Anthropic Reportedly Testing AMD GPUs as AI Giants Systematically Reduce Reliance on Single Compute Source

Anthropic is incorporating AMD into its compute supplier system. According to an interpretation by chip industry analysis firm SemiAnalysis, information disclosed in public GitHub code by AMD’s Senior Director of AI Business indicates that Anthropic will become an AMD customer. This move suggests that Anthropic is actively broadening its chip sources to reduce reliance on a single supplier.

On Monday, according to an analysis post published by SemiAnalysis on the social media platform X, the firm concluded that Anthropic would purchase AMD chips by interpreting GitHub code publicly released by AMD’s Senior Director of AI Business. SemiAnalysis stated in the post that it would further explain the code details and related background.

If the above information is accurate, AMD will gain an important new customer in the high-end AI training market, previously dominated by NVIDIA and Google. It also means that Anthropic’s compute infrastructure layout is accelerating towards diversification.

GitHub Code Accidentally Becomes a "Leak" Channel

The disclosure path for this information was quite unusual. It did not come from an official announcement or media report, but rather from code content published by an AMD internal executive on the public code hosting platform GitHub. SemiAnalysis professionally interpreted this batch of code and publicly released its analysis conclusions on the X platform, pointing out that it clearly involved information regarding Anthropic as an AMD customer.

Such accidental leaks of sensitive commercial information through public code repositories are not rare in the tech industry. Given that GitHub code is publicly searchable and the publisher was AMD’s Senior Director of AI Business, the credibility of the information is relatively high. In the post, SemiAnalysis promised to further dismantle the code logic and details to provide a more complete interpretation for the market.

Anthropic Had Previously Established Multiple Chip Sources

Prior to the news about AMD, Anthropic had already established cooperative relationships with multiple chip and compute suppliers. Google’s TPU (Tensor Processing Unit) is one of Anthropic’s important compute sources, which is closely related to Google’s strategic investment in Anthropic. Additionally, Anthropic had previously included Samsung in its supplier system.

This existing layout indicates that Anthropic’s compute procurement strategy does not rely on a single source; instead, it has long been working to build a diversified supplier matrix. The addition of AMD will further enrich this matrix and provide Anthropic with more optional high-performance computing resources beyond NVIDIA GPUs.

Compute Diversification Is a Systematic Choice for AI Companies

For Anthropic, introducing AMD holds multiple strategic values. As the scale of model training and inference continues to expand, issues such as supply chain risks, limited bargaining power, and technology path lock-in caused by relying on a single chip supplier are becoming increasingly prominent. By simultaneously adopting products from multiple suppliers such as Google TPUs, Samsung, and AMD, Anthropic can seek a better balance between cost control, supply stability, and technical flexibility.

For AMD, securing Anthropic as a customer has significant benchmarking importance. AMD has continuously increased its investment in the data center AI chip sector in recent years, and its MI series GPUs are regarded as one of the main competitive options to NVIDIA’s H series products. If Anthropic officially becomes an AMD customer, it will help AMD establish a stronger market presence among top-tier AI laboratory customers and bring substantial incremental revenue to its data center AI chip business.

Anthropic’s moves reflect a broader industry trend. As leading institutions such as OpenAI, Google DeepMind, and Meta AI continue to see rising demand for compute power, AI companies are systematically evaluating and expanding their chip supply sources to avoid the risk of excessive concentration on a single supplier. This trend objectively creates more opportunities for NVIDIA’s competitors, such as AMD and Intel, to enter the supply chains of core AI customers.