From Price Slasher to Price Setter: DeepSeek Hits $1 Billion ARR and Closes In on a $7.5 Billion Raise

The lab that taught the industry to expect near-frontier intelligence at a fraction of frontier prices has just delivered its most surprising number yet: DeepSeek’s annualized revenue run rate has crossed $1 billion, more than doubling from under $500 million a few months ago. And it did so by doing the one thing almost nobody expected — raising prices.

According to a report by Juro Osawa and Qianer Liu in The Information, the Chinese AI lab is simultaneously finalizing a fundraising round of roughly 50 billion yuan — about $7.5 billion — at a valuation of 500 billion yuan (roughly $74 billion), to be completed by the end of October ahead of a planned Shanghai listing. Reuters separately confirmed the revenue milestone on September 24.

The number that reframes DeepSeek

For most of the past two years, DeepSeek existed in Western coverage as a paradox: a research outfit that matched frontier labs on benchmarks while charging a pittance, subsidized by its parent High-Flyer’s quant-trading profits. Its January 2025 launch of R1 — priced at a fraction of OpenAI’s equivalent — triggered a market rout and forced every rival to cut prices. Founder Liang Wenfeng was “the price slasher.”

The new figures invert that story. Annualized revenue run rate — current monthly revenue multiplied by twelve — has climbed from under $500 million to over $1 billion in a matter of months. The Information attributes the jump primarily to the API price increases DeepSeek rolled out in August, which raised the price of a million tokens during peak hours from $0.28 to $1.32 for its cheapest tier — a 4.7x increase — with off-peak access at half price. Fortune reported at the time that increases across the lineup ran between roughly 2.3x and 4.5x depending on model and time slot.

That revenue survived — indeed, was created by — those hikes tells you demand for DeepSeek’s models is far less price-elastic than assumed. Enterprises building on V4-family models appear to have absorbed the new pricing rather than migrating, and the August repositioning of V4-Flash from loss-leader to sustainably-priced product converted usage into revenue.

Why the price hikes worked

Three factors made the strategy viable where it might have failed elsewhere.

First, DeepSeek’s models remained among the cheapest near-frontier options even after the increase. At $1.32 per million peak tokens, the entry tier still undercuts most Western equivalents substantially. The lab raised prices from an unsustainable floor, not toward a premium.

Second, capacity constraints were real. A Hacker News discussion of the August update captured the practical reality: DeepSeek lacked the GPU capacity to serve demand at the old prices, and rationing by price is the classic market answer. Serving more tokens at a negative gross margin — which 36Kr reported was likely the case for V4-Flash’s original pricing — was burning money to subsidize an image.

Third, the open-weight strategy built a moat that now monetizes. Because DeepSeek’s models can be run anywhere, they became the default backbone for Chinese enterprises, sovereign deployments, and cost-sensitive developers worldwide. When the official API repriced, much of that demand stayed because switching away from the ecosystem has its own costs.

The $7.5 billion raise and the Shanghai listing

The fundraising now being finalized is DeepSeek’s second external round, and its largest by an order of magnitude. Reuters first reported in July that the company was seeking to raise as much as 50 billion yuan at a 500 billion yuan valuation ahead of an onshore IPO; The Information now reports the round is in its final stages with an end-of-October target.

The structure marks a sharp break from the lab’s history. Liang historically routed all investment through a private limited partnership tied to High-Flyer, keeping outside capital at arm’s length. According to remarks attributed to the founder in a four-hour investor meeting whose contents circulated widely in Chinese media this summer, this round permits direct equity placement — a concession to the scale of capital the frontier now demands.

The choice of venue is equally deliberate. A Shanghai listing keeps DeepSeek inside China’s capital markets at a moment when Washington is tightening export controls and US listing of Chinese AI firms has become politically fraught. It also gives domestic institutional investors exposure to what is arguably China’s most important AI asset — and gives Beijing a mechanism to keep it that way.

What it means for the market

The straightforward read is commercial vindication. A lab long caricatured as a subsidized disruptor is now generating revenue at a run rate that would place it among the serious AI businesses globally — still far behind OpenAI’s reported figures, but in the same order of magnitude as the tier beneath the very top, and growing faster in percentage terms than almost anyone.

The subtler read concerns pricing power across the industry. If DeepSeek can raise prices 2-4x and accelerate revenue, the assumption that AI inference is destined for permanent deflation weakens. Frontier-adjacent intelligence, it turns out, is not a commodity — customers pay for capability, reliability, and ecosystem lock-in, even when a nominally cheaper open-weight alternative exists.

For China’s AI sector specifically, the milestone lands at a useful moment. Alibaba’s Qwen team cut voice-API prices by up to 95% the same week DeepSeek’s revenue jump made headlines — two strategies diverging in the same market. Qwen is buying volume and cloud attach; DeepSeek is harvesting margin from scarcity. Whether both can succeed simultaneously will shape how Chinese AI monetizes over the next two years.

The end-of-October close of the round, and the Shanghai listing to follow, will be the real test. At a 500 billion yuan valuation on a $1 billion run rate, investors are paying roughly 74 times current annualized revenue — a multiple that assumes the price-hike growth continues, capacity bottlenecks ease, and US chip restrictions don’t cap the models’ trajectory. Those are assumptions even the price slasher can’t guarantee.

One thing is certain: the era of treating DeepSeek as a pricing anomaly is over. It is now a revenue-generating, capital-raising, IPO-bound AI company — one that discovered that the most powerful deflationary force in the industry could, when demand outran supply, become its opposite.