DeepSeek has begun developing its first dedicated processor designed to run artificial intelligence models. According to Reuters, the chip is designed for the inference stage, which processes user queries and generates answers. The company aims to reduce its dependence on NVIDIA and Huawei chips.
The project is still in its early stages. DeepSeek is in talks with companies operating in the fields of microcircuit design, semiconductor manufacturing and memory technologies. At the same time, the startup has been recruiting chip engineers in recent months without posting open vacancies.
A similar strategy is being pursued by OpenAI and Anthropic. Both companies are developing their own chips to give them more control over the hardware infrastructure on which artificial intelligence models run.
However, DeepSeek faces difficulties in accessing the most advanced AI accelerators and high-performance memory technologies due to US export restrictions. Reuters reports that the company is preparing to attract foreign investment for the first time in parallel. DeepSeek is reportedly planning to raise $7 billion in investment, at a valuation of approximately $52–59 billion.
