Anthropic explores custom AI chip development with Samsung partnership

Anthropic is in preliminary talks with Samsung about co-developing a custom AI processor, marking a shift from the company's earlier interest in the idea toward concrete action, according to reporting from The Information.

The move reflects growing pressure among AI developers to secure reliable semiconductor supplies and reduce reliance on Nvidia's dominance in the market. When asked about the partnership discussions, Anthropic declined to offer specifics, reaffirming its commitment to maintaining a mixed approach to compute infrastructure across chips from major providers including Google, Amazon, and Nvidia.

Why custom chips matter#

The company faces an immediate challenge: critical details about the prospective Samsung chip remain unresolved. Anthropic has not yet determined its intended purpose, how it will integrate into existing hardware infrastructure, or what performance targets it should meet.

This uncertainty reflects a broader industry trend. Other AI leaders have moved aggressively to reduce Nvidia dependency. OpenAI recently unveiled its own inference processor, called "Jalapeño," developed in collaboration with Broadcom—a chip designed to deliver superior energy efficiency compared to competing options. Both Amazon and Google have similarly invested in proprietary tensor processing units as core components of their cloud offerings.

Samsung's expanding AI role#

Samsung is already deeply embedded in AI chip manufacturing as a critical partner to Nvidia, producing components necessary for training and running large language models. The South Korean company leverages Nvidia's software capabilities to guide its production processes, and the two organizations are jointly developing an advanced manufacturing facility domestically. Samsung has also been in discussions with Google regarding contributions to its next-generation AI chip lineup as manufacturing capacity tightens across the industry.