Anthropic in Early Talks With Samsung to Manufacture Custom AI Chip

Anthropic, the AI company behind the Claude model family, is in early-stage talks with Samsung Electronics to manufacture a custom AI chip, The Information reported on July 2, 2026. The project remains exploratory, with Anthropic yet to decide the chip's purpose, performance targets, or how it would fit into a server architecture.

Anthropic has already hired chip engineers to support the initiative, according to The Decoder. The move comes roughly one week after OpenAI announced its own custom AI chip, codenamed Jalapeno, developed in partnership with Broadcom.

Early stage, undefined scope

The discussions with Samsung are preliminary. Anthropic has not determined whether the chip would be used for training, inference, or both. No timeline for a tapeout or production has been disclosed. The company continues to rely on Nvidia GPUs for its current infrastructure and has publicly stated that Nvidia remains a critical partner.

• Anthropic has not decided the chip's power envelope or memory configuration.
• The chip's role in a server rack has not been defined.
• Samsung's foundry business would handle manufacturing if a deal is reached.
• Anthropic is also working with Micron to use Claude for designing better HBM, DRAM, and SSDs for AI workloads, though that partnership does not involve chip fabrication.
• OpenAI's Jalapeno chip is expected to be used for inference and is slated for production in 2027.

Industry context and what comes next

Anthropic's chip exploration signals a broader trend among AI labs seeking to reduce dependence on Nvidia and control infrastructure costs. Custom silicon allows companies to optimize for their specific model architectures and workloads)Skip. However, designing and manufacturing a chip from scratch is a multi-year, billion-dollar endeavor. Samsung's foundry has faced yield challenges in advanced nodes, which could complicate the timeline.

Anthropic has not commented publicly on the reports. The company is expected to continue using Nvidia GPUs for the foreseeable future while the chip project remains in the concept phase. No formal announcement is imminent.