Anthropic, Focused on In-House Chip Development, Discusses AI Chip Production with Samsung

Anthropic, the developer of the AI model Claude, is reportedly in discussions with Samsung Electronics regarding the production of AI chips.

Anthropic has begun early-stage work on developing its own AI chips and is in talks with Samsung Electronics as a potential manufacturing partner, the U.S. IT news outlet The Information reported on July 2 (local time), citing multiple sources.

Anthropic is considering utilizing the 2-nanometer (nm) manufacturing process and advanced packaging facilities of Samsung Electronics' foundry business division.

The 2nm process is currently the most advanced chip manufacturing technology in the industry, characterized by higher processor density and improved power efficiency.

Additionally, advanced packaging technology can reduce bottlenecks by placing the main processor closer to memory chips, thereby increasing data transfer speeds.

In May, during its Series H funding round, Anthropic announced that three major memory semiconductor manufacturers, led by Samsung Electronics, had participated as "strategic infrastructure partners," stating, "The technology from these companies plays a key role in the global supply of memory, storage, and logic chips."

At the time, Anthropic's mention of "logic chips" led industry observers to speculate that there was a high possibility of Samsung Electronics winning an order for Anthropic's AI chips.

This is because among the three major memory manufacturers, Samsung Electronics is the only one with a foundry business division capable of producing logic chips.


If Samsung Electronics ultimately secures the order for Anthropic's in-house AI chips, it will add another major client to its portfolio, following Tesla, Nvidia, and Apple.

Sources stated that while Anthropic is currently in discussions with several chip design firms, the project has not yet reached the stage of detailed design, testing, or manufacturing.

Anthropic, which hired Clive Chan—an early member of OpenAI's custom chip team—last month, is currently exploring the functions and performance levels of its AI chips, as well as methods for server integration.

Major AI developers are focusing on developing their own AI chips to gain power efficiency and control over their infrastructure.

Beyond Google, which has long-term experience in developing its own Tensor Processing Units (TPU), OpenAI also recently unveiled its first inference chip, "Jalapeño," in late June through a collaboration with Broadcom.

AI developers expect that these efforts will allow them to reduce their reliance on specific manufacturers like Nvidia and lower overall costs.

When asked for comment, Anthropic stated that Nvidia's graphics processing units (GPUs), Google's TPUs, and Amazon Web Services' (AWS) "Trainium" chips would continue to serve as the core of its computing resources, but declined to provide specific details regarding future plans.

Samsung Electronics declined to comment on the matter.

(Photo: AP, Yonhap News)
※ Please note: This article was translated by AI and may contain errors.