Anthropic explores partnership with Samsung for custom AI chip production

Anthropic has begun the early stages of developing a custom AI chip and is reportedly in talks with Samsung Electronics over the possibility of joining the project as a manufacturing partner.

According to The Information, Anthropic's custom AI chip plans could have significant implications for the broader semiconductor industry. The report said the startup is evaluating Samsung's 2-nanometer manufacturing process and advanced chip packaging capabilities, a move that could challenge Taiwan Semiconductor Manufacturing Co.'s dominance in the foundry market if Samsung secures the contract.

The report also noted that the project remains in its early stages, with no detailed chip design or manufacturing work underway. Anthropic has reportedly strengthened its chip development efforts by hiring Clive Chan, an early member of OpenAI's custom silicon team.

The development reflects a broader industry trend, with major AI companies increasingly designing proprietary chips to reduce reliance on third-party suppliers. OpenAI partnered with Broadcom on its custom AI chips, while Google, Amazon, Meta, and Microsoft have all launched similar initiatives. Despite growing competition, Nvidia continues to dominate the AI chip market, with estimated market share at 74%.|

More from Tech

Samsung's foundry ambitions are receiving support from multiple fronts. Intensifying the competition, Google is also reportedly considering Samsung to manufacture part of a future Tensor Processing Unit (TPU).

Amid these developments, Samsung Group and SK Group—the parent companies of Samsung Electronics and SK Hynix—have announced a combined investment of $518 billion over the next decade to build four new memory chip plants in South Korea. The commitment underscores the scale of investment being poured into South Korea's semiconductor ecosystem as it competes for AI infrastructure opportunities.

As reported by The Information, Anthropic's custom AI chip ambitions will not replace its existing hardware partnerships. The company said Amazon Web Services' Trainium chips, Google's Tensor Processing Units, and Nvidia's graphics processing units (GPUs) will continue to play a central role in scaling its compute infrastructure. Thealso reported that Anthropic is in discussions to use chips from Microsoft and UK-based startup Fractile, reinforcing its deliberate