DeepSeek Nears $7.4 Billion Close at $59 Billion Valuation in Historic First Funding Round

When DeepSeek’s V3 model landed in late 2024 and erased roughly $600 billion in market value from Nvidia in a single trading session, the world got its first look at a Chinese AI lab that could go toe-to-toe with OpenAI on a fraction of the compute budget. Eighteen months later, that same lab — which spent its early years operating entirely on founder capital, with zero outside investors — is closing what sources describe as one of the largest AI funding rounds ever assembled in China.

According to people familiar with the deal, DeepSeek is finalizing a raise of approximately 50 billion yuan (US$7.4 billion) at a valuation approaching $60 billion, completing a process it kicked off in the spring after years of deliberately rejecting external capital. The round is expected to close within weeks.

From Zero to $60 Billion in Eight Months

The speed of DeepSeek’s valuation ascent is staggering. In April, the lab was reportedly valued at $10 billion in informal secondary discussions. By early May, Bloomberg and the Financial Times were reporting that China’s state-backed “Big Fund” was in talks to lead a $3–4 billion round at a $45 billion valuation. Now, sources close to the round put the figure at just under $60 billion — a six-fold increase in the span of eight weeks.

The upward revision reflects how the geopolitical stakes have changed. As U.S. semiconductor export controls tighten further under the current administration, Chinese tech giants and state funds have converged on DeepSeek as the country’s most credible path to frontier AI independence. A strong valuation is not merely an ego metric — it signals strategic importance to Beijing.

The Investor Lineup

Tencent Holdings, China’s social media and gaming behemoth, is committing approximately 10 billion yuan to the round, making it the largest institutional backer. CATL, the world’s dominant electric vehicle battery manufacturer, is investing 5 billion yuan — a striking bet that reflects the company’s ambition to embed AI deeply into its energy storage and manufacturing operations. NetEase and JD.com are each contributing roughly 3 billion yuan.

Rounding out the cap table are several venture capital and growth equity firms including IDG Capital, Monolith Management, Loyal Valley Capital, and Shixiang Tech. China’s National Artificial Intelligence Industry Investment Fund — a government vehicle tasked with backing strategic technology sectors — is also expected to participate.

Perhaps the most telling commitment is Liang Wenfeng’s own. DeepSeek’s founder and CEO is personally investing approximately 20 billion yuan, or roughly $2.9 billion, constituting about 40% of the total round. For a founder who built his fortune running High-Flyer Quant, one of China’s most successful quantitative hedge funds, putting that kind of personal capital into an AI lab is not a diversification move — it is a concentrated bet that DeepSeek is playing for something much larger than a near-term exit.

A Strategic Pivot With Conditions

The willingness to raise external capital at all represents a meaningful evolution for a lab that prided itself on independence. DeepSeek’s management has reportedly told potential investors that the proceeds will go primarily toward foundational AI research — not toward product monetization, cloud infrastructure expansion for commercial customers, or near-term revenue growth. Investors entering the round appear to accept these terms, treating their stakes as long-duration strategic assets rather than conventional startup equity.

That framing matters because DeepSeek’s current competitive position is more complicated than its early 2025 moment of global shock might suggest. Its V4 model, released earlier this year, was widely praised as a major open-source advance, but evaluations by independent researchers suggest it trails the leading frontier models from OpenAI, Google DeepMind, and Anthropic on the most demanding tasks. Closing that gap requires sustained, capital-intensive research — exactly what this round is designed to fund.

The Hardware Constraint Thesis

DeepSeek’s funding story cannot be read without understanding the underlying semiconductor calculus. U.S. export controls have made it difficult — though not impossible — for Chinese AI labs to acquire the most advanced Nvidia chips (the H100, H200, and successor Blackwell series). DeepSeek’s landmark achievement with V3 and R1 was demonstrating that with algorithmic ingenuity — particularly around mixture-of-experts architectures and aggressive quantization — a Chinese lab could produce world-class results from older, lower-tier hardware.

That efficiency thesis is now an investment thesis. Backers are betting that DeepSeek can continue to close the hardware gap through software, even as the U.S. tightens export policy. It is, at its core, a bet on Chinese AI exceptionalism under constraint.

The company has also reportedly deepened its relationship with Huawei’s Ascend AI chip ecosystem. Earlier this year, reports emerged that DeepSeek’s training runs had incorporated Huawei hardware at scale — a significant endorsement of the domestic silicon alternative that Beijing has spent hundreds of billions nurturing.

Open Source as Strategic Moat

One underappreciated aspect of DeepSeek’s strategy is its commitment to releasing model weights publicly. While OpenAI and Anthropic keep their frontier models fully proprietary, and Google hedges with selective partial releases, DeepSeek has consistently made its flagship models available to download. This has driven extraordinary adoption: the R1 model has been integrated into dozens of third-party platforms and developer tools globally.

Open sourcing creates a distribution moat that is hard to replicate. Each integration, fine-tune, and derivative product builds the DeepSeek ecosystem and makes future, more capable releases more valuable by default. For a company that does not (yet) charge aggressively for API access, community adoption is its primary form of market penetration.

The new investor base — which includes Chinese tech platforms like NetEase and JD.com — will likely benefit from preferential access to DeepSeek’s next-generation models for their own products, creating aligned incentives without requiring the startup to abandon its research-first posture.

Implications for the Global AI Race

DeepSeek’s fundraise lands against a backdrop of extraordinary capital formation in AI globally. OpenAI’s IPO S-1 is in preparation. Anthropic completed its own major funding rounds earlier this year. Big tech companies have collectively committed over $700 billion to AI infrastructure capex through 2027.

What makes DeepSeek different is trajectory. It is not yet in the same tier as OpenAI or Google DeepMind on pure capability metrics. But it is closing the gap faster than most Western analysts expected, with a fraction of the capital, under hardware constraints that would have been considered crippling two years ago. A fully capitalized DeepSeek — with $7.4 billion to deploy on research, talent, and compute — is a materially different competitive threat than the scrappy quant fund spinout that shocked Davos in January 2025.

For the global AI ecosystem, the round confirms what was already becoming clear: the frontier AI race is no longer a two-country competition. It is a multi-front, geopolitically entangled contest, and DeepSeek has just secured its seat at the table.