Anthropic, the developer of the Claude AI assistant, is in early discussions with Samsung Electronics to manufacture a custom artificial intelligence chip, according to multiple reports. The talks, first disclosed by The Information and confirmed by TechCrunch and Korean media, represent a potential first step for Anthropic toward building its own silicon stack as compute demand becomes a critical cost and capacity constraint in the AI industry.
The negotiations focus on Samsung’s 2-nanometer process and its advanced packaging technology, which connects different processor components. While the discussions are preliminary and no design or manufacturing commitment has been finalized, Anthropic has also hired Clive Chan, a former member of OpenAI’s early custom-chip team. This move aligns with a broader industry trend: OpenAI partnered with Broadcom in 2024 to design the Jalapeño inference chip, Google has its tensor processing units (TPUs), Amazon developed Trainium, and Meta and Microsoft are working on their own silicon.
For Anthropic, a proprietary chip could reduce dependence on Nvidia GPUs, which currently hold about 74% of the AI chip market, and optimize inference costs for Claude-specific workloads. The company stated that chips from Amazon Web Services (Trainium), Google (TPUs), and Nvidia will remain central to its compute strategy, and it is also considering hardware from Microsoft and UK startup Fractile. The possibility of adding a Samsung-made processor underscores how leading AI firms are shifting from software-only models to controlling the full infrastructure stack.
For Samsung, winning an Anthropic mandate would be a strategic milestone in its effort to close the manufacturing gap with Taiwan Semiconductor Manufacturing Co., which held an estimated 73% of the pure-play foundry market in early 2026. Samsung has already secured a $16.5 billion Tesla AI chip deal in 2025 and is reportedly in talks with Google for a next-generation tensor processor using its 2nm technology.
The news comes as Anthropic resolves a standalone dispute with the U.S. Commerce Department that had restricted exports of its Mythos and Fable models. On July 2, the restrictions were lifted after Anthropic implemented new safety safeguards tested by the Center for AI Standards and Innovation, ending a weeks-long standoff that had frustrated some Silicon Valley leaders and foreign partners.
