The AI lab is exploring a chip collaboration with Samsung but has not yet determined the chip's purpose, server role, or performance targets.

Anthropic is in early discussions with Samsung to develop a custom AI chip, according to a report by The Information cited by TechCrunch, though key design decisions remain unresolved ^[1].

For developers and businesses that rely on Anthropic’s models, the move signals the company is working to reduce its dependence on third-party silicon — a constraint that has affected capacity and pricing across the AI industry ^[1].

What’s Known — and What Isn’t

Anthropic has not yet decided what the chip will be used for, how it will fit into a server, or how powerful it will be, according to The Information’s reporting ^[1]. The company told TechCrunch that a “diversified hardware stack that includes chips from Google, Amazon, and Nvidia will continue to be pivotal to its compute strategy,” and declined to comment further on a potential Samsung partnership ^[1].

The talks build on an April Reuters report that Anthropic was considering producing its own AI chips as a way to address chip shortages ^[1].

Industry Context

Anthropic is not alone in pursuing custom silicon. Several AI companies have moved in this direction both to tailor hardware for specific compute tasks and to reduce reliance on Nvidia, which remains the dominant force in AI chips ^[1].

The most direct competitive parallel is OpenAI, which last week announced a custom-built inference processor called “Jalapeño,” developed in partnership with Broadcom ^[1]. OpenAI claims the chip delivers better performance-per-watt than competing chips ^[1]. Amazon and Google each offer their own custom tensor processing units (TPUs) through their cloud platforms ^[1].

Samsung brings established AI-industry credentials to any potential deal. The company is already a major Nvidia partner, producing chips used to train and run AI models, and the two are jointly working on an AI chip factory in South Korea ^[1]. Samsung has also held separate discussions with Google about chip-making collaboration ^[1].