Chinese AI company DeepSeek is reportedly close to raising at least $12 billion, a financing round that would dramatically expand the resources available to one of the most closely watched challengers to U.S. frontier-model companies.
Bloomberg reported that DeepSeek is nearing at least 80 billion yuan, or about $12 billion, in new funding, after initially seeking roughly 50 billion yuan.
The final round could approach 100 billion yuan, according to people familiar with the financing, while Tencent Holdings and battery giant CATL are among the investors committing some of the largest amounts.
Reuters separately reported the Bloomberg figures but said it could not immediately independently verify the fundraising report.
Investor demand has moved beyond DeepSeek’s original target
The scale of the reported financing is notable because DeepSeek built much of its reputation around efficiency.
Its earlier models drew industry attention by challenging assumptions that frontier-level capabilities automatically required spending on the same scale as the largest American laboratories.
Now the company itself appears to be entering a much more capital-intensive stage.
Bloomberg reports that demand increased following the release of DeepSeek’s latest model and that the financing is taking place ahead of a potential initial public offering in early 2027.
That creates an interesting transition: a company known for doing more with constrained resources may soon have one of the largest pools of private capital in the AI sector.
DeepSeek is also moving closer to Huawei’s hardware ecosystem
The financing story is happening alongside an equally important infrastructure shift.
Reuters reported on September 30 that DeepSeek had partnered with Huawei to develop programming tools optimised for Ascend AI chips, part of China’s effort to reduce dependence on Nvidia’s hardware and software ecosystem.
DeepSeek’s own open-source repositories now include DeepGEMM-Ascend, which it describes as a port of its matrix-multiplication library to Huawei’s Ascend platform.
The library supports BF16, FP8 and FP4 operations and is designed to maintain the same API and development workflow used by DeepGEMM on other platforms.
That detail matters because Nvidia’s advantage is not only the chip itself. Developers have spent years building around CUDA and its surrounding software ecosystem.
Matching hardware therefore requires matching developer tooling.
Funding and infrastructure are becoming the same AI story
The potential $12 billion round cannot be viewed only as a valuation milestone.
Frontier AI increasingly requires capital for compute, research talent, inference capacity, model training and the engineering ecosystem underneath all of them.
DeepSeek’s collaboration with Huawei suggests that part of that expansion may happen inside a Chinese stack designed to function with less dependence on Nvidia.
For founders and infrastructure investors, this makes DeepSeek a useful signal about where the AI competition is heading.
The race is no longer simply which laboratory builds the strongest model.
It is increasingly about which ecosystem can finance and control the chips, programming tools, model architectures and deployment infrastructure required to keep improving those systems.
If DeepSeek completes a financing round of this size while deepening its support for Huawei hardware, it will strengthen something more consequential than one company’s balance sheet: China’s attempt to build a vertically integrated alternative to the dominant U.S. AI computing stack.
