Hangzhou-based DeepSeek is barreling toward one of the most closely watched initial public offerings in China’s artificial intelligence sector, with founder Liang Wenfeng personally committing roughly $3 billion of his own capital and setting his sights on a Shanghai STAR Market listing that could value the company at as much as $71 billion. According to reports circulating in Chinese financial media and aggregated by outlets including the South China Morning Post, the company is targeting a second-quarter 2027 debut, a timeline that reveals both Liang’s confidence and his willingness to keep DeepSeek off the public market longer than any of its major domestic peers.
That patience is now the defining feature of the DeepSeek story. Where rivals such as Moonshot, MiniMax and Z.ai have raced to public markets to convert model fame into share prices, Liang has openly told colleagues and investors that artificial general intelligence, not quarterly revenue, remains the company’s only real objective.
His willingness to sink an estimated $3 billion of personal wealth into the company, on top of the $7.4 billion DeepSeek raised in a June round, has become one of the most talked-about capital commitments in the industry. It also underscores an unusual governance reality: Liang, a former quantitative hedge fund manager, still controls the vast majority of DeepSeek’s equity and appears to have designed the IPO structure to preserve that control.
A different flavor of listing
The choice of the Shanghai STAR Market is itself a statement. Fortune and other outlets tracking the Chinese AI IPO rush have noted that STAR has become the venue of choice for what Beijing considers “national champions”, companies whose strategic value to the Chinese state supersedes their near-term commercial performance. Hong Kong, by contrast, has attracted the more consumer-facing model labs: MiniMax and Z.ai (formerly Zhipu AI) both debuted on the Hong Kong exchange in early January 2026, and Moonshot is preparing its own Hong Kong listing within six months at a valuation that has surged from about $30 billion in June to roughly $50 billion, as we reported in EastFrontier’s coverage of China’s AI IPO wave.
That geographic split matters. Shanghai STAR listings require regulatory blessing that effectively certifies a company as strategically important, and they subject issuers to tighter disclosure rules on technology stack, supply chain and end-use applications. Hong Kong offers faster access to international institutional capital and looser restrictions on consumer-facing narratives — a better fit for Moonshot, whose Kimi K3 model has pushed annual recurring revenue to roughly $300 million on the back of a viral consumer chatbot.
DeepSeek’s estimated revenue of $400 million to $500 million is comparable in scale but generated through a very different mix: API calls, enterprise contracts and, according to a US official cited in reporting we covered on July 23, engagements with Chinese military and intelligence customers. That profile, technically formidable, strategically sensitive, and controversial abroad — makes STAR a natural home.
The Liang commitment
The $3 billion personal check from Liang is unusual by any global standard. It exceeds the combined personal exposure of most Silicon Valley founders in their own AI ventures and dwarfs the founder stakes seen at Moonshot’s Yang Zhilin or MiniMax’s Yan Junjie. People familiar with the company describe Liang’s approach as a deliberate hedge against the pressures that come with outside capital: by anchoring the cap table himself, he retains the ability to reject commercialization detours that other investors might demand.
This is the same founder who, in a widely translated 2024 interview, argued that DeepSeek’s job was to advance frontier research rather than to build a super-app. Since then, DeepSeek has repeatedly forgone obvious monetization paths, no proprietary consumer chatbot buildout comparable to Kimi, no aggressive enterprise sales motion comparable to Z.ai’s push into government contracts, and has instead poured resources into open-weights model releases and infrastructure.
The forthcoming V4 generation, which retires the legacy deepseek-chat and deepseek-reasoner endpoints, extends that pattern: V4-Pro at 1.6 trillion parameters and V4-Flash priced at roughly one-hundredth of comparable US frontier models are moves engineered to reshape the global cost curve, not to maximize near-term margin.
Contrast with Moonshot and MiniMax
The strategic contrast with Moonshot is instructive. Moonshot’s Kimi K3 has drawn so much demand that the company had to suspend new subscriptions because of compute shortages, and it now faces a US government accusation, chronicled in our July 23 report, that it distilled Anthropic’s Fable model to build K3. Moonshot’s response has been to accelerate its Hong Kong IPO, capitalize on consumer momentum, and monetize while the window is open. Yang Zhilin’s company is, in effect, treating public markets as fuel for a commercial land grab.
MiniMax and Z.ai chose similar paths earlier this year. MiniMax leaned into its dominance in AI-generated video and character chat products; Z.ai, freshly capitalized, has been building out a one-gigawatt data center entirely on Chinese chips to serve enterprise and government customers. In each case the IPO was both a validation moment and a working-capital event.
Liang’s DeepSeek is playing a different game. By targeting a 2027 listing on STAR, he buys another eighteen months during which the company can iterate on architectures, release additional open-weights models, and cement its position as the technical reference point for Chinese AI without the quarterly reporting cycle that would inevitably tug it toward commercialization. The $71 billion valuation whispered in the market would make DeepSeek the most valuable of the “AI tigers” at listing, but Liang has told associates the number matters less than the runway.
What the IPO rush reveals
Taken together, the four listings — Z.ai and MiniMax already trading in Hong Kong, Moonshot filing within six months, and DeepSeek queued for STAR in 2027 — mark the maturation of China’s model-lab generation into public companies. It is a remarkably compressed cycle: barely three years separate DeepSeek’s founding from its planned IPO. The rush also reflects a strategic calculation in Beijing, where regulators have signaled they want strategic AI firms domiciled and listed at home rather than seeking capital abroad.
For Liang, the message he is sending to the market is that DeepSeek intends to be judged by a different yardstick. AGI, he has reportedly told his team, is the only goal. Commercialization can wait. The $3 billion he is personally putting behind that conviction ensures that, for now, no one else gets to change his mind.