Chinese startup DeepSeek is developing its own AI chip, according to three people familiar with the matter, a push that could reduce its reliance on Nvidia and Huawei processors it has depended on to train and run its globally popular models. The chip is designed for inference – the stage of AI computing in which a trained model generates responses for users – rather than for training new models, the sources said. YourDailyAnalysis notes that choosing to start with inference rather than training is itself a meaningful signal: inference chips are generally a lower technical bar to clear than training-grade silicon, making this a pragmatic first step rather than an attempt to leapfrog Nvidia’s most advanced hardware outright.
The strategic stakes are significant given DeepSeek’s profile. If successful, the expansion into chip development would mark a major shift for a company widely hailed in China as its AI champion, and could add to challenges already facing Huawei, which has supplied roughly half of China’s $50 billion domestic AI chip market since U.S. export bans cut off access to Nvidia’s most advanced offerings. YourDailyAnalysis flags Huawei’s position as the more exposed party in this story: its hold on that market is already weakening as Alibaba and Baidu develop competing in-house chips, and a DeepSeek entrant would be a third major customer turning at least partly self-sufficient.
The effort is still early-stage. DeepSeek has been reaching out to external partners and holding discussions with chip-design, foundry and memory companies for roughly a year, and has quietly increased hiring of chip-design engineers without public job postings, according to the sources, who were not authorized to discuss the matter publicly. That combination of external outreach and stealth hiring is consistent with a company trying to build serious chip-design capability without tipping off competitors or foreign regulators before the effort matures.
This move places DeepSeek inside a global pattern rather than a uniquely Chinese one. OpenAI unveiled its first custom inference chip, Jalapeno, developed with Broadcom, last month, while Anthropic has been weighing building its own AI chips. YourDailyAnalysis reads DeepSeek’s push as following that same logic – leading AI labs globally are converging on the view that long-term competitiveness requires some control over the underlying hardware rather than permanent dependence on Nvidia – but with an added dimension specific to DeepSeek: U.S. export controls make hardware independence not just a cost-optimization play but a matter of continued operational access to compute at all.
The obstacles are real and shouldn’t be understated. Designing a competitive AI chip typically takes years and significant capital, and DeepSeek faces additional hurdles beyond design: U.S. rules bar Chinese chip designers from the most advanced overseas foundries, while separate curbs have cut China’s access to high-bandwidth memory, a component critical to inference chips specifically. Notably, this chip push coincides with DeepSeek’s first embrace of outside capital – the company was slated to raise $7 billion in a maiden funding round valuing it between $52 billion and $59 billion, reversing its yearslong strategy of rejecting external investment, which suggests the chip ambitions may be a direct driver of the company’s sudden need for outside cash.
Watch whether DeepSeek’s fundraising closes at the reported valuation and whether any of that capital is explicitly earmarked for the chip effort, and watch for any formal partnership announcements with foundry or memory suppliers that would signal the project has moved past the exploratory stage. Your Daily Analysis views DeepSeek’s success or failure here as a meaningful bellwether for whether China’s second tier of AI champions, beyond Huawei, Alibaba and Baidu, can realistically build competitive chip capability, or whether that capability will remain concentrated in a small handful of state-favored players.
