BEIJING – The last time DeepSeek accepted outside money was five weeks ago. It walked away from that round at a $50 billion valuation with $7 billion in its account, the first institutional capital the company had ever taken. Bloomberg reported on Tuesday that DeepSeek is now in talks to raise another $1.5 billion at a $71 billion valuation. The interval is shorter than a fiscal quarter. The repricing is faster than anything the Chinese AI sector has produced before.
The talks, described by people familiar with the matter, would value DeepSeek at $71 billion, a 42 percent jump from the figure that was itself a first-round reference point less than a month earlier. A $1.5 billion raise at that level would bring DeepSeek’s total outside capital to roughly $8.5 billion in under two months, a pace that puts it in the same territory as some of the largest AI capital events in American history. DeepSeek declined to comment.
Alongside the funding discussions, TechCrunch reported that DeepSeek is targeting an initial public offering in 2027, though the timeline could accelerate to late 2026 depending on market conditions. No exchange has been identified. Whether DeepSeek lists in Hong Kong, Shanghai, or pursues a dual listing is a question that carries weight beyond finance: a US-listed DeepSeek would face political pressure in a Congress that has already moved to sanction Chinese AI companies over distillation attacks on American labs.
Founded in 2023 by Liang Wenfeng, who also built hedge fund High-Flyer Capital, DeepSeek became the most disruptive Chinese AI story of the decade in January 2026 when it released models that matched or outperformed leading US systems at a fraction of the training cost. The release set off a brief stock market panic in the United States, briefly erasing more than $500 billion in market value from AI-adjacent companies as investors revised their assumptions about how expensive frontier AI actually needed to be.
That efficiency thesis has held. As of June 2026, DeepSeek accounted for nearly 23 percent of tokens processed through Vercel’s enterprise AI gateway, a proxy for real production usage rather than benchmark performance. Anthropic held 32 percent. What the Vercel data captures is that DeepSeek is not a research curiosity; it is processing enterprise workloads at a scale that makes it the second-largest AI provider in at least one major commercial distribution channel.
The deeper tension in DeepSeek’s rise is the infrastructure it runs on. DeepSeek’s cloud operations rely primarily on Huawei Technologies chips, a consequence of US export restrictions that have blocked Nvidia’s most advanced processors from reaching Chinese buyers since 2022. That dependency was once treated as a ceiling on Chinese AI ambition. DeepSeek’s results, and now its valuation trajectory, suggest it has become something else: proof that the restrictions accelerated domestic Chinese alternatives rather than simply delaying them.
The investors in the new round have not been confirmed publicly. Tencent Holdings participated in DeepSeek’s prior $7 billion raise, alongside Beijing’s National Artificial Intelligence Industry Investment Fund, one of the state-backed vehicles the Chinese government has deployed to accelerate AI as a national strategic priority. The composition of DeepSeek’s backers matters because it maps where the company sits on the spectrum between private technology firm and strategic state asset, a question that any eventual IPO would force into the open.
The pace of DeepSeek’s capital accumulation is occurring as US AI financing is also running at record levels. Goldman Sachs last week reported a 78 percent surge in second-quarter profit, with AI-related financing cited as a primary driver. But the two capital stories point in different directions. US AI companies are building on scarce compute that costs tens of billions to acquire and deploy. DeepSeek has demonstrated it can produce competitive results on hardware that Washington never intended it to have, and it is raising money at a pace that suggests investors believe the efficiency gap is still closing.
The US export control architecture, designed to preserve American technological leadership in AI by limiting China’s access to advanced chips, has produced a result its architects did not fully anticipate: it forced DeepSeek to optimize. The company’s core innovation, building more efficient models that extract more value from constrained compute, is directly attributable to the restrictions the export regime imposed. The $71 billion valuation is, in a narrow sense, the market’s assessment of how far that optimization has gone and how much further it could go.
What the new round does not answer is where DeepSeek’s ceiling actually sits. The gap between its models and the leading US systems has narrowed but has not closed on the hardest tasks. The IPO exchange remains unconfirmed. The size of Tencent’s stake and the terms of Beijing’s state fund investment are not public. DeepSeek has not disclosed what it plans to do with the capital or what products are in active development. The company that declined to comment Tuesday raised more money this spring than most AI companies raise in a decade, and has still not said publicly what it intends to build next.
