DeepSeek Was the Underdog That Rattled Silicon Valley. Now It's Raising $7.4 Billion.

On January 27, 2025, a Chinese AI lab most Americans had never heard of wiped $589 billion off Nvidia's market cap in a single trading session, the largest one-day market cap loss in stock market history. The cause was a model called R1, quietly released by DeepSeek, a research lab in Hangzhou, funded entirely by hedge fund manager Liang Wenfeng. R1 matched the performance of OpenAI's best reasoning model. Its reported training cost was much lower than what US labs spend. The weights were released for anyone to use for free.

The story spread quickly: China, working around US chip export controls with less powerful hardware, had reached the frontier. Silicon Valley had a problem.

Seventeen months later, DeepSeek is raising $7.4 billion, its first outside funding ever, at a valuation of up to $59 billion. The scrappy underdog chapter is officially over.

From Zero to $7.4 Billion in One Round

Until this year, DeepSeek operated more like an academic research lab than a startup. There was no venture capital or outside investors. Liang Wenfeng funded it entirely through High-Flyer Capital, his quantitative hedge fund, and ran it with a culture that valued research output over making money. The lab gave its models away for free under MIT licenses and didn't pursue revenue.

That identity is now changing. This round, one of the largest initial raises in Chinese AI history, includes Liang himself investing about $2.75 billion of his personal funds, Tencent contributing around $1.4 billion, battery giant CATL investing approximately $700 million, and JD.com, NetEase, and IDG Capital all in talks. There are less than ten investors total, according to Reuters.

The money will go toward computing infrastructure — DeepSeek reportedly hit chip capacity limits during peak demand — and talent acquisition. Liang has told investors that the primary goal remains AGI and that open-sourcing models will continue. Whether that commitment lasts under the pressure of a $59 billion valuation remains to be seen.

The Part That Changes the Calculus

Buried in the investor list is a detail that matters most: China's national AI fund — a state entity within the Beijing-backed system that also supports Chinese chipmakers — is in advanced talks to participate. If it closes, DeepSeek will formally become linked to the state. The "independent research lab" framing, which was always somewhat fictional given China's regulatory environment, will completely disappear.

This matters beyond symbolism. A DeepSeek backed by the government, with Tencent's distribution and CATL's industrial reach, is not the same as the lean lab that surprised everyone in January 2025. It will become a national AI champion, supported by the full weight of the Chinese tech-industrial complex, with a clear mandate to achieve AGI.

There is also a legal cloud over the lab's methods. In February 2026, Anthropic claimed that DeepSeek, along with two other Chinese AI firms, created more than 24,000 fake accounts and generated over 16 million interactions with Claude, scraping the outputs to train their own models in violation of Anthropic's terms of service. Anthropic framed this as a national security issue. DeepSeek has not publicly responded. The allegation has never been settled, but it casts a shadow over the lab's open-source narrative: the lab that gives its models away for free may have partially built them by taking someone else's work.

The US response to DeepSeek's rise has been to impose export controls, restricting access to advanced chips to slow down Chinese AI development. Those controls failed to prevent R1. The $7.4 billion round is partly a direct response to them — funding domestic computing infrastructure built around Huawei chips and whatever hardware China can obtain outside US restrictions.

The Numbers in Context

At $59 billion, DeepSeek is large but not yet in the same league as its US rivals. OpenAI has a current valuation of $852 billion. Anthropic, which filed confidentially for an IPO last week, was last valued at $965 billion. DeepSeek operates at roughly one-fifteenth their scale.

But comparing valuations misses the point. DeepSeek's influence on the AI industry has been larger than its size suggests. It forced a global reconsideration of whether cutting-edge AI requires massive investment, and that question remains open. The $725 billion in combined AI spending from Amazon, Microsoft, Google, and Meta in 2026 — an arms race detailed when Google raised $85 billion and its stock still fell — partly responds to the anxiety DeepSeek created.

What This Means for You

If you hold Nvidia, the January 2025 shock already changed how the market views AI chip demand. This round doesn't reverse that — it increases the uncertainty. A better-funded DeepSeek with its own computing infrastructure is less reliant on the hardware that drives Nvidia's revenue.

If you follow the broader AI landscape, the DeepSeek round closes one chapter and opens a more complicated one. The lab that proved efficiency could outpace scale is now raising money to grow. The champion of open-source is accepting government funds. The underdog is becoming part of the establishment. What it builds next — and whether it maintains its open nature — will reveal more about the future of AI than any valuation multiple.

The $589 billion shock was just the opening move. This is the response.