NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

DeepSeek's New Investors Took About 17% — and Chinese Media Say a Listing Is Being Prepared

A roughly RMB 100B (≈ US$15B / HK$117B) second round closed in November, with new backers taking about 17% — and Chinese financial media say a STAR Market listing is in preparation.

2026-11-20 · 1167 words · NeuroAI
DeepSeek's New Investors Took About 17% — and Chinese Media Say a Listing Is Being Prepared

For years the most consistent thing about DeepSeek was its refusal to play the capital game. Then the game came to it, in a size that made the usual vocabulary of startup fundraising look small. A company built on restraint is now being measured by people who are not restrained at all.

Business registration changes completed on November 17 showed a new investor vehicle holding roughly 17% of the company. Chinese financial media reported that the second external round raised somewhere around RMB 100 billion (≈ US$15B / HK$117B), with a post-money valuation near RMB 615 billion (≈ US$87B / HK$678B). The same reports say the company has begun preparing a listing on Shanghai's STAR Market (科创板).

Key takeaways

  • A second external round of roughly RMB 100 billion (≈ US$15B / HK$117B) was completed in mid-November, according to business filings and financial-media reports. Some accounts put the final figure slightly higher.
  • The new investors entered through a partnership holding about 17%, with 31 outside parties reported as limited partners and the founder retaining control of the vehicle.
  • Chinese financial media put the post-money valuation near RMB 615 billion (≈ US$87B / HK$678B). Add the earlier summer round and the two together come to about RMB 150 billion (≈ US$22B / HK$171B).
  • CITIC Securities is reported to be advising on a STAR Market listing, with a prospectus possibly filed by year-end and a listing targeted for 2027. No official timetable has been announced.
  • The round is a financing event, not a validation of the valuation. The hard question — what a frontier model company is worth when its revenue is still small relative to its price — remains open.

What the round actually looks like

The structure is more interesting than the headline number. Rather than placing 31 institutions directly on the shareholder register, the parties were gathered into a single holding platform, with the founder controlling it as general partner and the outsiders sitting as limited partners. That is a governance choice with an obvious purpose: keep the cap table legible and control concentrated before a public review begins.

The money itself arrived in two waves. A first external round closed in the summer at roughly RMB 50 billion (≈ US$7.5B / HK$58B). The second, launched in mid-year and reportedly upsized twice as demand exceeded the original target, closed in November at roughly twice that size. Telecom groups, securities firms and a long list of industrial investors appear in the reported roster — a mix that says as much about strategic positioning as it does about expected financial return.

There is a reason a model lab would accept that. Frontier training is a capital-intensive business: chips, power, data centres and the engineers who keep them useful. A round of this size buys time and compute, the two things a research organisation cannot improvise. It also buys the ability to say no to customers who want exclusivity, which matters when your technology is likely to be embedded in other people's products.

The listing question

The IPO reporting needs to be held at arm's length. Several Chinese financial outlets name CITIC Securities as the sponsor and describe a filing before the end of 2026 with a 2027 listing target. Other summaries of the same events say no timetable has been stated. Both can be true if the preparations are real but the calendar is informal.

What is verifiable is the direction of travel. A company that once said it was in no hurry to list has hired advisers, filled a finance chief role and consolidated its shareholder structure — three moves that point the same way. For readers who follow markets, the significance is not one company's debut. It is that a frontier AI lab may become a publicly priced asset, which would give the rest of the sector something it currently lacks: an externally set number to argue with.

That number will be argued about loudly. Private markets can accommodate a wide range of valuations because each investor negotiates their own entry. Public markets produce one price every day, and they tend to ask boring questions — revenue, margins, concentration of customers — that are awkward for a company whose main asset is a research trajectory.

The number that will be argued over

Chinese financial media have reported an annualised revenue run rate above US$1 billion, roughly double an earlier figure, driven partly by API price changes earlier in the year. If accurate, that puts the reported valuation at a high multiple of current revenue. It is not unusual for a frontier lab to be priced on future capability rather than present sales, but it does mean the valuation embeds assumptions about growth that cannot yet be tested.

The honest framing is that this round prices a bet, not a business in the ordinary sense. The bet is that general-purpose models become infrastructure, that the company which trained a leading open model can capture a lasting position, and that compute bought today will still be relevant when the next architecture arrives. Each part of that is plausible. None is demonstrated by a financing announcement.

Honest limitations

This article relies on business-registration reporting, Chinese financial-media coverage of the round and a handful of English-language summaries. I have no access to the filings themselves. The round size, 17% stake and post-money valuation are reported figures that differ slightly between outlets — some give RMB 105 billion (≈ US$15.7B / HK$122B) for the second round, and earlier-round valuations are quoted anywhere from RMB 350 to 400 billion (≈ US$52–60B / HK$406–467B). I have used rounded numbers and attributed them rather than presenting a single precise figure.

The IPO timetable is especially uncertain: a sponsor being named is not the same as a prospectus being accepted, and reported target dates frequently move. The revenue-run-rate figure comes from press reports, not audited financials, and annualised run rate is a snapshot rather than a year of recognised revenue. I have also not verified the identities of all 31 investors, and descriptions of their ownership stakes vary by source.

Finally, none of this changes what the model can do. Financing and listing news says a great deal about capital markets and very little about research quality.

What readers can do now

  • If you use DeepSeek's API, ignore the valuation and watch the pricing page. Financing may fuel capacity, but what reaches you is uptime, latency and cost per token.
  • If you follow markets, wait for the prospectus. A public filing contains audited numbers; a funding rumour does not.
  • If you are simply curious, treat the headline as a scale marker. The scale is a measure of how expensive the AI race has become, not a scoreboard of who is winning.

Money arriving in this quantity is a sign that the industry has entered a new phase. What it buys will be visible in models and prices, not in the announcement. The next useful number will come from a filing, not a funding round.

Related coverage

Cite this article

NeuroAI (脑机网), "DeepSeek's New Investors Took About 17% — and Chinese Media Say a Listing Is Being Prepared," 20 November 2026, https://neuroai.site/a/na-capital-deepseek-round-star-ipo

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Capital & Markets” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories