For most of its short life, DeepSeek has been the strangest kind of AI company: a frontier research lab that refused to take outside money. Founded by Chinese hedge-fund billionaire Liang Wenfeng, the Hangzhou-based startup shot to global attention in January 2025 with R1, an open-source large language model that matched US rivals like OpenAI while reportedly training on a small fraction of their computing power. Liang owns 89.5% of the company through personal and affiliated holdings, and until recently he saw no reason to sell any of it.
That is now changing — dramatically. According to reporting from the Financial Times, China's biggest state-backed semiconductor investment vehicle, the China Integrated Circuit Industry Investment Fund (universally known as the 'Big Fund'), is in talks to lead DeepSeek's first-ever fundraise at a valuation of roughly $45 billion. Tencent is also reportedly weighing a stake. The price tag has more than doubled from $20 billion just weeks earlier, even though the final investor lineup is not yet locked in.
The story behind the deal is more interesting than the headline number. Liang originally wanted to raise only a nominal sum — just enough to set a share price so DeepSeek could offer employees stock options. The motive was defensive: rivals were poaching his researchers with bumper pay packages, and stock options are how AI labs typically keep their best people. But once word got out that DeepSeek was opening its books, investor demand exploded, and the Big Fund — which had assembled $47 billion in its third funding round in 2024 from the finance ministry, local governments, and state-owned banks — moved to lead.
Here's the catch: the Big Fund was built to bankroll Chinese semiconductor companies like SMIC, the country's most advanced chip foundry. Its mandate is hardware — equipment and materials — and it has not publicly backed any of China's other LLM operators. So an investment in DeepSeek would be a clear departure, signalling that Beijing now sees frontier AI labs as part of the same strategic perimeter as chips. President Xi Jinping has launched the Big Fund across three phases as part of a broader self-sufficiency drive in the face of US efforts to restrict Chinese access to advanced chip-production equipment. State capital flowing into DeepSeek fits that pattern: if you can't import the best chips, fortify the labs that squeeze the most performance out of the chips you do have.
What makes DeepSeek a curious recipient is that, unlike Chinese peers Zhipu (market value $52 billion) and Moonshot, it has deliberately avoided commercialisation. It is not selling AI services to enterprises or chasing consumer chatbot revenue. Investors are nonetheless striving to back it despite — perhaps because of — that lack of commercial focus, betting that pure model capability will matter more than near-term earnings. Liang himself may invest personally in the new round.
The deal isn't closed, and three questions hang over it. Will Tencent and Alibaba ultimately participate at the new valuation? Will DeepSeek keep releasing its model weights openly once state money is inside the cap table? And how will the Big Fund's stake be structured — does it come with strategic governance rights? The answers will reveal how independent China's frontier AI sector will be allowed to remain. For now, the direction is clear: a lab that prized its independence is becoming a national champion, whether its founder fully intended it or not.
Source: https://www.ft.com/content/deepseek-45bn-valuation-big-fund-tencent
A Chinese AI lab that refused outside money for years is suddenly worth $45 billion — and the lead investor is the same state fund Beijing built to win the chip war.
DeepSeek, the Hangzhou-based AI lab that stunned the world last year with its R1 model, is in talks for its first-ever fundraise at a roughly $45 billion valuation. The lead investor is China's 'Big Fund' — formally the China Integrated Circuit Industry Investment Fund — Beijing's main state vehicle for backing the country's chip industry.
Tencent is reportedly weighing a stake too, and founder Liang Wenfeng, who controls 89.5% of DeepSeek through personal and affiliated holdings, may invest his own money. The valuation has more than doubled from $20 billion just weeks ago, and the final investor list isn't locked in yet.
This isn't a normal Silicon Valley funding round. It's closer to a government picking a national champion — the way countries once bankrolled flag-carrier airlines or shipyards.
If you use ChatGPT, Claude, or any AI tool, the question of who builds the foundation models — and under what flag — shapes everything from what those tools will say about Tiananmen Square to which country sets the safety rules. The US is restricting China's access to advanced chips; China is responding by financially fortifying labs that can squeeze more performance out of the chips it already has. Careers in AI research, semiconductor engineering, and tech policy are being shaped by exactly this kind of move.
This is the clearest sign yet that Beijing has chosen DeepSeek as a frontier-AI national champion, the way it once anointed Huawei in telecoms or SMIC in chip fabrication. Watch three things: whether DeepSeek keeps releasing open-weight models once state money is inside the cap table, whether Tencent and Alibaba stay in the round at the new price, and whether the Big Fund demands governance rights. Each answer tells you how independent China's AI sector will be allowed to remain.