DeepSeek, the Chinese AI laboratory that sent shockwaves through Silicon Valley in early 2025 by training world-class language models on a shoestring budget, is now reportedly raising its first round of venture capital — and the valuation has already doubled from initial estimates.
According to the Financial Times and Bloomberg, the Hangzhou-based lab is in advanced talks to raise capital at a valuation that has surged from roughly $20 billion to as high as $45 billion in just a few weeks. If the round closes at that figure, it would make DeepSeek one of the most valuable private AI companies on the planet — and the first Chinese AI lab to crack the top tier of global valuations alongside OpenAI, Anthropic, and xAI.
From Hedge Fund Side Project to National Strategic Asset
The story of DeepSeek is almost too improbable to be true. Founded by Liang Wenfeng, a Chinese hedge fund billionaire who made his fortune running quantitative trading strategies at High-Flyer Quant, DeepSeek began as an internal research project. Wenfeng controls nearly 90% of the company and, until now, had never sought outside investment. The lab was entirely self-funded, operating more like a research institute than a startup.
That independence is precisely what made DeepSeek dangerous — in the competitive sense — to American AI giants. In January 2025, the lab released DeepSeek-V3, a large language model that matched the performance of OpenAI's GPT-4 and Anthropic's Claude on key benchmarks. The jaw-dropping part? It was trained for approximately $5.6 million, a fraction of the hundreds of millions spent by US labs. The model was also open-weight, meaning anyone could download and run it locally, a sharp contrast to the closed API-only approach of most American frontier models.
The release triggered a panic in Silicon Valley. NVIDIA's stock dropped 17% in a single day as investors questioned whether America's chip dominance was as unassailable as previously believed. The episode became known as the "DeepSeek shock," and it permanently altered the strategic calculus of the global AI race. If a Chinese lab could build frontier models with consumer-grade chips and a tiny budget, what would they accomplish with real capital and state backing?
Now, with this funding round, we are about to find out.
Why DeepSeek Is Raising Capital Now
According to sources familiar with the matter who spoke to the Financial Times, Wenfeng's decision to bring in outside investors was driven by a single, urgent problem: talent retention.
DeepSeek's researchers have become some of the most coveted AI talent in the world. The lab's open-weight releases, published papers, and publicly available model weights have served as a continuous demonstration of its team's capabilities. That visibility has made DeepSeek employees prime targets for poaching by both Chinese tech giants and international labs willing to pay premium salaries.
By raising a venture round, Wenfeng can offer employees equity in a company that may soon be worth $45 billion. That potential upside is a powerful retention tool — and arguably necessary, given that DeepSeek's compensation has historically been below market rates for top AI researchers. The lab has prided itself on a lean, research-first culture, but lean cultures struggle to compete when competitors are throwing around eight-figure stock packages.
The timing is also geopolitically significant. China is in the middle of an aggressive push to develop homegrown AI capabilities that can operate independently of American technology. The country faces increasingly strict US export controls on advanced semiconductors, particularly NVIDIA's H100 and Blackwell GPUs. DeepSeek's ability to train frontier models on less advanced hardware has made it a critical piece of China's technological sovereignty strategy.
The Investors: State-Backed, Strategic, and Very Chinese
The funding round is reportedly being led by the China Integrated Circuit Industry Investment Fund, commonly known as the "Big Fund." This is a state-backed investment vehicle created by the Chinese government to support the domestic semiconductor industry. Its involvement signals that Beijing views DeepSeek not merely as a promising startup, but as a strategic national asset.
The Big Fund has historically invested in chip design firms, foundries, and equipment manufacturers. Investing in an AI lab represents a new direction — one that reflects the Chinese government's understanding that AI model capabilities and semiconductor independence are two sides of the same coin. You cannot have one without the other, and DeepSeek's work on optimizing training for Huawei's Ascend chips makes it a direct contributor to both goals.
Huawei Technologies, China's hardware giant and the country's primary domestic chip manufacturer, has developed the Ascend series of AI accelerators as an alternative to NVIDIA's GPUs. DeepSeek's models have been explicitly optimized to run efficiently on Ascend hardware, and the combination of DeepSeek's algorithms with Huawei's chips is viewed in Beijing as a powerful duo capable of rivaling the NVIDIA-OpenAI alliance.
Also reportedly in talks to participate are Tencent and Alibaba, China's two largest cloud computing providers. Their involvement would serve multiple purposes: securing preferred access to DeepSeek's models for their own cloud platforms, gaining insight into the lab's roadmap, and ensuring that DeepSeek's technology remains integrated into China's domestic tech ecosystem rather than drifting toward Western partnerships.
None of the companies involved — DeepSeek, the Big Fund, Tencent, or Alibaba — have commented publicly on the fundraising.
