The Chinese AI venture DeepSeek has halted its second funding round, initially aiming to raise at least $1.5 billion. This decision stems from dissatisfaction expressed by the startup's founder, Liang Wenfeng, regarding media coverage of his remarks to investors, according to Bloomberg.
DeepSeek has informally notified some potential investors that it will not be signing agreements in the upcoming days as they had anticipated. However, the company may revisit the funding process later.
Details from the Meeting
A meeting took place on May 20 and lasted about four hours. Portions of the meeting minutes were leaked to Yicai, which published excerpts on July 23. As of the publication date, neither DeepSeek nor Liang had confirmed the authenticity of the document.
According to the minutes, Liang pointed out that China's primary shortfall compared to the U.S. in artificial intelligence is not due to a lack of talent but rather limited access to computing resources. He also highlighted restrictions on purchasing advanced graphics processors and lower investment levels in the sector compared to the U.S.
The document suggests that DeepSeek lags behind leading U.S. developers by 12 to 18 months. Notably, the startup has achieved comparable results using only about 5% of the processing power of its U.S. counterparts. The company aims to reduce this gap to three to six months.
Moreover, DeepSeek is collaborating with Huawei to optimize its models for Ascend chips, thereby reducing its reliance on Nvidia's CUDA software environment. Liang indicated that the V3 model operated on American chips but utilized their own TileLang compiler, minimizing dependence on CUDA. He noted that there is an opportunity for the Chinese chip industry to substitute imports, although production volumes remain a bottleneck.
Valuation and IPO Preparations
In its second funding round, DeepSeek aimed to secure at least 10 billion yuan ($1.5 billion). The company's pre-deal valuation was approximately 480 billion yuan ($71 billion).
The startup's first external funding round was concluded at the end of May, raising over 50 billion yuan ($7.4 billion) at a valuation of around 367.5 billion yuan ($54.3 billion). Investors included Tencent, CATL, NetEase, JD.com, IDG Capital, and the National AI Industry Investment Fund, with Liang himself contributing the largest single investment of 20 billion yuan.
In parallel, the company has begun preparations for an initial public offering and may submit its application this year. However, in the leaked transcript, Liang mentioned that considerations of going public or maximizing profits were not part of the project's inception. He stated that the first few dozen employees did not view their work through that lens.
Additionally, it was reported in early July by Reuters that DeepSeek is developing its own processor for its models.