DeepSeek, a Chinese startup known for its large language models that compete with those from US firms like OpenAI and Anthropic, is planning to enter the silicon business. According to Reuters, the company has been working on this transition for about a year, engaging with potential partners in the hardware and silicon sectors and hiring engineers for the project. The focus is on data center chips for inference, not training, as part of a broader effort to lessen dependence on both Huawei and Nvidia.
The decision comes amid US export controls that have limited Nvidia's presence in China, where Huawei controls about half of the data center chip market. DeepSeek is not alone in this endeavor; Chinese tech giants like Alibaba and Baidu are also making similar moves. The push for self-sufficiency in silicon is driven by the need to navigate export restrictions and secure access to critical compute resources as AI models scale.
The company's goal is to establish a competitive position in a market where data center access is likely to remain constrained, with multiple firms vying for compute power as they expand their AI capabilities. The shift reflects a growing trend among both Chinese and US-based AI companies to develop custom silicon solutions. Source: arstechnica
