Six months after its Hong Kong debut, Zhipu AI wants a Shanghai listing that would raise about $2.2 billion and ask investors to believe a 2,000 percent stock surge still has room to breathe.
Zhipu AI did the hard thing first. It listed in Hong Kong on January 8, 2026, raised HK$4.35 billion at HK$116.20 a share, and became the first major large language model developer to trade publicly. Then the harder thing arrived: explaining why a company with less than 1 billion yuan in 2025 revenue should be treated by the market like China's local answer to OpenAI.
By June 22, the stock had touched HK$2,980 intraday, after a 42 percent single-day move that pushed its market value past HK$1 trillion. From the IPO price to that peak, the gain is roughly 2,400 percent. You don't need a spreadsheet to see the problem. A rally that large stops being only a vote on product quality. It becomes a vote on how much future investors are willing to pay for before the company has earned it.
The latest spark was GLM-5.2, Zhipu's open-source flagship model released on June 13. The company says the model uses a mixture-of-experts architecture with 744 billion total parameters and 40 billion active, carries a one-million-token context window, and ships under the MIT license without regional restrictions. Business Insider reported this week that GLM-5.2 has drawn attention from Silicon Valley developers for long coding tasks and agentic workflows, the same territory where Anthropic's Claude Fable 5 has become the model to beat.
The benchmark detail matters because Zhipu isn't selling a vague AI dream. Independent rankings have placed GLM-5.2 second globally on the Code Arena front-end development benchmark, behind Claude Fable 5. JPMorgan raised its 2026 to 2030 revenue forecasts for Zhipu by 7 percent to 16 percent after the launch, citing stronger pricing power in China's enterprise AI market. Two licensing deals with major Chinese banks and a state-owned cloud provider were disclosed shortly after. That is the part bulls can point to without blushing.
Now Zhipu wants to do it again, this time in Shanghai. On June 1, the company announced plans for an A-share listing on the STAR Market, targeting 15 billion yuan, about $2.2 billion, through a new issuance equal to 2 percent to 8 percent of post-issuance share capital. Guotai Haitong Securities is serving as the tutoring institution after the application cleared the STAR Market's acceptance stage on June 17. The use of proceeds is unusually plain: 12 billion yuan for AI model development, 2 billion yuan for the MaaS platform, and 1 billion yuan for working capital.
Here's the thing: the business is growing, but the valuation has sprinted ahead of it. Caixin Global reported that Zhipu's 2025 revenue was 724 million yuan, up 132 percent year on year, while its adjusted net loss reached 3.18 billion yuan. In the first quarter of 2026, its MaaS API business was running at an annualized rate of about $250 million. That's real demand. At a HK$1 trillion market cap, though, the stock is trading above 1,200 times trailing revenue. Even JPMorgan's most optimistic case puts the 2027 price-to-sales ratio just above 100 times.
Frankly, that only works if you believe Zhipu becomes one of the few durable platforms in Chinese enterprise AI. If you don't, the math is brittle. Baidu, ByteDance, Moonshot AI and other well-funded Chinese players aren't politely leaving the market open. They have distribution, cash, engineers and, in some cases, cloud businesses that can absorb margin pressure for far longer than a newly listed company can.
The stock's reaction to the Shanghai plan told you investors understand the tension. Zhipu shares fell on the day the A-share proposal was announced. Secondary issuance is dilutive, and the people who bought at HK$200 or HK$500 didn't buy because they wanted a long lecture on funding model development. They bought because AI momentum was working. The Hong Kong IPO's cornerstone investors face their own decision soon, with the six-month lock-up expiring on July 8, 2026.
The Shanghai listing would test a different audience. Hong Kong gave international investors a clean proxy for Chinese AI at a time when US-listed alternatives were scarce. A STAR Market listing would put Zhipu in front of domestic funds and retail investors who have watched Nvidia's US run and want a local champion. As Caixin Global reported, the deal would also make Zhipu the first Hong Kong-listed AI company to complete a full H-share to A-share dual listing, with Beijing's securities regulator signaling support for the structure.
Zhipu has earned more respect than the usual story stock. Its GLM models have benchmark credibility, its enterprise customers include state-linked clients, and its API business is showing genuine revenue traction. But you still have to separate the company from the price. One can be good while the other is already asking too much.
If the $2.2 billion raise clears, Zhipu will have more money for model training, infrastructure and the MaaS platform. That gives it runway. It doesn't settle the larger question Shanghai investors are about to answer: whether Zhipu is being valued as China's next AI platform, or as if it has already won that fight.
Also read: MoEngage acquires Aampe to put a dedicated AI agent behind every single customer • Superhuman bets on trust as it buys AI detection startup GPTZero for its 19 million users and $30M ARR • Cerebras Systems beat its first earnings bar and still watched its stock fall 10 percent
