DeepSeek Seeks $300 Million in First Fundraise at $10+ Billion Valuation

DeepSeek, a barely two-year-old Chinese AI startup, is raising its first external funding of at least $300 million at a valuation exceeding $10 billion after being entirely backed by its parent hedge fund High-Flyer Capital Management, which posted 56.6% returns in 2025. The company's open-source R1 reasoning model, trained for just $5.6-6 million using older export-restricted Nvidia H800 chips, drew comparisons to OpenAI's systems and triggered roughly $1 trillion in erased market value across AI companies. Surging demand for DeepSeek's dramatically lower-cost models has pushed infrastructure to its limits, causing frequent outages and climbing costs that necessitate the fundraise. Domestic Chinese investors are most likely to participate, as U.S. venture firms face regulatory pressure and national security concerns, while export controls on advanced chips continue shaping the company's trajectory.

DeepSeek's first external fundraising round at $10+ billion valuation

• ▪DeepSeek is in talks to secure at least $300 million in its first fundraising round.
• ▪DeepSeek valuation estimates have ranged from as low as $1 billion to well above $20 billion.
• ▪DeepSeek's first fundraising round could value the Chinese AI startup at more than $10 billion.
• ▪DeepSeek founder and CEO Liang Wenfeng has long resisted outside investors.
• ▪DeepSeek is preparing to raise outside capital for the first time.
• ▪High-Flyer Capital Management posted returns of about 56.6% in 2025.
• ▪High-Flyer Capital Management is the quantitative hedge fund that gave birth to DeepSeek.
• ▪DeepSeek has been backed entirely by High-Flyer Capital Management until now.

R1 model's market impact and cost-efficient approach

• ▪DeepSeek engineers focused on efficiency techniques such as KV cache compression and selective activation.
• ▪DeepSeek released its R1 model in early 2025.
• ▪DeepSeek's R1 model drew comparisons to OpenAI's top systems.
• ▪DeepSeek's R1 model was reportedly trained for around $5.6 to $6 million.
• ▪The market reaction to DeepSeek's R1 model was estimated at roughly $1 trillion in erased value across AI-related companies.
• ▪DeepSeek's R1 model was trained using older, export-restricted Nvidia H800 chips.

Infrastructure demands and scaling challenges

• ▪DeepSeek needs more GPUs, more servers, and far more compute capacity to keep up with demand.

Strategic implications for global AI competition

• ▪DeepSeek shook Silicon Valley and Wall Street with its low-cost R1 reasoning model.

Perspective of DeepSeek competitors (OpenAI, other well-funded AI labs)

• ▪The $1 trillion market value erasure across AI-related companies reflects investor concerns about the competitive moat of high-spending AI firms.

Perspective of High-Flyer Capital Management

• ▪High-Flyer Capital Management's willingness to accept external investors marks a shift from founder Liang Wenfeng's long-standing resistance to outside capital.
• ▪High-Flyer Capital Management's 56.6% returns in 2025 demonstrate the financial success that enabled exclusive funding of DeepSeek until now.

Perspective of Domestic Chinese investors

• ▪Domestic Chinese investors are positioned as the most likely participants in DeepSeek's $300 million fundraising round due to regulatory advantages.

1 source