AMD will sell tens of billions worth of Anthropic AI servers and invest up to $5 billion in the company
AMD Makes Its First Investment in Anthropic / Photo: AMD
Chipmaker Advanced Micro Devices (AMD) and Anthropic have struck a deal worth tens of billions of dollars to supply servers for artificial intelligence, according to The Wall Street Journal (WSJ). The parties have not disclosed the exact value of the equipment. The agreement strengthens AMD’s competitive position against industry leader Nvidia and provides Anthropic with the necessary computing power, the newspaper explains.
Under the terms of the agreement with AMD, developer Claude will deploy Helios servers with a capacity of up to 2 gigawatts, powered by AMD Instinct MI450 Series AI chips (GPUs). Deliveries of the first gigawatt will begin in the first half of 2027.
Analysts at Futurum Group estimate the cost of a single Helios rack with 72 GPUs at $5–5.5 million, according to CNBC. AMD does not disclose the power consumption of a single rack, but TechTimes estimates that Helios’ peak power consumption is about 140 kilowatts. Based on these assumptions, a 2-gigawatt infrastructure would consist of approximately 14,300 racks, and its cost could be around $70–80 billion.
AMD is also investing up to $5 billion in Anthropic as it reaches certain milestones in its server deployment. This is the company's first investment in an AI developer, the WSJ notes.
The AI company runs its computations on various types of chips, including Google’s tensor processors, Amazon’s Trainium chips, and Nvidia’s graphics processing units. As part of the deal, Anthropic will purchase some AMD chips for its own data centers, while leasing some computing capacity through other major cloud providers. According to AMD CEO Lisa Su, Anthropic and AMD are jointly selecting data centers for these chips. Additionally, Su said that AMD has signed an engineering collaboration agreement with Anthropic under which it will use Claude models to improve the performance of its technologies, the WSJ reports.
According to a source at the newspaper who is directly familiar with the negotiations, AMD is also discussing the possibility of providing financial support for future leases of Anthropic’s data centers.
The deal with AMD will help Anthropic meet the significant growth in demand for AI tools, which has made the company one of the leaders in the AI race, the WSJ notes. The AI developer is forced to rapidly scale up its computing power. In April, the company stated that demand for its Claude models and products had created an “unavoidable strain” on its infrastructure, which affected “reliability and performance” for users, especially during peak hours, according to CNBC.
In May, Anthropic entered into an agreement with Elon Musk’s SpaceX to use computing capacity at the Colossus 1 data center in Memphis. According to SpaceX’s prospectus, under this deal, Anthropic will pay SpaceX $1.25 billion per month through May 2029. In April, the AI company signed a multibillion-dollar agreement with Amazon, as well as an agreement with Google and Broadcom for several gigawatts of computing power. In addition, Anthropic is in preliminary talks to lease computing capacity from Meta, CNBC reports.
AMD’s stock rose 2% during trading on July 22. Since the beginning of the year, its value has more than doubled. Reuters notes that in October 2025, AMD also entered into a multi-year agreement with OpenAI, which is expected to generate tens of billions of dollars in annual revenue for the chipmaker and give the creator of ChatGPT the option to purchase up to 10% of the chipmaker’s shares.
This article was AI-translated and verified by a human editor
