DeepSeek, the Hangzhou-based AI lab that sent shockwaves through Silicon Valley in January 2025 with its open-source R1 model, has told prospective investors in its second fundraising round to stop — for now. The round would have valued the company at roughly 500 billion yuan, or about $71 billion. The trigger was not a market crash, a regulatory objection, or a board dispute. It was the viral spread of comments its founder made in a private investor meeting.
🔍 THE BOTTOM LINE: DeepSeek’s founder Liang Wenfeng holds near-total voting control and can halt a multi-billion-dollar round on his own say-so. He did exactly that — reportedly because he was unhappy about how his own words were being recirculated online. That’s not a company being buffeted by forces beyond its control. It’s a company being steered by one man’s instincts, and investors just got a reminder of what that means for their money.
What Happened
According to Reuters, which cited Bloomberg’s original reporting, DeepSeek verbally informed some would-be backers they would not be signing investment agreements in the coming days as expected. The company indicated it might restart the process later.
The planned valuation — approximately $71 billion — represented a 37% step up from DeepSeek’s first outside round, which closed in June 2026 at roughly $52 billion. That first round raised about $7.4 billion from Tencent, battery maker CATL, and China’s state-backed National Artificial Intelligence Industry Investment Fund. Liang Wenfeng himself was the single largest investor, reportedly putting in around $3 billion of his own money.
The second round had only begun preliminary discussions in mid-July, weeks after the first close. The cadence was aggressive even by US megaround standards. DeepSeek needs the capital — it is building its own data centres and buying AI chips, with its push into AI agents driving sharply higher compute demand.
The Leak That Stopped a Deal
The proximate cause of the pause, according to Bloomberg, was Liang Wenfeng’s frustration over online reports of comments he made to investors during DeepSeek’s first fundraising round. Those remarks, drawn from a leaked account of a roughly four-hour investor meeting, ranged across the company’s strategy and US-China AI competition. One notable claim: the gap between the two countries’ AI capabilities comes down to computing power, not talent.
The account circulated quickly through Chinese tech and investor circles in the days before the pause. DeepSeek has not confirmed the leaked comments are accurate. But the damage — if that’s the right word — was done. The leak turned a communications problem into a financing one.
As Unite.AI reported, the episode turns on a fundamental tension: DeepSeek courts capital while guarding its secrecy, and the viral spread of a private investor meeting collided with that instinct head-on.
Why Liang Can Simply Walk Away
What makes this pause unusual is the concentration of power. Liang reportedly holds the large majority of DeepSeek’s equity and nearly all of its voting rights, with most outside capital routed through a limited partnership he controls and locked up for years. The decision to stop was effectively his alone. The backers wiring in billions have little say over the timing.
This fits the ethos Liang has built the company around. In a collected set of 52 statements from the investor meeting, he repeatedly framed restraint as strategy: not pursuing user numbers, not going closed-source, not building 3D/video generation or world models. “Restraint is a strategy to increase the probability of achieving AGI,” he said. Products are “byproducts along the journey to AGI.”
A founder who frames capital discipline as strategy is also one willing to halt a $71 billion raise over how his words get repackaged online. The same concentration of control that reassures aligned backers — who know Liang isn’t going anywhere — also leaves them exposed to his read of the moment.
The Chinese AI Capital Rush, Briefly Paused
The backdrop matters. DeepSeek’s pause stands out against a Chinese AI sector racing to convert model momentum into capital. Rival Moonshot AI is seeking approval to list in Hong Kong. GPU designer MetaX has filed confidentially for its own Hong Kong IPO. Beijing is weighing new limits on exporting its best AI and chip designs.
Reuters separately reported that DeepSeek had started early deliberations on a potential IPO on Shanghai’s Nasdaq-style STAR Market. So the company that spent its first years insisting it had no interest in outside capital — no fundraising, no IPO, no rush to commercialise — is now simultaneously pausing a private round and exploring a public listing. That’s not a retreat from capital markets. It’s a founder who wants to set the terms.
DeepSeek’s open-source model strategy, which produced DeepSeek V4 as a genuine GPT-5 competitor, has given the company enormous leverage. The question for investors is whether that leverage compounds or whether the founder’s willingness to walk away from the table makes the bet less attractive, not more.
NZ Angle: What a Funding Pause Means for Sovereign AI
DeepSeek’s pause is a reminder that the global AI investment landscape is not a smooth, market-driven process. It is shaped by founder decisions, geopolitical pressure, and — in China’s case — state backing with strategic strings attached. For New Zealand, which is exploring its own sovereign AI options, the lesson is that access to frontier models from any single source is never guaranteed. A founder’s mood can pause a $71 billion round. A geopolitical decision can cut off chip supplies entirely. Diversification — across models, suppliers, and jurisdictions — is not a hedge. It’s a necessity.
❓ FAQ
Has DeepSeek permanently cancelled its funding round? No. The company told investors the round is suspended, not cancelled, and indicated it may resume the process at a later date. Reuters noted that DeepSeek could not immediately be reached for comment outside regular business hours.
Why does Liang Wenfeng have so much control? He holds the large majority of DeepSeek’s equity and nearly all voting rights. Outside capital is routed through a limited partnership he controls, with shares locked up for years. This structure means strategic decisions — including whether to accept investment — are effectively his alone.
What were the viral comments about? The leaked account of a four-hour investor meeting covered DeepSeek’s strategy and US-China AI competition. The most circulated claim was that the gap between US and Chinese AI comes down to computing power rather than talent. DeepSeek has not confirmed the leaked comments are accurate.
Does this affect DeepSeek’s products or API? Not in the short term. DeepSeek’s open-source models, including V4, remain available. The company has said it charges “a reasonable profit, not a profit-maximising price” and that low cost is a deliberate strategy to make models accessible and to allow the company to train larger models on limited compute.
🔍 THE BOTTOM LINE
Whether the pause proves brief or lasting will say more than the $71 billion valuation did. If the round resumes within weeks, the episode reads as a founder managing his message — uncomfortable with leaks, unwilling to let outside narratives set the terms of his financing. If it stalls, it becomes the first real friction in a fundraising run that, until now, only pointed up. For investors weighing the next round, the fresh data point is simple: getting into DeepSeek means getting in on Liang’s timeline, at a pace set by a founder who has shown he will pull back when the moment feels wrong.
📰 Sources
- Reuters — DeepSeek tells prospective investors of funding pause
- Bloomberg — DeepSeek Said to Tell Backers of Funding Pause After Viral Posts
- Unite.AI — DeepSeek Pauses Second Funding Round After Viral Founder Remarks
- KuCoin News — DeepSeek Founder Liang Wenfeng Shares Key Insights from 4-Hour Investor Meeting
