Chinese artificial intelligence startup DeepSeek is reportedly developing its first in-house AI chip in a strategic move aimed at reducing its dependence on chips supplied by Nvidia and Huawei, according to a Reuters report citing three sources familiar with the matter.
The chip is being designed primarily for AI inference, the process where trained AI models generate responses to users, rather than for training new AI systems. If successful, the project would represent a major expansion beyond DeepSeek’s traditional focus on developing advanced AI models and could reshape China’s fast-growing AI semiconductor industry.
DeepSeek rose to global prominence after launching highly efficient AI models that attracted worldwide attention for their strong performance at comparatively low costs. The company has been widely viewed as one of China’s leading AI innovators.
Its move into chip development signals an effort to gain greater control over the hardware powering its AI products while reducing reliance on external suppliers amid ongoing U.S. technology restrictions.
Sources told Reuters that the Hangzhou-based company began exploring custom chip development about a year ago. Since then, DeepSeek has quietly increased recruitment of semiconductor engineers while holding discussions with chip designers, manufacturing foundries and memory suppliers. Hiring has reportedly been conducted privately without public job advertisements.
DeepSeek has not publicly commented on the project.
DeepSeek has historically relied on both Nvidia and Huawei processors to train and deploy its AI models.
The company’s popular R1 reasoning model was trained using Nvidia’s H800 chips, which were specifically designed for the Chinese market before Washington later prohibited their export to China.
Since those restrictions, DeepSeek has increasingly adopted Huawei’s Ascend AI chips. Earlier this year, the company introduced its V4 model optimized for Huawei’s Ascend platform, while Huawei confirmed that its processors were used during training of DeepSeek’s V4-Flash model.
The growing reliance on domestic technology comes as the Chinese government encourages local technology companies to reduce dependence on foreign semiconductor suppliers.
Although Huawei has emerged as China’s largest AI chip supplier following U.S. export controls, its dominance is increasingly being challenged.
Chinese technology giants Alibaba and Baidu have also accelerated development of proprietary AI chips, intensifying competition within China’s estimated $50 billion AI semiconductor market.
DeepSeek’s entry into the sector could further diversify China’s domestic AI hardware ecosystem.
DeepSeek’s strategy mirrors a growing trend among major AI companies seeking greater independence from Nvidia.
Recently, OpenAI introduced its first custom inference chip, named Jalapeno, developed alongside Broadcom. Meanwhile, AI startup Anthropic has also reportedly explored building its own AI processors.
Owning proprietary chips allows AI developers to optimize performance, lower operating costs and reduce dependence on third-party suppliers.
Demand for inference chips has surged as AI applications become more widely adopted across industries.
Unlike training chips, inference processors handle everyday AI tasks such as generating chatbot responses, translating languages and creating images. These chips are typically designed to consume less power while delivering faster responses for real-world applications.
However, developing competitive AI processors remains technically demanding and expensive. Chip development often requires years of engineering work, while U.S. export restrictions continue limiting Chinese companies’ access to the world’s most advanced semiconductor manufacturing facilities and high-bandwidth memory technologies essential for high-performance AI chips.
DeepSeek’s hardware ambitions coincide with a broader shift in its business strategy.
Reuters previously reported that the company is preparing to raise approximately $7 billion in its first external funding round, potentially valuing the AI startup between $52 billion and $59 billion.
The fundraising would mark a significant departure from DeepSeek’s long-standing policy of rejecting outside investment, providing additional resources to support its expansion into semiconductor development and strengthen its position in the increasingly competitive global AI industry.
