Anthropic is developing custom semiconductors optimized for its AI model, "Claude," as part of a "multi-chip strategy" to improve processing speed and power efficiency compared to general-purpose GPUs. The company announced on the 5th that it is building an in-house chip design team by hiring engineers with expertise in both hardware and software, aiming to co-design AI models and semiconductors to address large-scale customer demand. Anthropic currently uses AI chips from Amazon Web Services (AWS), Google, NVIDIA, and AMD, and plans to combine its self-designed chips with existing ones for AI training and inference. The company is reportedly in discussions with Samsung Electronics, considering using Samsung’s 2-nanometer foundry process and advanced packaging once detailed designs are finalized, according to U.S. IT media outlet The Information. The 2-nanometer process is currently used by Samsung to produce Tesla’s AI chips. Analysts suggest that securing Anthropic’s production could provide a breakthrough for Samsung’s foundry business and demonstrate its competitiveness in the 2-nanometer process. The growing competition in AI chip development is also seen as a positive factor for South Korea’s semiconductor industry, potentially driving demand for advanced foundry services, HBM, server DRAM, high-performance storage, and packaging.