DeepSeek's Revenue Run Rate Hits $1 Billion, Chinese AI Lab Seeks $7.5 Billion at $75 Billion Valuation

DeepSeek, the Hangzhou lab that rattled Silicon Valley in early 2025, told investors its annualized revenue run rate has crossed $1 billion, more than double the under-$500 million pace reported just months earlier. The Information first reported the figure, citing two people with direct knowledge, and multiple outlets including Reuters and Dealroom followed with confirming details.

CEO Liang Wenfeng shared the number at a recent investor meeting, according to The Information. The growth came almost entirely from one place: DeepSeek raised its API prices by 2.3x to 4.5x last month, and customers didn't leave. Liang told investors demand stayed strong anyway, according to the same reporting cited by AI Weekly.

A company jacks up prices on developers by as much as 350%, and usage holds. Either DeepSeek's models are still cheap enough relative to competitors that nobody blinked, or enterprises building AI products don't have anywhere else to go that's this cheap. Both are probably true. DeepSeek's API still runs at an 82.9% gross margin through July, according to figures reported by The Information and repeated across Crypto Briefing, KSL News (via Reuters) and Hugging News.

The free chatbot app that most people associate with DeepSeek, the one that competes with ChatGPT, makes the company nothing. Zero. Every dollar of that $1 billion run rate comes from developers and businesses paying to plug into DeepSeek's models through its API, according to Runtime Wire's reporting on The Information's findings. The consumer product is a loss leader for brand awareness, not a business.

The Fundraise, and the Numbers Don't Quite Match

DeepSeek is finalizing a second external funding round, targeting roughly 50 billion yuan at a 500 billion yuan valuation, aiming to close by the end of October 2026. Converted at reported exchange rates, that lands most outlets, including KSL News and the Economic Times, at roughly $7.45 billion raised at a $74 billion valuation. Crypto Briefing and Hugging News rounded it to $7.5 billion at $75 billion. Dealroom reported a notably different figure, $6.88 billion at $68.8 billion, attributing it to the same underlying reporting. None of these numbers are final. The round has not closed.

If it lands anywhere near $75 billion, that's a roughly 44% jump from the $52 billion valuation DeepSeek closed at in June 2026, when it raised about $7.4 billion in its first outside funding round, according to Crypto Briefing. Four months, 44% up. That's the kind of number that should make anyone ask whether this is being priced off actual cash flow or off fear of missing the next DeepSeek moment.

DeepSeek has hired CITIC Securities to lead an eventual IPO on Shanghai's STAR Market, targeted for 2027, and recently brought on its first CFO, Yan Wentao, a former Hillhouse Capital partner, according to Hugging News. Investor meetings in Beijing and Hangzhou reportedly required attendees to surrender their phones and other devices before entering, a level of operational security unusual even by Chinese tech standards.

One Outlet Got the Story Wrong, Then Fixed It

Reuters' original dispatch, picked up by the Economic Times, described DeepSeek as aiming to raise the 50 billion yuan "in an IPO" ahead of a Shanghai listing. That's not accurate. The funding round is a private fundraise, separate from the eventual public listing that CITIC Securities is preparing for 2027. Reuters issued a correction on its wire copy, changing the language to specify a "fundraise." This fix is reflected in KSL News's version of the story but not caught in the Economic Times republication, which still describes it as an IPO raise.

The Overhang Nobody's Resolved Yet

China's Cyberspace Administration opened a probe into DeepSeek and Moonshot, another Chinese AI firm, over allegations that the companies routed sensitive user data to Anthropic's Claude models, according to AI Weekly. No findings, charges, or conclusions have been announced. It's an open investigation, not an established violation, and it's unclear how or whether it affects the October fundraise timeline or the 2027 listing plan.

Separately, Liang told investors DeepSeek expects to bring on Huawei-made training chips as early as the fourth quarter of 2026, and that internal testing shows its smaller models run fine on ordinary gaming GPUs for inference, according to Dealroom. More than 70% of the company's current compute is dedicated to training new models rather than running existing ones, per Reuters. Whether the Huawei chip timeline holds, and whether the CAC probe turns up anything, are the two loose threads that will determine if DeepSeek's next valuation jump is built on as solid ground as the last one.