DeepSeek’s reported $7.4 billion fundraising would not be just another AI startup financing. It would mark a new phase in China’s AI race—one where private capital, industrial policy, cloud distribution, and even battery-sector money converge around a single Hangzhou lab that rattled global markets two years ago by claiming it could build competitive models on a fraction of the usual budget.
What has reportedly happened
According to people familiar with the matter cited by Reuters, DeepSeek is set to raise about 50 billion yuan, or roughly $7.4 billion, in its first funding round, with Tencent Holdings and battery maker CATL among the investors. It would be the company’s first external raise after years of running on capital from founder Liang Wenfeng and his quantitative hedge fund, High-Flyer. CNBC
The numbers, if the deal closes as described, are striking. The fundraising could value the company at between 350 billion and 400 billion yuan—roughly $52 billion to $59 billion—after the investment. The structure is unusual too: Liang Wenfeng has committed 20 billion yuan of his own money, Tencent is considering 10 billion yuan, and CATL is looking at 5 billion yuan, with the planned number of investors kept below ten. That makes the founder, not an outside fund, the single largest check. CNBCThe Manila Times
Beyond the headline trio, DeepSeek is also in final talks with China’s national AI fund—backed primarily by the China Integrated Circuit Industry Investment Fund—along with gaming developer NetEase and e-commerce giant JD.com. One caveat matters for readers: this is reported, not confirmed. Tencent and CATL declined to comment, while DeepSeek, Liang, NetEase, JD.com and the fund did not immediately respond to Reuters’ requests. No official company disclosure has accompanied the figures, and reporting notes that final terms could still shift. Treat the valuation as a market signal, not a settled fact. CNBCCNBC
Why investors are circling
DeepSeek’s pull is not hard to understand. The lab became China’s national AI champion in early 2025, when its V3 and R1 models drew widespread praise in Silicon Valley and challenged U.S. assumptions about Chinese AI capabilities. The R1 reasoning model in particular triggered a sharp repricing of AI economics. DeepSeek claimed it trained V3 on just $5.6 million worth of processors, and the subsequent selloff in U.S. tech stocks wiped out roughly $1 trillion in value at one point before markets recovered. The Manila TimesFortune
The momentum has continued. On April 24, 2026, the lab released a preview of its V4 frontier model, and the design choices behind it are central to why strategic investors are paying attention. DeepSeek launched two open-weight Mixture-of-Experts variants—V4-Pro at 1.6 trillion parameters with 49 billion active per token, and a leaner V4-Flash—available on Hugging Face, through the DeepSeek API, and on its consumer app. Pricing remains the company’s signature weapon: DeepSeek recently made a 75% price cut on V4-Pro permanent, lowering uncached input pricing from about $1.74 to $0.435 per million tokens and output from $3.48 to $0.87. aitoolinsightThe Tech Portal
The China angle: AI as industrial strategy
Here the story stops being a conventional venture deal. The most consequential detail in V4 is not its parameter count but its silicon. DeepSeek worked closely with Huawei so that V4 runs on Huawei’s Ascend AI processors, and Huawei announced its chips would offer full support for the models. This is the first frontier-class Chinese model engineered to train and serve without Nvidia in the loop—a direct response to a hardware ceiling imposed from Washington. Fortune
The investor lineup reinforces the industrial-policy reading. A national semiconductor fund, China’s largest internet platform, a leading game publisher, an e-commerce distributor, and the world’s dominant EV battery maker is not a typical cap table. CATL has recently pushed into AI data centers, exploring power equipment and energy-storage solutions as AI workloads drive demand for large-scale, reliable power—which is why a battery company appears alongside cloud and chip money. The mix, as Reuters frames it, underscores China’s effort to build an increasingly self-sufficient AI industry, from models to the energy infrastructure needed to power them. CNBCThe Manila Times
The compute bottleneck remains
Capital does not automatically solve DeepSeek’s hardest constraint. U.S. export controls still shape what Chinese labs can buy. In a rule effective January 15, 2026, the Bureau of Industry and Security revised its license review policy to allow Nvidia H200-class chips and less advanced semiconductors to be considered for export to China on a case-specific basis—a meaningful loosening from the blanket presumption of denial in place since 2022, but one wrapped in certification and third-party testing requirements. Federal Register
Even that opening is conditional and contested. Nvidia itself continues to flag the risk in its filings, warning that the U.S. has imposed broad unilateral controls on GPUs and that additional, more restrictive controls are likely, which could prohibit exports to entire markets and materially affect its business. For DeepSeek, the strategic logic is to design around the uncertainty rather than depend on it: optimize for domestic accelerators, squeeze efficiency, and price aggressively to win distribution. Whether Ascend supply can scale fast enough to train and serve trillion-parameter models at frontier quality is the open question that money alone cannot answer. sec
Analyst’s View
From a country-risk and credit perspective, the most important feature of this round is the blurring line between private financing and state strategy. A cap table that pairs Tencent and JD.com with a national semiconductor fund means DeepSeek’s fortunes are increasingly tied to Beijing’s self-reliance agenda. For an investor, that cuts both ways: it implies durable access to strategic capital, compute allocation, and domestic distribution, but it also raises the company’s profile as a target for foreign export-control authorities and data-governance scrutiny. Strategic backing and regulatory exposure tend to scale together.
For market positioning, a confirmed round near a $59 billion valuation would revive the question that unsettled equity markets in early 2025—whether high-performance models can be built and deployed far more cheaply than the prevailing U.S. capex assumptions imply. That matters less for DeepSeek’s own balance sheet than for the valuation premium attached to Western AI leaders and the multi-year cloud and semiconductor spending built on the premise that compute scarcity is a durable moat. A credible low-cost, domestically-supplied Chinese stack chips at that premise.
On operational and execution risk, the practitioner’s caution is straightforward: financing reduces runway risk, not supply risk. DeepSeek’s execution now hinges on Huawei’s ability to ramp Ascend production, on the maturity of China’s non-Nvidia software stack, and on partnerships with platforms that can provide distribution and power. The capital is necessary but not sufficient.
The round is best read as a test, not a verdict—of whether China can convert genuine AI engineering momentum into durable commercial and strategic scale while operating inside a hardware perimeter it does not fully control. The answer will not come from the size of the check, but from what gets shipped on Chinese silicon in the eighteen months that follow it.
