For years, Liang Wenfeng funded DeepSeek out of his own pocket through High-Flyer, his quantitative hedge fund. He told investors to stay away. He ran it like a research lab, not a startup. That era just ended.

On June 3, Reuters reported DeepSeek is raising about 50 billion yuan ($7.4 billion) in its first outside funding round, with investors valuing the company at up to $59 billion. The round includes Tencent, CATL, NetEase, JD.com, and China's national AI fund, with fewer than 10 investors total.

Why $59 Billion Makes Sense and Also Sounds Insane

DeepSeek's valuation seems wild until you look at the math. OpenAI was valued at $300 billion earlier this year. Anthropic just closed $50 billion at what sources say was a valuation north of $100 billion. Chinese AI labs have been systematically undervalued compared to their American counterparts. DeepSeek built models that compete with GPT-4 and Claude on benchmarks while spending a fraction of what US labs burn on compute. That efficiency is the product, not just the models.

But here's what bugs me: DeepSeek raised this money from fewer than 10 investors. That's unusual for a round this size. Most AI mega-rounds spread risk across 20 or more participants. A concentrated cap table means DeepSeek is cherry-picking strategic partners, not just cash. Tencent wants to compete with Alibaba's Qwen. CATL makes batteries for every EV in China. JD.com runs logistics. These aren't passive financial investors. They're building the supply chain around DeepSeek's models.

The Numbers That Matter

Metric Detail
Round size ~50 billion yuan ($7.4 billion)
Valuation Up to $59 billion
Round type First outside funding round
Investors Tencent, CATL, NetEase, JD.com, national AI fund
Investor count Fewer than 10
Prior funding Self-funded via High-Flyer quant fund
Known models DeepSeek-V4, DeepSeek-R1, DeepSeek-Coder

The $7.4 billion is roughly what Anthropic raised in a single tranche of its Series E. For a Chinese lab with a lean team and proven efficiency, this is enough capital to buy serious compute without the cash-burning trajectory that concerns US investors.

The Tencent Signal Is the Real Story

Tencent investing in DeepSeek is the headline most people will skip, but it matters more than the valuation. Alibaba has poured resources into Qwen and made it the default Chinese open-source model. Tencent, until now, has been the quiet one in the AI race. Investing in DeepSeek signals they've decided building in-house isn't fast enough. They'd rather back the best independent lab and integrate its models into WeChat, gaming, and cloud services.

This is the same play Microsoft made with OpenAI. Tencent gets distribution advantage. DeepSeek gets resources without losing independence. The question is whether Liang Wenfeng can maintain the research-lab culture that made DeepSeek efficient in the first place once Tencent's money comes with strings.

What This Means for the Open-Source AI Race

DeepSeek's models have been the most consequential open-weight releases of the past year. DeepSeek-R1 proved you could build a reasoning model that matched OpenAI's o1 on math benchmarks. DeepSeek-V4 delivered frontier-level performance at a fraction of the compute cost. Every open-source AI lab from Mistral to Meta has had to respond to what DeepSeek shipped.

Now those models have $7.4 billion behind them. The immediate question is whether DeepSeek uses the money to scale training runs or to expand into products. Liang Wenfeng has historically been allergic to product thinking. He built models and let the community figure out what to do with them. Tencent and JD.com will want more than that. They'll want integration, APIs, and enterprise offerings that compete directly with Alibaba's cloud AI stack.

What Surprised Me

The timing. DeepSeek chose to raise now, right after Anthropic's massive round and during a period when Chinese AI companies are under intense regulatory scrutiny. Liang Wenfeng reportedly is directly involved in the negotiations, which suggests he's not handing this off to a CFO. He wants control over who gets in.

The bigger picture: this round effectively ends the myth that DeepSeek is just a scrappy side project. $59 billion puts them in the same conversation as OpenAI and Anthropic. The difference is they got there without spending billions on marketing or hiring thousands of employees. They built efficient models, let the community validate them, and now the money is coming to them.

That's either the most disciplined AI company in the world, or the last independent Chinese AI lab that will ever exist. Probably both.

Sources

  • Reuters: DeepSeek slated to raise $7 billion in maiden fundraising round (June 3, 2026)
  • Yahoo Finance: DeepSeek could be valued at up to $50 billion in first fundraising
  • The Information: China's DeepSeek is Raising Money for First Time
  • WSJ: China to Invest in DeepSeek at $50 Billion Valuation