What $45 Billion Means in the Global AI Hierarchy
A $45 billion valuation would place DeepSeek in rare company. For context, OpenAI was valued at approximately $157 billion in its last funding round in late 2025. Anthropic raised capital at around $40-50 billion. xAI, Elon Musk's lab, has been valued at roughly $50 billion. Cohere and Mistral, the leading European labs, are valued in the single-digit billions.
DeepSeek at $45 billion would effectively tie it with Anthropic and xAI, creating a three-way race for second place behind OpenAI. The valuation would also validate a fundamentally different business model: DeepSeek has released most of its models as open weights, meaning it does not charge API fees or operate a subscription chatbot service. Its revenue, if any, comes from enterprise licensing, custom model development, and cloud partnerships — not from the direct-to-consumer model that powers OpenAI's valuation.
This raises an important question: what are investors actually buying? A $45 billion valuation for a company that gives away its primary product for free suggests that the investors are not expecting traditional software margins. They are betting on geopolitical value, on DeepSeek's role as a national champion, and on the possibility that the lab will eventually monetize through enterprise channels or government contracts.
It also suggests that the investors are not primarily motivated by financial returns in the conventional sense. A state-backed fund investing in a strategic technology company is more similar to defense procurement than venture capital. The goal is capability, not cash flow. That distinction matters when trying to understand whether DeepSeek's valuation is a market price or a policy decision.
The Open-Weight Question
DeepSeek's commitment to open-weight releases has been both its greatest strength and its most confounding strategic choice. On one hand, open weights have made DeepSeek the most popular AI lab among developers, researchers, and privacy-conscious users who do not want to send their data to American cloud providers. The lab's models are among the most downloaded on Hugging Face, and they have been integrated into countless open-source projects and local deployment tools.
On the other hand, giving away your model weights makes them difficult to monetize. OpenAI and Anthropic generate billions in revenue by gating access to their most capable models behind API keys and subscription tiers. DeepSeek's open-weight models can be downloaded once and run indefinitely without generating any revenue for the lab.
The funding round may signal a shift in this strategy. With investors to answer to, DeepSeek will face pressure to demonstrate a path to revenue. That could mean more closed-weight releases, premium tiers for enterprise users, or a pivot toward hardware and cloud services rather than pure model research. Alternatively, the state-backed nature of the investors may mean that profitability is not the primary metric — and that DeepSeek can continue its open-weight strategy as a form of technological soft power.
Either way, the next few months will reveal whether DeepSeek's open-weight philosophy is compatible with a $45 billion valuation. If the lab can find a way to monetize its influence without abandoning its developer-friendly approach, it may create a template for how AI companies can succeed outside the American closed-weight paradigm.
Implications for the US-China AI Race
The DeepSeek funding round lands at a pivotal moment in the US-China technology competition. The Trump administration has expanded export controls on advanced AI chips to China, added Chinese AI labs to investment restriction lists, and pressured allies to limit Chinese access to semiconductor technology. The goal is to slow China's AI progress by choking off its supply of cutting-edge hardware.
DeepSeek's entire existence is a rebuke to that strategy. The lab has demonstrated that algorithmic innovation can partially compensate for hardware disadvantage. Its models run on Huawei chips that are technologically inferior to NVIDIA's latest offerings, yet they achieve competitive performance through efficient architecture, optimized training recipes, and aggressive quantization.
A $45 billion valuation — backed by state money — would provide DeepSeek with the resources to push this advantage further. More researchers. Better compute infrastructure. Longer training runs. The gap between Chinese and American AI capabilities, which many analysts believed was widening in America's favor, may start to narrow again.
For US policymakers, the response options are limited. Sanctioning DeepSeek directly would be symbolically meaningful but practically ineffective — the lab does not rely on American technology, capital, or markets. Expanding controls on Huawei would slow chip development but would not address the algorithmic innovation that allows DeepSeek to do more with less. The most effective response would be for American labs to redouble their own efficiency research, but that requires a level of coordination between government, academia, and industry that the US has struggled to achieve.
What Happens Next
If the round closes at $45 billion, DeepSeek will instantly become the most closely watched AI company outside the United States. The lab will face scrutiny from investors, regulators, and competitors that it has never experienced before. Its research will be parsed for strategic implications. Its hiring will be tracked by intelligence agencies. Its model releases will move markets.
For the broader AI industry, DeepSeek's success — or failure — will serve as a test case for whether the open-weight model can survive at scale. If a $45 billion company can give away its weights and still thrive, it may shift the industry's center of gravity away from the closed API model that has dominated since GPT-3. If it cannot, the experiment will be remembered as a noble but economically unsustainable deviation.
Either way, DeepSeek has already changed the game. The question now is whether it can afford to keep playing — and at $45 billion, the answer appears to be yes.