Facing US export controls, China’s DeepSeek plans to make its own chips (arstechnica.com)

Chinese startup DeepSeek, known for developing large language models that rival those from US firms like OpenAI, is planning to produce its own data center chips for inference. This strategic shift comes in response to US export controls that limit access to key hardware from companies like Nvidia. Over the past year, DeepSeek has engaged with potential hardware partners, supporting its move to reduce dependency on established players like Huawei, which currently dominates the Chinese data center chip market. This initiative highlights a significant trend amid increasing chip restrictions from the US, as both Chinese and American companies explore custom chip designs to gain more control over their technology stack. Recently, OpenAI and Broadcom unveiled a new chip, Jalapeño, aimed at scaling inference operations, showcasing a broader competitive landscape where firms seek innovative solutions amid constrained data center resources. DeepSeek's entry into silicon production could bolster China's AI capabilities and reshape the chip supply dynamics in a market where demand for computational power continues to surge.