Deepseek builds inference AI chip and seeks $7B funding

Deepseek is designing its own AI chip aimed at inference, and the company is exploring outside capital as it pursues that hardware plan. Three people familiar with the matter told Reuters the chip is for inference, not training, and Deepseek is looking to bring in $7 billion at a valuation of $52 to $59 billion.

What is Deepseek building?

Deepseek is building an inference-focused AI chip, not a chip for training new models, according to three people familiar with the matter. The design effort targets the phase where a trained model generates responses for users; the three sources explicitly distinguished inference from training as the chip's purpose.

The project remains in early stages. The startup has been quietly hiring chip engineers for months without public job listings and is talking with chip design, manufacturing, and memory companies as part of the effort.

How is Deepseek proceeding and what challenges will it face?

Deepseek is engaging chip design firms, manufacturing partners, and memory companies while hiring engineers, but the company faces hurdles from export controls and a difficult supply chain. Reuters says the startup is talking to chip design, manufacturing, and memory companies, and has quietly hired chip engineers for months without public job listings.

One clear constraint is US export controls, which the source says have cut Chinese companies off from access to the most advanced chips and memory. That restriction is presented as a structural barrier to procuring top-tier components and tooling, a headwind for any Chinese company trying to internalize hardware production.

The move could reduce Deepseek's reliance on existing suppliers. Reuters notes the chip could lower the company's dependence on Nvidia and Huawei chips, though building production-ready silicon remains a long and costly path.

Why it matters

Deepseek attempting to design its own inference silicon signals a push by Chinese AI startups to control more of their stack and reduce dependency on external suppliers. The startup is also pursuing a large outside financing round: Reuters says Deepseek is looking to bring in $7 billion at a valuation of $52 to $59 billion, figures that underline the scale of the company's ambitions.

If Deepseek succeeds, it could alter procurement dynamics by shifting workloads away from Nvidia and Huawei hardware for inference tasks. The story sits alongside other firms pursuing bespoke chips: Reuters points out OpenAI and Anthropic are also working on their own chips, showing hardware development is no longer limited to incumbent chipmakers.

What to watch

Watch whether Deepseek closes the reported $7 billion financing and at what valuation within the $52 to $59 billion range Reuters cited; that will indicate how much external capital the company can marshal for hardware. Also watch for public hires or formal partnerships with chip design, manufacturing, or memory firms, which would signal the project moving from quiet engineering to industrial execution.

A successful early prototype announcement or a named foundry partnership would be the next concrete milestone to confirm progress. Conversely, any public statement that the effort is suspended or scaled back would show how export controls and supply constraints are shaping Chinese AI hardware plans.