AINEWS
On a 4 hour call, DeepSeek's Liang Wenfeng told Chinese strategics he didn't want their money, and walked out with $7.4B in a deal that turns visible disinterest into pricing power, with one caveat.
On a mid-May afternoon, a small room of Chinese strategic investors dialed into a Tencent Meeting video call, China's equivalent of Zoom, to hear DeepSeek founder Liang Wenfeng explain why they should give his company roughly $7.4 billion. Two attendees per firm. No agenda. Four hours and 118 questions. Liang spent most of them telling the room the lab didn't need the money. "We have no intention of wanting more," Liang said, and he kept saying it until the round closed.
This was DeepSeek's first external raise. The Hangzhou-based lab, whose open-weight models triggered a 2024-2025 global selloff in AI infrastructure stocks, had been funded entirely by the High-Flyer hedge fund (Chinese name Huanfang Quant) since its founding. The mid-2026 round was a one-time strategic capitulation, not the start of a recurring posture. The deal terms put it among the largest first-time raises any Chinese AI company has closed: roughly ¥50 billion (~$7.4B) raised at a post-money valuation that a Chinese regulatory filing pegged around $52 billion in July, with secondary coverage bracketing it in the $50 billion-plus range and oztalking citing a figure closer to $59 billion. Three sources, one cluster: somewhere around $50B to $60B post.
With a scarce asset in front of a concentrated, capital-rich bidder pool, the seller's repeated "no" forces buyers to raise their own offers out of fear of losing the slot. The transcript reads like an auction run in reverse. The seller keeps signalling exit; the buyers keep walking the price up. The call's structure made the signal land cleanly. Two attendees per firm meant investors had no way to coordinate a price, no agenda meant no off-deck negotiations, and a founder-controlled Q&A meant the only variable buyers could move was the bid.
Two features of the Chinese market made the deal work. First, the bidder pool: the oztalking-cited roster puts Tencent around ¥10 billion (~$1.4B), CATL, the world's largest EV battery maker, around ¥5 billion (~$700M), JD.com, NetEase, and IDG Capital around ¥3 billion each (~$430M), and the National AI Industry Investment Fund around ¥1 billion (~$140M). Flush, strategic, and motivated by both commercial interest and alignment with China's AI industrial policy. Second, the limited-partnership structure through which the capital routes, managed by Liang personally and designed to preserve his absolute control over the lab despite the new money. Liang is not selling DeepSeek. He is selling access, on his terms, with the structure guaranteeing that.
Three conditions made DeepSeek's restraint work: a scarce asset, a concentrated bidder pool, and a seller who could credibly walk away. Most companies, even most AI companies, hit none of them. The Chinese strategic-cash context is also load-bearing. The same set of names underwriting China's AI industrial policy paid the premium, and the terms that looked generous were, in part, the price of staying close to a flagship lab. Restraint as pricing power is a real lever, but the conditions it requires are rarer than the transcript makes them look.
The next test is the next round, if there is one. The High-Flyer stake is reportedly being wound down, the LP is the only formal governance channel, and Liang's posture in the transcript leaves little room for a second walk-back. If a second deal happens, the same room will be smaller, or the same posture will be harder to keep. Either way, the 4-hour call set a price for what "we don't want more" is worth in this market.