DeepSeek, the Hangzhou AI company that triggered a U.S. tech stock sell-off in January 2025 with its low-cost R1 reasoning model, is raising approximately 50 billion yuan ($7.4 billion) in its first external funding round. The deal could value the company at up to 400 billion yuan ($59 billion), according to Reuters, making it one of the largest private technology financings ever arranged in China.

Investor Roster and Capital Structure

Tencent is weighing roughly 10 billion yuan and battery maker CATL about 5 billion yuan, positioning the two as the round’s anchor backers, Reuters reported. Additional participants include NetEase, JD.com, Hong Kong-based IDG Capital, Monolith Capital, and several state-backed Chinese AI funds, according to TechStartups.

Founder Liang Wenfeng is contributing 20 billion yuan ($2.9 billion) of his own capital, a substantial personal commitment from someone who has financed the company largely through his quantitative hedge fund, High-Flyer.

The investor base is deliberately compact. Fewer than ten participants are expected in the round, with no retail allocation and no roadshow, Proactive Investors noted. Tencent brings distribution. CATL brings energy infrastructure capability. The national AI fund brings state backing.

The Valuation Gap

DeepSeek’s $59 billion ceiling valuation is a fraction of its American counterparts. Anthropic filed confidentially for an IPO on June 1 at a $965 billion valuation. OpenAI raised $122 billion in March at $852 billion. SpaceX, which merged with xAI earlier this year, is targeting at least $1.8 trillion in its IPO, according to Proactive Investors.

DeepSeek is raising less than 5% of the combined $155 billion that SpaceX and Alphabet are seeking from public markets in the same window.

Compute Is the Constraint

The capital appears directed at the one input that efficiency alone cannot produce: compute hardware. U.S. export controls have throttled high-end accelerator shipments into China, and DeepSeek needs billions in fresh capital to secure domestic silicon, stockpile available chips, and fund next-generation models without relying on Liang’s personal fortune, TechTimes reported.

DeepSeek built its R1 and V3 models on a substantially smaller compute budget than Western labs, partly by necessity given export restrictions. That achievement challenged the prevailing assumption that frontier AI required the capital expenditure only U.S. hyperscalers could sustain.

The Agent Infrastructure Angle

For teams building on open-weight models, a better-capitalized DeepSeek extends the price pressure its models already apply across the industry. DeepSeek’s open releases give developers leverage against incumbent API pricing from OpenAI and Anthropic. A $7.4 billion war chest means DeepSeek can sustain aggressive pricing and model releases longer, strengthening the case for agent builders who route workloads to the cheapest capable model available.

The performance gap remains real. Independent testers have found DeepSeek’s models competitive on reasoning and coding but weaker on tool use, long-context stability, and real-world reliability compared to the best closed systems, according to TechTimes. For agent orchestration, where tool-use reliability is critical, that gap matters more than raw benchmark scores.

The round has not been publicly confirmed by DeepSeek, Tencent, or CATL. People familiar with the talks say it could close within weeks.