DeepSeek has temporarily suspended its latest round of fundraising discussions after informing some prospective investors that planned investment agreements would not be finalized in the near term, Bloomberg reported on Saturday, citing people familiar with the matter.

The halt comes on the heels of widely circulated comments attributed to founder Liang Wenfeng concerning China's artificial intelligence capabilities and its dependence on Nvidia chips. Sources indicated that Liang was displeased with reports characterizing remarks allegedly made during investor meetings tied to DeepSeek's June funding round, which secured $7 billion.

The Chinese AI startup had been seeking at least 10 billion yuan ($1.5 billion) in new capital at a pre-money valuation of approximately 480 billion yuan ($71 billion). This effort began just weeks after the company closed a record $7 billion financing round led by investors including Tencent, CATL, and a state-backed AI investment fund.

Although the fundraising process has been paused, discussions are still active and DeepSeek may restart the process at a later stage. For AI developers pursuing large-scale model training and deployment, access to capital is closely tied to the ability to secure computing infrastructure, including high-performance accelerators and data center capacity.

The company is also preparing groundwork for an initial public offering that could be filed later this year, as it seeks additional capital to expand computing capacity. DeepSeek continues to prioritize open-source AI development and long-term research into artificial general intelligence.

Founder: Computing Resources, Not Talent, Separate China from the US

DeepSeek reportedly believes that computing power — rather than talent — is the primary factor distinguishing China's AI sector from that of the United States, according to unconfirmed investor meeting minutes reported by Yicai.

Wenfeng reportedly stated that restrictions on advanced Nvidia AI chips and comparatively lower capital investment have constrained China's AI progress. Advanced AI chips have become a central constraint for Chinese AI companies as US export controls limit access to some of Nvidia's most capable accelerators used for training and running frontier models. However, the company estimates it trails leading US firms by only 12 to 18 months while utilizing roughly one-twentieth of their computing resources.

According to the reported meeting details, DeepSeek currently possesses computing resources equivalent to approximately 20,000 Nvidia H-series GPUs and aims to narrow the technology gap to three to six months.

The startup is also said to be collaborating with Huawei to optimize its models for Ascend chips, reducing its reliance on Nvidia's CUDA software ecosystem.