic as a "customer" carrying the maximum 30 priority boost points, the same tier as hyperscale buyers such as Meta. Anthropic had not been publicly confirmed as an AMD customer at that point, and its disclosed suppliers were Nvidia, Amazon and Google. Wednesday's release states that the Helios deployment "builds on Anthropic's existing use of AMD Instinct MI355X GPUs," which Tom's Hardware called the first official confirmation that Claude workloads already run on AMD silicon.

Where AMD sits in Anthropic's supplier stack

Recent data center deals with SpaceX and TeraWulf preceded it, alongside infrastructure arrangements with Google, Broadcom and Amazon, per The Verge. Anthropic said in April that demand for its Claude models and products had caused "inevitable strain" on its infrastructure, which hurt "reliability and performance" for users, particularly at peak hours. In May the company agreed to take all the compute at SpaceX's Colossus 1 data center in Memphis, paying $1.25 billion a month through May 2029 under terms disclosed in SpaceX's prospectus, CNBC reported. It has also held preliminary talks about leasing capacity from Meta.

Counting Nvidia GPUs, Amazon's Trainium chips under Project Rainier, 3.5 gigawatts of Google TPU capacity coming from Broadcom in 2027 and 300 megawatts rented from Colossus 1, AMD becomes roughly the sixth distinct silicon source feeding Claude, by Tom's Hardware's tally.

The Helios queue

Anthropic's first AMD gigawatt sits behind other buyers. OpenAI's first gigawatt of AMD capacity is due in the second half of this year, and Oracle is scheduled to begin deploying 50,000 MI450 GPUs this quarter. Microsoft committed on Monday to running Helios racks at scale on Azure, which makes Anthropic AMD's second major Helios customer announcement inside a week.

Anthropic's run-rate revenue passed $47 billion in May, CNBC reported, up from about $10 billion for all of last year, and the company closed a round that month valuing it at $965 billion. It filed a confidential prospectus with the SEC in June and has yet to disclose further plans.

Anthropic Weighs Microsoft Maia Chip Deal as Claude Demand Grows

Anthropic is in early talks to rent servers powered by Microsoft’s Maia AI chips, The Information reported Thursday, citing two people familiar with the discussions. The talks would give the Claude ma

The Implicator

Meta Turns to Amazon Graviton Chips as AI Agents Shift Compute Math

Meta has signed a multiyear deal to run AI workloads on Amazon's Graviton processors, giving AWS one of its largest outside validations for homegrown server chips. The deployment begins with tens of m

The Implicator

Google Cloud Locks In Thinking Machines Lab With Multi-Billion GB300 Deal

Mira Murati's Thinking Machines Lab has signed a new multi-billion-dollar agreement with Google Cloud for AI infrastructure built on Nvidia's latest GB300 chips, Google announced Wednesday at its Clou

The Implicator

Marcus Schuler

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai