Inside DeepSeek: The Rise of a Chinese AI Powerhouse
DeepSeek, the Chinese artificial intelligence startup that rattled global markets with its low-cost R1 model last year, increased its registered capital by 50% on April 27, with founder and CEO Liang Wenfeng personally funding the entire increase and dramatically raising his direct equity stake from 1% to 34%, according to business registry data from Tianyancha cited by multiple outlets.
The Hangzhou-based company lifted its registered capital from 10 million yuan to 15 million yuan, with Liang subscribing to the full 5.1 million yuan increase. While Liang's direct personal stake surged, his overall control of the company remains unchanged, as he previously held his interest largely through shell corporations.
First External Funding Round
The capital restructuring comes as DeepSeek pursues its first-ever external funding round. The Information first reported on April 17 that DeepSeek was seeking at least $300 million at a valuation of $10 billion. Within days, the target valuation had roughly doubled, with Alibaba and Tencent entering discussions to invest at a valuation exceeding $20 billion, according to Bloomberg, citing The Information.
The Wall Street Journal reported that Shunwei Capital, a venture firm backed by Xiaomi founder Lei Jun, is also in talks to invest. According to Chinese financial publication Caijing, Alibaba and Tencent are expected to invest a combined $1.8 billion at a roughly $20 billion valuation, though terms have not been finalized.
The fundraise marks a sharp reversal for a company that had long rebuffed outside capital. As recently as March 2025, Liang told colleagues he was wary that external investors might interfere with DeepSeek's operations, and was particularly cautious about government-linked backers, according to The Wall Street Journal.
A Reclusive Founder
Liang, now about 41, has remained almost entirely absent from public life since a celebrated appearance at a symposium hosted by China's Premier Li Qiang on January 20, 2025 — the same day DeepSeek released its R1 model. Since then, from "Liang Wenfeng down, the team has gone collectively silent," according to Recode China AI.
The capital increase and funding push coincide with the launch of DeepSeek V4, the company's long-awaited successor model released in preview on April 25. The model, built partly on Huawei chips, claims to rival offerings from OpenAI and Anthropic in coding and reasoning tasks, though Bloomberg reported it "doesn't meaningfully narrow the US lead in AI capabilities".
DeepSeek originally operated on the profits of High-Flyer, Liang's quantitative hedge fund, which posted average returns of 56.6% across its funds in 2025 and manages more than 70 billion yuan. The shift to outside capital signals that the escalating costs of frontier AI development may have outpaced even those resources.
