DeepSeek has verbally informed prospective investors that it will not sign investment agreements for its second funding round, pausing a raise that had targeted a valuation of approximately 500 billion yuan, roughly $74 billion, and sought to raise up to 50 billion yuan ($7.4 billion). The decision follows the viral leak of a private investor call in which CEO Liang Wenfeng made a series of unusually candid admissions about China’s structural disadvantages in the global AI race, triggering a wave of online commentary that the company found difficult to manage.
What Liang Wenfeng Actually Said About China’s Compute Gap
The controversy stems from quotes given to investors in the lead-up to an earlier funding round. According to an unauthenticated English PDF of the transcript that circulated on GitHub, Liang stated that China’s lack of compute resources and its continued reliance on Nvidia explains the capability gap between Chinese and American frontier models. He reportedly noted that this compute deficit leads to a talent brain drain, not because Chinese engineers lack ability, but because researchers who want to push the frontier of model training need access to hardware that China’s export-controlled environment cannot reliably provide.
Liang also reportedly implied that Nvidia was “the agent of its own downfall” by forcing Chinese companies to build workarounds through Huawei’s Ascend ecosystem. By pushing Chinese firms off its hardware and software stack, Nvidia is inadvertently accelerating the development of a domestically rooted alternative that, if it matures, would make Chinese AI companies permanently independent of US silicon. Liang reportedly predicted a breakthrough would come “in a year or so.” The viral quotes were initially posted by a Tencent-owned news publication, but the link has since been taken down.
The authenticity of the transcript remains unconfirmed. However, Kyle Chan of the Brookings Institution noted that he was “a bit less skeptical” the interview was genuine after the funding pause was confirmed, the company’s behavior is consistent with a real attempt to manage fallout from a real leak rather than a fabrication.
The Funding Pause and the STAR Market IPO
Bloomberg reported that anonymous sources close to DeepSeek say Liang is experiencing “frustration” over the online reaction, which he did not anticipate. The company has verbally communicated to prospective investors that it will not proceed to signing investment agreements for the time being, though it “might choose to resume the deal process at a later date.”
The pause comes at a sensitive moment. DeepSeek has reportedly begun the process of exploring an initial public offering on China’s STAR Market, a move that would represent a significant shift in the company’s capital strategy. The STAR Market, designed for high-tech companies, would give DeepSeek access to domestic retail and institutional investors without the geopolitical exposure of a US listing. The company closed a $7.4 billion round at a $50 billion valuation earlier this year and had been targeting a $71 billion valuation in the new round.
A domestic IPO would also give DeepSeek a degree of insulation from the kind of US regulatory action that has threatened the listings of other Chinese technology companies. The STAR Market has become an increasingly attractive venue for Chinese AI companies that want access to public capital markets without depending on the continued stability of the US-China financial relationship.
Market Fallout and the High-Flyer Quant Fund Slump
The fallout from the viral posts has extended beyond the funding round itself. Liang’s quant fund, High-Flyer Capital Management, slumped 15.7% in the week of July 17, a significant drawdown for a fund that has historically been one of China’s top-performing quantitative strategies. High-Flyer is the vehicle through which Liang built his fortune before pivoting to found DeepSeek, and its performance is closely watched as a proxy for the health of China’s AI-adjacent financial ecosystem.
Fortune notes that Forbes currently rates Liang as the 50th richest person in the world. The combination of the funding pause, the quant fund drawdown, and the public controversy over his leaked remarks represents an unusual moment of vulnerability for a founder who has, until now, maintained an unusually low public profile for someone running one of the world’s most consequential AI companies. DeepSeek’s communications team has not issued any public statement on the matter.
The Deeper Strategic Question Behind the Leak
The viral posts have sparked a broader debate in China about the country’s AI trajectory. Liang’s candid assessment of the compute gap, delivered in what he believed was a private setting — aligns with the structural reality that DeepSeek’s own fundraising history has highlighted: China’s capital markets are deep enough to fund AI startups at enormous valuations, but the underlying hardware constraints remain a binding limit on what those companies can actually build.
Gizmodo reported that the leaked transcript circulated widely among AI researchers and investors before the funding pause was confirmed, with many noting that the candor of Liang’s remarks, particularly his acknowledgment of Nvidia dependence, was striking for a CEO of a company that has publicly positioned itself as a symbol of China’s AI self-sufficiency. The irony is that Liang’s leaked remarks, however uncomfortable for DeepSeek in the short term, articulate a problem that China’s entire AI industry is grappling with.
The compute gap is real, the talent brain drain is real, and the dependence on Nvidia, despite years of domestic substitution efforts, remains a structural vulnerability. Whether the funding pause is a temporary tactical retreat or the beginning of a more fundamental reassessment of DeepSeek’s capital strategy remains to be seen.