DeepSeek pauses $1.5B round after Liang transcript leaks
The Chinese lab halted its second raise near a $71B valuation after a leaked four-hour investor briefing by founder Liang Wenfeng went viral.
What happened
DeepSeek has put its second funding round on hold after a lengthy recording of founder Liang Wenfeng addressing first-round backers circulated widely online, according to reporting citing people familiar with the matter. The interrupted round had been targeting at least 10 billion yuan (roughly $1.5 billion) at a valuation near $71 billion, following an earlier raise of about $7 billion completed in June.
The leaked material — a transcript running nearly four hours and reproduced across Chinese tech outlets and X — captures Liang walking shareholders through DeepSeek's roadmap in unusually candid terms. He is quoted framing the US-China gap almost entirely as a compute problem: the lab is "about two years behind" frontier labs, he reportedly said, yet operating on roughly one-twentieth of their computing resources, with the goal of narrowing the gap to six months without a comparable hardware budget.
The strategic picture
Liang also restated the positions that have defined DeepSeek's identity: a refusal to chase consumer-app scale in the mold of ByteDance or Tencent, a commitment to keeping the lab's strongest models open-weight, and a stated priority of AGI research over near-term profit. The candor is precisely what made the leak damaging — private guidance on strategy, pricing logic and dependence on Nvidia silicon was never meant for public consumption, and its disclosure complicates active fundraising conversations.
Why it matters
DeepSeek is the most closely watched of China's open-weight labs, and its financing arithmetic is a proxy for how investors price Chinese frontier ambition against US export controls. A pause at a $71 billion valuation signals that even the country's marquee AI name is not immune to governance and disclosure risk. More consequentially, Liang's own framing — capability parity blocked chiefly by chips, not talent — lands directly in the middle of the Washington debate over whether to restrict Chinese models, giving both hawks and open-source advocates fresh ammunition.