DeepSeek, the Chinese AI lab known for training frontier-class models at a fraction of the cost of its US rivals, is negotiating a new funding round that would value the company at roughly $71 billion, according to Bloomberg. The financing, about $1.5 billion, would land five weeks after investors priced DeepSeek near $50 billion in its first outside funding round. That is a valuation jump of more than 40 percent in barely a month, and it is happening specifically to set up an initial public offering.
TechCrunch, citing Bloomberg’s July 14 report, said DeepSeek is preparing for a public listing as soon as late 2026, though 2027 is the more likely window. A pre-IPO primary round like this one functions as a marker for bankers and future public investors: it tells the market what private capital believes the company is worth just before it opens its books to everyone else.
That distinguishes this raise from the $7 billion round DeepSeek closed in June, which introduced outside investors to the company’s cap table for the first time and set the $50 billion baseline. This new round is not about diversifying ownership. It is about pricing the company ahead of a listing, at a valuation investors have not yet tested in public markets.
The valuation math has some support in usage data. DeepSeek accounted for nearly 23 percent of the tokens processed in June by Vercel’s enterprise AI gateway, compared with Anthropic’s 32 percent share, per the gateway’s own reporting. Token share is not revenue, and DeepSeek has not disclosed subscription or API revenue figures to accompany the funding talks.
DeepSeek’s cloud service runs on chips made by Huawei Technologies, a detail TechCrunch highlighted as evidence the lab keeps closing the capability gap with US labs despite export controls on advanced semiconductors. Its investor base already includes Tencent and Beijing’s National Artificial Intelligence Industry Investment Fund, both of which have incentives to support a high valuation heading into an IPO regardless of what public markets ultimately pay.
DeepSeek did not respond to TechCrunch’s request for comment, and neither Bloomberg nor TechCrunch reported terms beyond the $1.5 billion target and $71 billion price. Until DeepSeek discloses revenue, margins, or an underwriter, the $71 billion figure remains a negotiating position rather than a market-tested price.
For operators evaluating open-weight alternatives to US frontier models, DeepSeek’s valuation trajectory is now a leading indicator of how public markets might price Chinese AI labs once the IPO window opens. Anyone benchmarking model costs against DeepSeek should watch whether the $71 billion figure holds once outside auditors and public investors get a look at the books.
TechCrunch, citing a Bloomberg report, first reported the talks on July 14, 2026.
