11 SEP 2026 — DeepSeek has retained CITIC Securities and three other underwriters for a listing on Shanghai's STAR Market. That is a bank mandate, not a filing. There is no prospectus, no price, no raise amount, and the reports do not agree on whether the listing lands this year or next.
What is actually confirmed
The Hangzhou company has engaged four underwriters, CITIC Securities among them, to run the pre-listing process for a STAR Market float. Some reporting says the process begins this year; the Wall Street Journal has separately reported a 2027 listing.
A concurrent pre-IPO financing round would value DeepSeek at around 500 billion yuan, roughly 74.5 billion dollars, before the new money goes in.
For scale, the company raised 7.4 billion dollars in June at a valuation above 50 billion. A move to about 74.5 billion is a rise of roughly half in three months.
Hiring banks is the earliest public step
An underwriter mandate is an early, formal step. It signals intent and the start of preparation, and the company is now paying people to do that work. It does not commit the company to a date, a price or a size, and mandates are quietly unwound often enough that the step is routinely over-read.
The STAR Market adds its own timeline. Listings there run through a registration-based review with the Shanghai exchange and the securities regulator, and an AI company at this valuation will not move through it quickly.
The two reported timings are not in conflict; they describe different stages. Beginning the process this year and listing in 2027 are compatible statements, and only one of them is an event a reader can put in a calendar.
Why a domestic listing at all
DeepSeek could not list in New York on any realistic reading of current United States rules, so Shanghai is less a choice than the available option — and the STAR Market has become the venue where China's AI champions are expected to raise.
That has consequences for what the listing will disclose. A STAR Market prospectus will be written for Chinese regulators and domestic institutions, and the questions a Western analyst would ask about training compute, chip sourcing and export-control exposure are not the questions it is built to answer.
The prospectus, the one document that could clarify how DeepSeek actually operates, is therefore coming — and it is not being written for the audience most curious about it.
What a listing changes for everyone else
A public DeepSeek would be the first frontier lab anywhere with a continuous obligation to report numbers, and that matters well beyond China.
Every argument about this industry currently runs on private marks and company-supplied figures. A listed lab has to publish revenue, cost of revenue and capital expenditure on a schedule, audited, with a regulator behind it — which would give the sector its first outside view of what training and serving frontier models actually costs a company that does it at scale.
The numbers would come with caveats. Chinese disclosure standards differ, the filings are in Mandarin and written for domestic institutions, and a company under export controls has obvious reasons to be sparing about compute. Even so, an audited quarterly number from any frontier lab is more than the field has now.
The valuation question nobody can check
Roughly 74.5 billion dollars would make DeepSeek one of the most valuable private AI companies anywhere, on a business whose published revenue picture is thin.
What is observable is that DeepSeek competes hard on price. We covered its peak and off-peak pricing structure, which raises rates when demand is high — a mechanism that reveals a company managing capacity rather than one with capacity to spare.
A valuation of this size on a low-price, capacity-constrained business is a bet on volume and on domestic market position rather than on margin. That may well be right. It is also the part a prospectus would have to substantiate, and no prospectus exists.
What to watch
The filing itself, which is the first document with numbers in it. Whether the pre-IPO round closes at the reported figure, since a pre-IPO mark and a listing price are different animals and the gap between them is the market's actual verdict.
And what the prospectus says about compute. For a Chinese frontier lab operating under export controls, the disclosure of what it trains on and where those chips came from is the most consequential paragraph the document will contain — and the one most likely to be written carefully.
