A tiny Chinese textile company's regulatory filing just did what DeepSeek itself never has: put a public number on what the AI lab is worth.
You would not expect to learn DeepSeek's valuation from a company that mostly makes elastic webbing and hair ties. But that is exactly what happened this week. Reuters reported on July 16 that Anhui Korrun disclosed a fund tied to one of its subsidiaries had put 2.90 billion yuan into an indirect 0.8265% stake in DeepSeek. Do the math and the Hangzhou AI lab comes out valued at 350.88 billion yuan, or about $51.82 billion. That's now the hard number.
It matters because DeepSeek has spent the past year being loud in the one place that counts, the model market, and almost silent everywhere else. The company has not announced the round. It has not named a lead investor. It has not given the public a clean cap table. Founder Liang Wenfeng built DeepSeek out of High-Flyer, his quant hedge fund, and the company's public posture has mostly been: ship the model, say little, let everyone else argue.
That silence is why the paperwork carries weight. A separate filing from Jiuan Medical showed its Hong Kong unit put 750 million yuan into a Shixiang Capital vehicle for roughly a 0.21% indirect stake in DeepSeek, another data point that points to the same rough valuation. Anhui Korrun said investors paid into its vehicle by June 17 and that it was told the investment closed on July 15. Small, unrelated listed companies are now doing the disclosure work DeepSeek will not do itself.
The valuation is already moving #
The $52 billion figure is not sitting still. Bloomberg reported this week that DeepSeek has begun preparing for a mainland China IPO filing that could come as soon as this year, with a possible market debut in 2027. The Wall Street Journal also reported that the company is preparing for a Shanghai listing and is talking to investors and banks. For a company that had never taken outside money until recently, that is a fast turn.
DeepSeek is also weighing another private round. The Financial Times reported that the company is discussing fresh funding at a pre-money valuation of about $71 billion, just weeks after closing its first external financing. Bloomberg put the same target at at least 480 billion yuan. If that price holds, DeepSeek will have moved from a roughly $52 billion post-money marker to a $71 billion private-market target in a matter of weeks. Frankly, that is a lot of belief to place on a company that still communicates like a research lab with the lights off.
The market is not irrational to care. DeepSeek's R1 reasoning model, released on January 20, 2025, turned the company from a Chinese AI name into a global one. The company said at the time that the model was trained using about 2,048 Nvidia H800 chips at a claimed cost of $5.6 million, a figure that hit Silicon Valley exactly where it was vulnerable: compute spend. R1 briefly became the most downloaded free app on Apple's US App Store and helped trigger a brutal selloff in US tech stocks, including Nvidia. One model changed the conversation.
China's AI market is not waiting #
DeepSeek is not raising into an empty field. Moonshot AI, the Beijing company behind Kimi, has just released Kimi K3, which Xinhua said has 2.8 trillion parameters and a 1 million token context window. The Financial Times reported that Moonshot is positioning K3 against Anthropic's Opus 4.8 and preparing a launch that pushes China further into open-weight frontier models. That is the pressure on DeepSeek. You cannot live forever on the shock of R1 while rivals keep shipping.
Still, DeepSeek's valuation tells you something specific about where Chinese capital is moving. Investors are not paying only for a chatbot. They are paying for a company Beijing can point to as proof that Chinese labs can compete despite US chip controls, and for a founder who has made open models look commercially dangerous to closed US rivals. According to the Financial Times, DeepSeek's new fundraising push is tied partly to the capital needed for computing capacity and infrastructure. That is the real bill underneath the valuation.
This is a strange way to price one of China's most important AI companies. No glossy funding announcement. No triumphant founder interview. No clean investor deck for the public. Instead, the number arrived through securities filings from companies better known for medical devices and elastic webbing than frontier models. But you work with the facts you have. Right now, the facts say DeepSeek is worth about $52 billion on paper, may soon ask investors to value it near $71 billion, and is already preparing for the public markets before it has learned to speak in public.
Also read: Moonshot's Kimi K3 Becomes the Largest Open-Weight Model Ever Built • Meta puts Muse Spark 1.1 on OpenRouter, then locks out the rest of the world • A Former DeepMind Researcher Just Raised $55 Million Before Building Anything