The Numbers: Under $500M to $1B in Months
DeepSeek's annualized revenue run rate has reached $1 billion, more than double the figure of under $500 million recorded a few months earlier, according to pymnts.com reporting published 24 September 2026. The underlying report, dated 23 September, cited unnamed sources. ARR Club's tracker logs the same trajectory: $1 billion now, under $500 million before. Doubling in months, from that base, means demand expanded faster than distribution alone could explain.
A run rate is a projection, not booked revenue. It multiplies the most recent month by twelve, which flatters any business with accelerating sales. In DeepSeek's case the doubling still signals material growth, because half a billion annualized sits well past the stage where a handful of contracts move the number. The open question is how the gain splits between volume and price, and the pricing record supplies part of the answer.
API rates rose by factors of 2.3 to 4.5 during the same window, per arr.club. If revenue doubled while unit prices multiplied several times over, volume held roughly steady or grew despite the increase. That suggests demand elasticity near zero in the current market for the company's systems. Price, in other words, did a large share of the work the headline attributes to growth.
Price Hikes That Didn't Break Demand
The pricing move was steep by any standard: increases of 2.3 to 4.5 times across the API, with no significant effect on customer retention. Arr.club attributes the retention to high global demand for the company's cost-effective AI systems and for the architectures that power agents. Buyers of inference capacity had somewhere to go. They stayed. That is the entire finding, and it is more informative than the revenue figure itself.
Increases of that magnitude normally function as a churn test. Passing one implies customers priced DeepSeek against alternatives and still found it favorable, which is consistent with the cost-effective positioning the evidence describes. The inference has limits: the record does not state churn rates, contract sizes, or which customer segments absorbed the increase. It establishes only that retention held. The mechanism behind that outcome remains unstated in the available sources.
One allocation detail frames the pricing decision. The company dedicated over 70 percent of its compute to training next-generation models rather than serving inference, per arr.club. That choice caps serving capacity at a moment when demand is high. Raising prices under a capacity constraint is rationing, and the revenue result indicates the constraint bound. Whether training investment converts into a model that sustains this pricing is the unresolved variable.
Capital Plans: Funding Round and Shanghai Listing
DeepSeek is planning a $7.5 billion funding round and aims for a future listing on the Shanghai Stock Exchange, per arr.club. Pymnts.com's headline ties the company to a $75 billion Shanghai IPO. The two sources agree on the listing destination and the scale of the raise; neither states a timetable, an exchange board, or named investors. The record is silent on prior funding rounds in the permitted sources, so no baseline for dilution can be established here.
The $75 billion figure against a $1 billion run rate is a multiple of 75 times annualized revenue. That pricing rests on growth trajectory and strategic position rather than current earnings, and it presumes the run rate is a floor rather than a peak. The open question is whether public-market investors in Shanghai will underwrite that multiple on the same terms private buyers did, given that the revenue doubling is months old and partly price-driven.
Three signals are worth tracking. First, whether the $7.5 billion round closes at its stated size, which would test private appetite at the current valuation. Second, whether the Shanghai listing proceeds on a defined schedule, since Chinese exchange listings for AI firms carry regulatory steps the record does not detail. Third, whether the run rate holds now that the price increases have lapped, which would separate durable demand from a one-time repricing.