DeepSeek just tested what happens when the cheapest frontier lab in the world starts charging more, and so far its customers are paying. The Hangzhou-based company's annualized revenue run rate has crossed $1 billion, more than double the sub-$500 million pace it was running at a few months ago, according to The Information, which cited two people with direct knowledge of the figures. Chief executive Liang Wenfeng shared the number at a recent investor meeting as DeepSeek races to close a roughly 50 billion yuan (about $7.5 billion) funding round by the end of October at a 500 billion yuan valuation, or about $75 billion.
The timing is the story. In August, DeepSeek raised API prices on its models by between 2.3 and 4.5 times, depending on the model. Liang told investors the increase did not shrink the customer base, according to the report. The change, which took effect on August 16, also introduced peak and off-peak billing for the flagship V4-Flash and V4-Pro models. V4-Pro output tokens now cost $3.96 per million at peak, up from a flat $0.87, while V4-Flash output rose to $1.32 per million from $0.28. The company said at the time it was adopting the new structure to "allocate resources more reasonably." Even after the hike, DeepSeek's prices remain among the lowest of any major model provider.
From Research Lab to Revenue Machine
The new run rate is a sharp departure from DeepSeek's booked figures. Earlier reporting by The Information put revenue for the first seven months of 2026 at about 475 million yuan, or $70.7 million, roughly ten times its revenue for all of 2025. Over that period the API business carried an 82.9% gross margin, while overall gross margin stood at 44.6%, and the company still posted a net loss. The gap between $70.7 million in booked sales and a $1 billion forward pace reflects what a run rate is, a recent month extrapolated across a full year, but it also means the headline number is only as durable as the demand behind it. The report does not break out how much of the jump came from higher prices versus more usage or new customers.
Nearly all of that money comes from developers paying for API access; DeepSeek's consumer chatbot remains free. Liang has also signaled that monetization remains secondary to research, with the company still dedicating more than 70% of its compute to training next-generation models, according to The News International.
The Second Raise and the Road to Shanghai
The new financing is DeepSeek's second outside round in a matter of months. Its first, which closed in June at about $7.4 billion, drew investors including Tencent and battery giant CATL, and was structured so that capital flowed into a limited partnership managed by Liang rather than directly into DeepSeek, preserving his control. The second round has not closed. DeepSeek has tapped CITIC Securities as lead underwriter for a listing on Shanghai's STAR Market, with a public debut targeted for 2027. At 500 billion yuan, the company would be valued at roughly 74 times its current run rate.
The commercial milestone lands in a week when DeepSeek's founder was conspicuously absent from the geopolitical stage. Liang, along with Moonshot AI founder Yang Zhilin, was left off every reported version of the business delegation accompanying President Xi Jinping to his September 24 summit with President Donald Trump, which leaned instead toward EV and manufacturing names such as BYD, Xiaomi and CATL. Chen Zhiwu, chair professor of finance at the University of Hong Kong, described the omission as "a measure to protect them, because at this stage their attendance could put them in an awkward position," according to the South China Morning Post. Alicia Garcia-Herrero, chief economist for Asia-Pacific at Natixis, offered a blunter read: "China knows the U.S. needs a breather, but China itself does not. That's why Xi doesn't want to bring the AI team along."
Why It Matters
For two years DeepSeek has been the industry's price floor, the lab that forced Western rivals to justify their margins. The August hike was the first real test of whether that position was a moat or a crutch. If Liang's account holds, customers stayed even as their bills rose several-fold, which suggests DeepSeek's developer base is buying capability and ecosystem, not just cheap tokens. That is a notable signal at a moment when soaring inference costs are pushing enterprise buyers to shop for alternatives, and it strengthens the case that Chinese labs can build durable businesses rather than simply undercutting the market.
It also changes the investor calculus ahead of a listing. A research-first lab once bankrolled by Liang's quantitative hedge fund, High-Flyer, is now presenting a revenue story to public-market investors. A 500 billion yuan valuation would make DeepSeek the most valuable pure-play AI company to list outside the United States, handing China's domestic markets a flagship AI stock. The caveat is that a run rate is a snapshot. Without disclosure of customer counts, usage trends or revenue by model, investors are being asked to trust that one strong stretch reflects a lasting trajectory.
What to Watch
The end-of-October deadline is the immediate marker: a clean close at 500 billion yuan would lock in the valuation and start the clock on a STAR Market filing, while a slip or a downsized round would signal hesitation among domestic investors. Watch too for whether rival Chinese labs hold their prices to poach DeepSeek customers, which would reveal how sticky its developer base really is. And with a U.S. House select committee already pressing Liang by letter and Chinese regulators probing data practices at DeepSeek and Moonshot, the political risk around China's most visible AI lab is not shrinking as its revenue grows.
“China knows the U.S. needs a breather, but China itself does not.”— Alicia Garcia-Herrero, Chief Economist for Asia-Pacific, Natixis