Anthropic has entered early-stage discussions with Samsung Electronics to manufacture a custom AI chip, according to a report published Thursday by The Information, joining a growing list of frontier AI labs seeking to reduce their reliance on Nvidia hardware.
The talks are exploratory, and Anthropic has not yet decided what the chip will be optimized for, how powerful it will be, or how it will fit into a server rack. According to three sources familiar with the matter cited in the report. The discussions reportedly center on Samsung’s 2-nanometer manufacturing process. The company’s most advanced node, along with its advanced chip-packaging capabilities.
Anthropic recently hired Clive Chan, an early member of OpenAI’s custom chip team. A move reported as a signal that the project has moved beyond casual exploration. The talks with Samsung are not exclusive; Anthropic is also in discussions with Microsoft regarding its Maia AI chips and with UK-based startup Fractile about inference hardware, according to the report.
An Anthropic spokesperson did not confirm or deny the Samsung discussions when contacted by TechCrunch, saying only that the company’s diversified hardware stack, including chips from Amazon Web Services, Google and Nvidia, “will continue to be pivotal to its compute strategy.” The company said it had nothing further to add regarding a potential Samsung partnership.
The discussions follow a similar move by rival OpenAI, which unveiled a custom inference chip called Jalapeno last month, built in partnership with Broadcom and manufactured by Taiwan’s TSMC. Google has developed its own Tensor Processing Units. Amazon has built Trainium and Inferentia chips for AWS, and Meta has reportedly struck a chip deal with Samsung as well.
Nvidia remains dominant in the AI chip market, holding an estimated 74% share. According to data cited by The Information, a figure that has grown even as competitors pursue custom silicon, driven by AI training demand outpacing the growth of alternative chip suppliers.
Samsung has been expanding its foundry customer base as it competes with TSMC, (the industry’s dominant contract chip manufacturer). for leading-edge production contracts. The company’s foundry division already produces chips for Tesla, Nvidia and Apple, according to the Korea Herald. Samsung Foundry President Han Jin-man told employees last month that the division could return to profitability by 2028.
The timing of Anthropic’s talks follows a setback for Samsung’s prior custom chip discussions with OpenAI. Samsung had been developing an Arm-based inference chip for OpenAI before those talks stalled in early June. Due to what Korean media described as strategic differences between the two companies. OpenAI chief executive Sam Altman subsequently canceled a planned visit to Seoul that had been expected to advance that relationship.
Samsung and SK Group separately announced a combined $518 billion, decade-long investment to build four new chip manufacturing facilities in South Korea. A commitment industry analysts say gives Samsung the capacity expansion needed to support large anchor customers such as Anthropic.
Anthropic has separately committed to more than $100 billion in AWS infrastructure purchases and a $50 billion U.S. data center buildout with Fluidstack, according to reporting cited by Let’s Data Science. The company was valued at $965 billion following a funding round in May that included Samsung as an investor.
