Chinese AI lab DeepSeek has gone from quiet outlier to geopolitical lightning rod in a matter of weeks, as Beijing moves to anoint it as a national AI champion at a valuation approaching $45 billion. The money is big, but the message from China’s state is even bigger.
DeepSeek’s rise didn’t begin with term sheets; it began with a technical shock to the system. In early 2025, the lab vaulted into prominence with a large language model that matched or closely tracked the capabilities of leading U.S. systems from OpenAI and Anthropic — but trained “on a fraction of the compute power and at a fraction of the cost” of those American rivals.
Crucially, the model wasn’t just a research stunt. DeepSeek has “kept reasonable pace with the top models in the world in areas like reasoning and coding while remaining open weight,” with versions “freely available on Hugging Face.” In a landscape increasingly defined by closed, tightly controlled models, that open-weight stance helped the lab build a global following among engineers and researchers.
Behind the scenes, however, DeepSeek was anything but a typical startup. Founded in July 2023 by Chinese hedge fund billionaire Liang Wenfeng, the company spent “its entire existence refusing outside money,” staying internally funded and tightly controlled — Liang reportedly holds nearly 90% of the equity.
For a while, that worked. But as DeepSeek’s models improved, so did the pressure. Rival labs and tech giants began poaching its researchers, and the lack of equity upside became a competitive liability. According to reporting cited by TechCrunch, “faced with competitors poaching DeepSeek’s researchers, Liang opted to raise funds in order to offer employees shares in the company.”
The original plan was relatively conventional by mega-unicorn standards. In mid-April 2026, DeepSeek was reportedly exploring a roughly $300 million raise at a valuation of about $10 billion, with Chinese cloud titans Alibaba and Tencent in the mix as potential investors.
That was already ambitious for a lab that had never taken external capital. But the real shock came in the speed and scale of what happened next.
By early May, the prospective valuation had more than quadrupled. Bloomberg and the Financial Times reported that DeepSeek “is in talks to raise its first round of venture capital, and in just a few weeks, its potential valuation has soared from $20 billion to $45 billion.” What began as a classic big-tech–backed growth round had mutated into something very different.
The key change: who would lead the deal.
On May 6, reports surfaced that China’s main state-backed semiconductor investment vehicle — the China Integrated Circuit Industry Investment Fund, better known as the “Big Fund” — was now in talks to lead DeepSeek’s first external funding round at that eye-watering ~$45 billion valuation.
This isn’t just any state investor. Since 2014, the Big Fund has been the central firehose of Chinese state capital for chips, having “deployed more than $50bn into Chinese chip-design, fabrication, packaging, and equipment companies.” Until now, its mandate was “overwhelmingly focused on the silicon side of the AI stack” — fabs, foundries, memory, and EDA tooling, not software or model labs.
A direct, lead investment in a frontier AI model company like DeepSeek would be “by some distance, the largest extension of the fund’s mandate to date.” As The Next Web frames it, the deal is not just about valuation; “DeepSeek's $45bn valuation is also Beijing's strategic statement.”
The strategic context is straightforward and stark. The U.S. has used semiconductor export controls “to limit China’s access to leading-edge AI compute,” and Beijing has been searching for “a response strategy that recovers parity.” The Big Fund moving into AI models signals that China’s leadership has reached a conclusion: the path to AI parity no longer runs only through chips.
As TNW puts it, “The Big Fund leading a DeepSeek round is, in that frame, the recognition that the response strategy now runs through model capability rather than purely through chip capability. If China cannot acquire Nvidia’s leading-edge GPUs at the volume required, it will, on this evidence, finance the model labs that have demonstrated they can produce frontier results without them.”
DeepSeek isn’t just any model shop. It has been “optimized to run on chips made by China’s hardware giant Huawei Technologies,” a company already at the center of U.S.-China tech tensions.
That combination — state-backed capital, Huawei silicon, and a frontier-scale, cost-efficient model lab — “is considered a powerful duo for the nation to develop its own AI to rival the United States.” In other words, this is about more than building another ChatGPT competitor. It’s about constructing a domestically controlled AI stack, from hardware to models, that cannot be throttled by Washington.
China’s private cloud giants are also circling. Tencent and Alibaba “are reportedly in talks to participate” in the round, according to Bloomberg reporting cited by TechCrunch, turning DeepSeek into a potential nexus where state, platform capital, and national security priorities converge.
From a venture-capital perspective, the velocity of DeepSeek’s revaluation is extreme. TNW notes that “there is, in venture-capital pricing, a particular kind of ascent that does not normally happen to a company that has spent its entire existence refusing outside money.” Going from an internal, no-VC outlier to a state-backed juggernaut in one leap breaks the usual startup narrative arc.
From Beijing’s perspective, that is precisely the point. The DeepSeek deal is being read as a pivot in strategy: away from a chips-only obsession toward an AI-stack–wide approach where model capability is a strategic asset on par with lithography tools and fabrication plants.
If the round closes as reported, DeepSeek will move “from a self-funded outlier to a state-backed national champion.” And the Big Fund’s move into model labs could open the door for similar bets on other Chinese AI outfits that can show they can squeeze more capability from constrained compute.
Different actors will spin this saga in different ways.
Beijing’s narrative: DeepSeek is proof that China can innovate around U.S. sanctions. It can build cutting-edge AI models on domestic chips and do so more efficiently than American labs. Backing DeepSeek at $45 billion is about securing a flagship to rally the ecosystem and signal that model capability is now a core national priority.
Western policy view: Expect the DeepSeek-Huawei-Big Fund triangle to be read as confirmation that export controls are biting — and that China is now redirecting state resources to the model layer to compensate. For Washington, this strengthens the argument that AI and chips are inseparable in the security calculus.
VC and tech industry perspective: On one hand, DeepSeek’s meteoric valuation will be seen as yet another sign that AI remains the only game in town. On the other, the scale and speed of state intervention may reinforce the idea that frontier AI in China is no longer a purely market-driven race but a state-coordinated industrial campaign.
Researcher and open-source community view: DeepSeek’s open-weight stance has made it a darling among developers, but state backing could complicate that image. Will a lab now positioned as a national champion, running on Huawei hardware and financed by the Big Fund, keep releasing powerful models as freely as before? That tension between openness and national strategy will be closely watched.
Negotiations are ongoing, and DeepSeek “could not be immediately reached for comment,” leaving some details — including exact round size, structure, and participating private investors — still in flux.
But even before the money lands, the outlines are clear. A lab that began as a hedge-fund-funded curiosity has, in less than three years, become the centerpiece of Beijing’s next phase in the AI race: a bet that if you can’t buy the world’s best chips, you build the world’s most efficient models on the chips you have.
If DeepSeek’s $45 billion crown is fastened on, it won’t just be a victory lap for one lab. It will be Beijing’s way of rewriting the rules of AI power — with model capability, not silicon alone, at the top of the stack.
