DeepSeek has suspended its second funding round, telling prospective backers that investment agreements will be delayed indefinitely while founder Liang Wenfeng deals with the fallout from his own remarks. The Chinese AI startup, which raised a record-breaking $7.4 billion in June, was seeking an additional $1.5 billion at a $74 billion pre-money valuation. Now those talks are frozen, with Liang reportedly furious that comments from a private investor meeting circulated online and drew coverage he considered distorted. (Source: Bloomberg/Yahoo Finance)

The leaked transcript reveals why the remarks traveled so far. Liang told investors that "Nvidia is digging its own grave," claimed Huawei's 950 super-node could replace Nvidia's GB200 and GB300, and declared that DeepSeek had closed the gap with American AI leaders to "about two years" while using "only one-twentieth of their computing power." (Source: Hello China Tech transcript compilation) The performance was defiant, technically specific, framed as a David-and-Goliath story of Chinese innovation triumphing over American chip restrictions. Algorithms loved it.

The market responded differently. Where viral clips celebrated Liang as a techno-patriot, institutional investors saw volatility. The funding pause suggests that capital, even Chinese capital, treats AI nationalism as entertainment rather than investment thesis. Tencent committed 10 billion yuan to the first round. CATL put in 5 billion. These are firms with regulatory relationships and cross-border supply chains to protect. They do not need their portfolio companies becoming geopolitical lightning rods.

Liang's control structure means he can make unilateral decisions. He owns 1% directly and 99% through a partnership vehicle, having personally contributed 20 billion yuan to the first round. (Source: Exovera corporate profile, SCMP) That concentration of power let him pause fundraising on principle. It may also be what unsettled backers in the first place. The governance model assumes founder judgment is sufficient. The past two weeks tested that assumption.

American startups operate in a capital market that largely ignores nationalist signaling. OpenAI's Sam Altman does not need to proclaim US technological supremacy to raise funds; doing so would strike investors as odd. Chinese AI firms face the opposite pressure: state-affiliated funds and public opinion reward patriotic framing, even as global capital and technical reality impose cooler assessments. Liang attempted to satisfy both constituencies. He satisfied neither.

DeepSeek's technical claims lack independent benchmarking. The assertion that four Huawei chips equal one Nvidia chip, or that Huawei hardware can fully replace Nvidia's latest offerings, remains unverified. The company's own framing—that it trails by two years despite radical compute constraints—acknowledges a gap that patriotic rhetoric tends to obscure. Liang identified the compute gap as fundamental: "The talent gap is fundamentally a compute gap." US export controls were designed to create exactly this.

The firm is simultaneously preparing for an IPO filing as soon as year-end 2026, targeting Shanghai's STAR Market. (Source: The Straits Times/Reuters) Public market investors will be less forgiving of founder volatility than private backers. The funding pause may be an attempt to reset narrative control before regulatory filings require audited disclosures and mandated quiet periods. It may reflect uncertainty about whether the $74 billion valuation can withstand due diligence uninterested in nationalist applause.

Patriotic AI narratives play well on social media. They play poorly in term sheets. Liang's transcript offered a future where DeepSeek rewrites the narrative entirely, narrowing the lag to "6 months or 3 months." Tencent and CATL, which together contributed 15 billion yuan to the first round, have not said whether they will join the second. The silence is its own answer.