A price rise that demand absorbed
DeepSeek’s annualised revenue run rate has crossed $1 billion, more than double where it stood a few months ago, The Information reported on Thursday. Chief executive Liang Wenfeng gave the figure at a recent investor meeting, according to the report, and Reuters subsequently carried it.
A run rate is not a year’s revenue. It is what twelve months at the current pace would produce, and for a company whose pricing changed last month it is a particularly fresh number. That is the point of it here: DeepSeek raised API prices by between 2.3 and 4.5 times, and customers stayed. The company told investors its API business ran at an 82.9 per cent gross margin through July.
Why the price went up
DeepSeek built its reputation on being cheap. It undercut Western labs by a wide margin and made that the argument for its models. Raising prices several-fold is a reversal of that positioning, and the company appears to have concluded it could afford one.
The allocation of its own hardware explains part of the decision. DeepSeek told investors that more than 70 per cent of its computing goes to training the next model and less than 30 per cent to serving users — an unusual split for a company with a billion-dollar run rate, and one that makes inference capacity something to ration by price rather than expand. Its most recent release, DeepSeek-V4.1-Flash, came earlier this month.
The raise and the listing
The revenue disclosure arrives alongside a fundraising. DeepSeek is working to close a round of 50 billion yuan, roughly $7.5 billion, at a target valuation of 500 billion yuan — about $74 billion — and is aiming to complete it by the end of October.
Separately, the company has retained CITIC Securities to prepare for a possible listing on the Shanghai Stock Exchange’s STAR Market. Timing, valuation and the size of any offering remain undecided, according to the reporting.
What this changes
A $1 billion run rate puts DeepSeek in a different category from the one it occupied when it was mainly a demonstration that frontier-adjacent models could be trained cheaply. It is now a revenue business with pricing power, a domestic capital market willing to fund it at scale, and a listing in preparation.
The figures here come from the company, relayed by reporters, and none of them have been audited in public. The listing process is where that changes: a STAR Market filing would require disclosure on a schedule DeepSeek has never had to meet. That, rather than the next investor meeting, is the moment the numbers become checkable.
