Leaked Comments = DeepSeek Intends to Slash Costs

DeepSeek has suspended its second funding round, one that was targeting a $71 billion pre-money valuation, after private comments from founder Liang Wenfeng leaked and went viral across Chinese social media. The comments themselves weren't damaging. They were candid, which in this market amounts to the same thing.

Key Points

• Bloomberg reports DeepSeek verbally told prospective backers it would not sign investment agreements as expected, effectively pausing the round
• The round had targeted at least 10 billion yuan at a pre-money valuation between 480 and 500 billion yuan, roughly $71 to $74 billion
• Leaked transcripts, unverified by Bloomberg, reportedly show Liang telling investors China still trails the U.S. in AI funding and computing infrastructure and that DeepSeek remains dependent on Nvidia hardware
• The pause comes weeks after DeepSeek closed a $7 billion first round in June, backed by Tencent, CATL, and China's National AI Industry Investment Fund
• Liang's fortune reportedly hit $36 billion after the first round, making him the wealthiest AI model founder in the world, ahead of Anthropic's Dario Amodei and OpenAI's Greg Brockman

The Comments Weren't the Problem: The Timing Was

Nothing in what Liang reportedly said is controversial to anyone who follows the industry closely. Chinese labs using Nvidia chips despite export controls, and trailing U.S. labs on raw compute, are widely understood facts. What made this newsworthy is where those comments landed: directly against Beijing's preferred narrative that China's AI sector has already closed the gap with Silicon Valley, at the exact moment DeepSeek needed investors to buy into a valuation ten times its size from eight months prior. A founder's private candor and a public fundraising narrative are not always compatible, and this week they collided.

A Pause Isn't a Collapse, but It's Not Nothing Either

Bloomberg describes this as a temporary pause, not a canceled round, and DeepSeek could resume talks once attention fades. That framing is probably accurate. It's also worth noting what got interrupted: a valuation jump from roughly $50 billion in June to $71 billion in the same summer, based largely on momentum and narrative rather than a corresponding leap in disclosed revenue or compute capacity. Momentum-driven valuations are efficient at capturing upside and just as efficient at capturing a founder's unscripted honesty as downside risk. Anyone tracking frontier lab valuations, ours included as part of standing growth strategy conversations with clients, should treat this as a data point on how thin the story-to-fundamentals ratio can get across the industry, not just in China.

What This Says About the Broader AI Funding Environment

This is the second time in as many months that a widely covered AI story has hinged on the gap between public confidence and private admission. Investors are pricing labs on projected dominance; founders, when unguarded, describe something closer to competitive parity or catch-up. That gap doesn't make the technology less real. It does make the valuations attached to it worth stress-testing before treating them as settled fact, whether the lab in question is in Hangzhou or San Francisco.

For teams evaluating AI vendors on the strength of headline valuations, that stress test is worth doing before it's forced on you. If you want help thinking through which parts are durable versus narrative-driven, that's a conversation our AI marketing services team has regularly.