$45 billion is the number now attached to DeepSeek’s first outside funding talks.
Only weeks earlier, reports had framed the company around a far smaller $300 million raise at roughly a $10 billion target. Before that, founder Liang Wenfeng’s posture was even clearer: DeepSeek did not need venture money. The latest version is a different story.
Why the lead investor matters
The most important part of the reported round is not just the step-up in valuation. It is the identity of the possible lead investor: the China Integrated Circuit Industry Investment Fund, better known as the Big Fund.
Created in 2014 to finance semiconductor self-sufficiency, the Big Fund has historically backed hardware names such as SMIC and YMTC. A move into DeepSeek would mark its first reported investment in a Chinese large-language-model company, according to media accounts citing the Financial Times.
That changes the meaning of the round. The Big Fund is not a conventional VC. It brings capital, but it also signals industrial alignment. If it enters DeepSeek’s cap table, the company’s future compute path will almost certainly be pulled closer to domestic chips from Huawei Ascend, Cambricon and Hygon.
- DeepSeek’s training infrastructure would become more directly linked to Chinese silicon road maps.
- The pairing of domestic models and domestic chips would move from an engineering preference to an industrial-policy priority.
- Government and state-owned enterprise deployment decisions would become easier to justify.
Tencent is also reported to be at the table. That mix would give the round two legs: state-backed industrial policy on one side and a private internet giant’s market reach on the other.
How a $45 billion valuation can be argued
On conventional SaaS or AI revenue multiples, DeepSeek is difficult to price. It has no public history of outside financing and has disclosed little commercial data. The reference point in this negotiation is not revenue. It is strategic scarcity.
| Company | Latest valuation | Profile |
|---|---|---|
| OpenAI | $852B | Closed model, broad commercialization |
| Anthropic | $380B | Closed model, enterprise market |
| DeepSeek | ~$45B | Open model, research-led |
| Mistral | ~$8B | European open model |
| Cohere | ~$6B | Enterprise API |
DeepSeek is valuable because it is the rare Chinese lab that can be discussed in the same benchmark conversation as Anthropic and OpenAI. Its V4-Pro work, reported around a 1.6 trillion-parameter MoE design with 49 billion active parameters and a one-million-token context window, pushed Chinese open models into a tier few expected. The added point is that the system can be made to run beyond Nvidia hardware.
Where the money would go
Reports differ on the size of the round. Some mention $300 million to $400 million, while others say terms remain unconfirmed. The intended uses are more consistent: compute, compensation and a deeper migration toward domestic chips.
Compute comes first. DeepSeek’s next model cycle is already expected after V4, and a trillion-parameter MoE system still requires heavy training infrastructure even with aggressive engineering efficiency. Expansion at the Ulanqab data-center site needs capital.
Talent is the second pressure point. Chinese AI researchers are being bid up by ByteDance, Alibaba and other large platforms. DeepSeek’s earlier academic-style compensation model is harder to maintain when top researchers can command much richer packages.
The third use is chip migration. Moving a training stack from CUDA toward Ascend and MindSpore is not a simple port. It is a major engineering rebuild, and the pace of that rebuild will shape DeepSeek’s model cadence over the next few years.
A different DeepSeek after the Big Fund
State-backed capital would not necessarily take control from Liang Wenfeng, but it would add a new structure around the company. DeepSeek’s decisions would have to account for domestic chip makers, government customers and state-backed industrial expectations.
The result could be more disclosure, more formal governance and a clearer strategic boundary around open models. What used to be a founder-led research culture funded by High-Flyer would become a company carrying a national-AI-infrastructure role. The cost of a $45 billion valuation may be larger than the headline number.
Sources: DeepSeek reportedly seeks first funding round at $45 billion valuation (TechNode); China's Chip Fund in Talks to Lead DeepSeek Funding, FT Says (Bloomberg); China's AI Fund May Lead $44B DeepSeek Funding Round (Winbuzzer); CocoLoop; DeepSeek Eyes Landmark $50 Billion Valuation as China's AI Race Intensifies (YourNews)