DeepSeek draws China’s biggest AI bet
DeepSeek has become China’s most valuable artificial intelligence startup after raising more than $7.4 billion in its first outside funding round, a deal that values the company at more than $50 billion and places it at the center of Beijing’s campaign to build a homegrown rival to U.S. AI leaders. The Wall Street Journal reported the financing included major Chinese corporate backers and founder Liang Wenfeng’s own capital. :contentReference[oaicite:0]{index=0}
The round marks a dramatic turn for a company that was little known outside technology circles before its low-cost reasoning model shook global markets in January 2025. DeepSeek’s R1 release promoted open-source access and claimed performance comparable to leading U.S. reasoning systems, helping fuel questions about whether American companies’ vast spending on chips and data centers would remain a durable advantage. :contentReference[oaicite:1]{index=1}
Investors in the new round include Tencent, battery maker CATL, JD.com, NetEase, IDG Capital and Monolith Management, according to the Journal. Tencent invested about $1.5 billion, CATL about $740 million and a Chinese government AI fund about $150 million, the report said. Liang contributed about $3 billion and retained control through a limited partnership structure with a five-year holding requirement for outside investors. :contentReference[oaicite:2]{index=2}
The funding gives DeepSeek one of the largest war chests in China’s technology sector at a time when U.S. export controls have made advanced Nvidia chips harder for Chinese companies to obtain. The company is expected to use the money for research, computing infrastructure and commercial products, including agent-style AI tools that can carry out multistep tasks for businesses. :contentReference[oaicite:3]{index=3}
A challenge to U.S. dominance
DeepSeek’s rise has been closely tied to Liang, a former quantitative hedge fund manager who founded the Hangzhou-based company after building computing capacity through High-Flyer Capital Management. Forbes identifies Liang as the founder of both DeepSeek and High-Flyer, and said DeepSeek gained global attention after releasing R1 in January 2025. :contentReference[oaicite:4]{index=4}
The company’s technical pitch is efficiency. DeepSeek’s V3 technical report described a mixture-of-experts model with 671 billion total parameters, with only 37 billion activated for each token, and said the system used architectural choices aimed at cheaper training and inference. The report said V3 required 2.788 million H800 GPU hours for full training. :contentReference[oaicite:5]{index=5}
That approach has made DeepSeek a symbol of China’s effort to compete despite restrictions on advanced semiconductors. It has also intensified price pressure across the AI industry, where cheaper open-source and Chinese models are increasingly used for routine business tasks while companies reserve more expensive systems for complex work. The Journal reported this month that such alternatives are contributing to a broader AI price war. :contentReference[oaicite:6]{index=6}
Still, DeepSeek remains far smaller than the largest U.S. AI companies by valuation and faces political risk abroad. U.S. officials have scrutinized Chinese AI firms as strategic competitors, and reports this month said Washington had considered adding DeepSeek to a trade blacklist. :contentReference[oaicite:7]{index=7}
For Beijing and Chinese investors, the new financing is a vote of confidence that DeepSeek can turn a technical breakthrough into a durable national champion. For rivals in Silicon Valley, it is another sign that the race to dominate AI will be fought not only on model quality, but on cost, access and the ability to build under geopolitical pressure.
