- DeepSeek, the Hangzhou-based AI lab founded by quantitative hedge fund High-Flyer Capital Management’s Liang Wenfeng, is negotiating its first external funding round at a valuation of approximately $20 billion, according to Bloomberg and The Information.
- Tencent Holdings Ltd. and Alibaba Group Holding Ltd. are among the parties in active discussions to invest in the round, which was reported on April 22, 2026.
- DeepSeek has operated without outside institutional capital since its founding, making this its maiden financing event.
- No deal has been announced; negotiations were ongoing as of April 22, 2026.
What Happened
Tencent Holdings Ltd. and Alibaba Group Holding Ltd. are in discussions to participate in DeepSeek’s first external financing round, Bloomberg and The Information reported on April 22, 2026. The round is being negotiated at a valuation of approximately $20 billion. DeepSeek, which is backed by Chinese quantitative hedge fund High-Flyer Capital Management and led by founder Liang Wenfeng, has not previously accepted capital from outside investors.
Why It Matters
Bloomberg described the prospective investment as representing “a milestone for the country’s artificial intelligence sector.” DeepSeek became one of the most closely watched AI labs globally after releasing the DeepSeek-R1 reasoning model in January 2025, which posted benchmark scores comparable to OpenAI’s o1 at a reported fraction of the training cost. The release triggered a roughly 17% single-day decline in Nvidia’s stock price as investors recalibrated assumptions about compute demand for frontier AI development.
DeepSeek has operated without venture capital or strategic investment while producing models that benchmark favorably against well-funded Western labs. A round at $20 billion would rank among the largest AI company valuations globally, placing it near the tier of Anthropic and xAI in terms of reported valuation.
Technical Details
DeepSeek-R1, released in January 2025 under an MIT open-source license, uses a mixture-of-experts architecture trained with a reinforcement learning pipeline that the company said did not rely on distillation from proprietary Western models. DeepSeek reported that training the underlying DeepSeek-V3 base model cost approximately $5.576 million in GPU compute—substantially below cost estimates for comparable frontier models from Anthropic, Google, and OpenAI.
At release, DeepSeek-R1 scored 79.8% on AIME 2024 and 97.3% on MATH-500 in the company’s published benchmarks. The lab subsequently released DeepSeek-V3-0324 in March 2026, which posted competitive results on third-party evaluation platforms including Chatbot Arena, without an accompanying technical paper at release.
Who’s Affected
Tencent Cloud and Alibaba Cloud compete directly in China’s enterprise AI and cloud services markets. An equity stake in DeepSeek would give either or both companies a strategic position in a lab whose open-weight models are already widely integrated into domestic commercial applications. For Chinese enterprises, DeepSeek has become a widely adopted foundation model layer; outside investment could accelerate its inference capacity and developer ecosystem.
U.S. AI companies and chip exporters face continued competitive pressure if DeepSeek uses outside capital to expand compute infrastructure, particularly given ongoing U.S. export controls on advanced semiconductors destined for China. DeepSeek has not disclosed what GPU hardware underpins its current inference operations.
What’s Next
As of April 22, 2026, neither DeepSeek, Tencent, nor Alibaba had publicly confirmed the discussions. If a deal closes, it would likely require review under Chinese regulatory frameworks governing strategic technology investments. DeepSeek has not published a formal product roadmap, though the lab has shipped multiple model updates across its V3 and R-series lines since January 2025.
