The chip is being designed for inference, the stage at which a trained model generates responses for users, rather than for training, the sources said.

The effort began roughly a year ago and remains at an early stage, with the Hangzhou-based company holding discussions with chip-design, foundry and memory partners and quietly increasing its hiring of chip-design engineers, without public job postings.

If successful, the move would mark a significant strategic shift for a company that has become a standard-bearer for China’s AI ambitions, and could add further pressure to Huawei, whose share of the domestic AI chip market is already being eroded by rivals Alibaba and Baidu.

DeepSeek has relied on both Nvidia and Huawei hardware to date; its R1 model was trained on Nvidia’s H800, a chip designed to meet earlier US export-control thresholds. Nvidia shares slipped by around 1.6–2% in premarket trading following the report.

Mimicking OpenAI and Anthropic’s playbook

The move also shakes up the wider tech landscape, mimicking the hardware strategies of OpenAI and Anthropic as both look to control their own recurring cost centres. OpenAI unveiled its first custom inference chip, Jalapeno, developed with Broadcom, last month, while Anthropic has reportedly been weighing development of its own AI silicon.

While the market’s attention is on the technical pivot away from Nvidia and Huawei, the development points to a larger geopolitical reality: Western chip restrictions appear to be accelerating Chinese firms’ push toward technological independence rather than containing it.

Bill Conner, president and CEO at AI integration and automation company Jitterbit, said the winner of the AI race won’t just be decided by hardware or processing power, but by governance and trust:

“DeepSeek may be challenging US’s big tech dominance and demonstrates that advanced AI development can be achievable with fewer resources, a particularly significant outcome for China amid further US chip restrictions. The question, then, is whether current restrictions are containing competitors, or are they quietly accelerating their independence?

“Despite the media’s warning that China could be overtaking the US, the AI race is not solely about processing power, model performance or affordability. Trust, governance, interoperability and accountability are decisive differentiators. Nations that embed responsible AI principles early will shape global standards.”

Manufacturing hurdles remain

For DeepSeek, the strategic stakes are high as US export controls prevent Chinese firms from buying Nvidia’s most advanced chips, and Beijing has been pressing domestic technology champions to develop alternatives. Manufacturing constraints add a further hurdle: US restrictions bar Chinese chip designers from the most advanced overseas foundries, while separate curbs limit access to high-bandwidth memory, a component critical to inference chips.

The chip development coincides with DeepSeek’s first move to accept outside capital. The company was reported in June to be raising $7bn in its maiden funding round, at a valuation of between $52bn and $59bn.

RELATED STORIES

The $7bn opportunity: Why DeepSeek still dominates the AI market

China’s Big Fund in talks to lead DeepSeek investment at $45bn valuation