DeepSeek Funding Round: Valuation, Backers, IPO Plan

China’s most prominent AI lab is raising money at a pace that matches its Silicon Valley rivals. According to reporting this week from Bloomberg and Chinese financial outlet Caixin, DeepSeek is in talks with investors for a second external funding round just weeks after closing its first, with the new deal reportedly targeting a pre-money valuation as high as 500 billion yuan — roughly $74 billion. The Hangzhou-based company is said to be raising about 50 billion yuan, or $7.4 billion, and has begun early work toward an initial public offering on Shanghai’s STAR Market. The details remain fluid, and the reported figures vary by source, but the direction is unmistakable: DeepSeek is moving from research curiosity to one of the most valuable AI companies on the planet.

Two rounds in two months

DeepSeek closed its first-ever external funding round in mid-June 2026, a milestone for a company that had run for years on the balance sheet of its founder’s hedge fund. That round raised roughly $7.4 billion at a post-money valuation reported near $50 billion to $60 billion, and it drew a marquee list of Chinese backers. Tencent contributed about 10 billion yuan and battery giant Contemporary Amperex Technology (CATL) put in about 5 billion yuan, making the two the company’s largest outside shareholders. NetEase and JD.com also participated.

That first raise was deliberately limited to a small circle of strategic investors, and according to Caixin, it left a queue of others eager to buy in. The new round now under discussion is, in effect, a response to that unmet demand: a chance for institutions who missed the June deal to back what many in China regard as a national AI champion. Reports differ on the precise headline number — a recent Chinese regulatory filing implied a valuation closer to $52 billion, while Bloomberg’s sources describe the fresh round targeting up to $74 billion — but both point to a valuation that has roughly doubled in a matter of months.

Revenue catching up to the hype

DeepSeek burst into global view in early 2025 when it showed that a Chinese lab could train frontier-class models at a fraction of the compute budget Western competitors were spending. What has changed in 2026 is that the attention is now translating into revenue. The company’s annualized revenue has climbed to between $400 million and $500 million, according to figures cited in multiple reports, driven largely by paid API access to its models sold through cloud channels.

That is still a small number next to the AI labs it is often compared to. Anthropic’s confidential IPO filing put its annualized revenue near $47 billion, and OpenAI has said it reached roughly $25 billion in annualized revenue earlier this year. But DeepSeek’s cost structure is the story investors keep returning to: a lab that reached the frontier on comparatively lean spending, and that continues to release open-weight models rather than lock everything behind a paid API. Its willingness to ship weights openly has made its models a reference point for the broader argument that open models are closing the gap with closed frontier systems.

The founder at the center

Much of DeepSeek’s valuation story runs through one person. Founder Liang Wenfeng built the company using profits from his quantitative hedge fund, High-Flyer, and personal capital, and he retains roughly 78% of the equity along with tight control through a special governance structure. As the reported valuations have climbed, so has his estimated net worth — pegged in recent reports at around $36 billion, a figure that, if accurate, would place him above the founders of OpenAI and Anthropic as the wealthiest individual whose fortune is tied to building AI models.

That concentration cuts two ways. It gives DeepSeek an unusually stable ownership base and a founder who is unlikely to be pushed off course by outside investors. But it also means a listing will require navigating the disclosure and governance expectations of a public exchange around a company whose control sits overwhelmingly with one person.

The road to a STAR Market listing

DeepSeek has reportedly hired investment banks to prepare for an IPO on the STAR Market, Shanghai’s Nasdaq-style board for technology companies, with an internal goal of filing an application this year and completing the listing in 2026 or 2027. Choosing a domestic Chinese venue rather than a U.S. exchange is itself a signal. It keeps the company inside China’s regulatory perimeter, sidesteps the political friction that has dogged Chinese firms listing in New York, and taps a pool of domestic capital that has been hungry for exposure to a homegrown AI leader.

It also lands DeepSeek’s debut in the same window as a wave of U.S. AI listings. Both Anthropic and OpenAI have moved toward the public markets this year, and a DeepSeek IPO would give investors on both sides of the Pacific a way to price the AI boom directly. The company is also pushing beyond software: it has been reported to be working on its own AI inference chip to reduce its dependence on foreign hardware, a vertical-integration bet that mirrors moves by its larger rivals.

What it means

DeepSeek raising a second round within weeks of its first tells you more about investor appetite than about the company’s need for cash. A lab that reached the frontier on lean spending does not obviously require $7 billion twice in a summer; the raises are being pulled forward by demand, not pushed by burn. For DeepSeek, the benefit is a war chest heading into a capital-intensive phase — bigger training runs, custom silicon, and the compute contracts that a public listing will force it to justify.

The clear winners are the June backers. Tencent and CATL bought into a valuation near $50 billion weeks before reports of a round that could value the company around $74 billion, an implied markup that looks handsome on paper before a single share trades publicly. The strategic logic is just as important as the arithmetic: Tencent gets a preferred lane to a leading model maker, and CATL, a battery company, gets a foothold in the AI-and-energy nexus that is coming to define data-center economics.

The open questions are the ones every AI valuation now faces. Revenue of $400–500 million against a $74 billion target implies a multiple that only makes sense if growth stays near-vertical — and DeepSeek’s open-weight strategy, while excellent for adoption and mindshare, is not the most direct path to the recurring, high-margin API revenue that public investors reward. A STAR Market listing keeps the company on home turf, but it also means its financials, governance, and founder concentration will be tested against public-market scrutiny for the first time.

Watch three things next: whether the new round closes at the reported terms or a lower number as diligence proceeds; whether DeepSeek’s revenue run-rate keeps pace with the valuation as its next model generation ships; and whether the STAR Market filing actually materializes this year. If it does, DeepSeek would become the first frontier-class AI lab to test the public markets from inside China — a data point the entire industry, and every investor trying to value it, will be watching closely.