Chinese AI developer Z.AI raised $5 billion from a Hong Kong share placement and a concurrent convertible bond sale, according to a stock exchange filing on Sunday. The company, formerly known as Zhipu AI, is racing to fund the computing infrastructure it needs to compete with US rivals like OpenAI and Anthropic.
Z.AI launched the sales on Friday. It offered 21.97 million new Hong Kong shares at HK$714 each, about $91, a 10 percent discount to Friday’s closing price of HK$793. That raised roughly $2 billion. Alongside it, the company sold 20.14 billion yuan, about $3 billion, of zero-coupon convertible bonds due September 2027. The yuan-denominated bonds will be settled in US dollars and were issued at 100.5 percent of face value.
The initial conversion price on the bonds is HK$892.50, a 25 percent premium to the placement price and 12.6 percent above Friday’s close. Z.AI can redeem all, but not part, of the bonds from February 18, 2027, if its shares trade at or above 130 percent of the conversion price for 20 out of 30 trading days. The placement and the bond sale are running at the same time but are not dependent on each other closing.
Where the money goes
About 60 percent of the net proceeds will fund research and development of next-generation models and the company’s fully self-training system, according to the filing. Another 15 percent goes to expansion plans, and the rest to optimizing the capital structure, replenishing working capital and general corporate purposes. Term sheets circulated on Friday also listed strategic investments and potential acquisitions among the intended uses.
The raise comes less than two months after Z.AI pulled in $4 billion from a follow-on share sale in July. The company went public in Hong Kong in January. Three large raises inside eight months is a pace that few listed companies anywhere can match, and it reflects how expensive frontier model training has become.
Chinese AI developers are spending heavily on computing infrastructure and talent as they try to close the gap with better-funded American labs. Rival MiniMax also listed in Hong Kong earlier this year, while Moonshot AI and DeepSeek are pursuing potential listings in Hong Kong and Shanghai respectively, Reuters has reported. The whole sector is repricing around one assumption: whoever secures the most compute over the next two years sets the terms of competition.
A crowded domestic field
Z.AI faces stiff competition at home. DeepSeek, Moonshot AI and MiniMax are all investing heavily in large language models, and each has its own angle: DeepSeek on efficient open models, Moonshot on long-context consumer assistants, MiniMax on multimodal products. Z.AI’s GLM model family is widely used by Chinese enterprises and developers, and the company positions itself as the closest thing China has to an OpenAI-style full-stack lab, with its own models, developer platform and consumer chatbot.
The bond structure is worth a note. Zero-coupon convertible bonds issued at a slight premium, with a negative-to-zero yield, let the company raise $3 billion at effectively no interest cost, with the real price paid later in dilution if the stock performs. The early redemption clause from February 2027 gives Z.AI a way to force conversion once shares rally, capping the debt’s lifetime.
| Deal | Size | Terms |
|---|---|---|
| Share placement | ~$2.0B | 21.97M shares at HK$714, 10% discount |
| Convertible bonds | ~$3.0B | Zero-coupon, due Sept 2027, conv. HK$892.50 |
| July follow-on | ~$4.0B | Secondary share sale |
The compute race behind the raise
Every large Chinese AI lab is constrained by the same thing: access to training compute. US export controls block the most advanced Nvidia chips from Chinese buyers, which pushes domestic labs toward Huawei’s Ascend processors and whatever Nvidia products Washington still permits. DeepSeek’s V4 models, for example, run on Huawei Ascend 950 hardware rather than Nvidia.
That makes capital even more important in China than in the US, because domestic chips are scarcer and often less efficient per dollar. A lab with $5 billion more cash can lock up domestic compute capacity, hire researchers away from rivals and fund the multi-month training runs that frontier models require.
The discount on the share placement, about 10 percent, is standard for Hong Kong top-ups but still stings existing holders. The stock closed Friday at HK$793, and new shares went out at HK$714. Investors who bought in January at the IPO have done well regardless; the July follow-on and this placement both found demand quickly, which says something about how Hong Kong money views the AI sector right now.
For Hong Kong’s market, the deal is another sign that Chinese tech issuers can raise serious money at home again after a long drought. For Z.AI’s American competitors, it is a reminder that the funding gap is closing from the other side. OpenAI and Anthropic still raise more in absolute terms, but a Chinese lab with $9 billion of fresh capital since July is no longer an underdog on resources. What remains uncertain is whether the compute those funds buy, largely domestic silicon under export-control constraints, translates into models that keep pace with the best US systems. That question will be answered by the next generation of releases, not by the size of anyone’s treasury.
There is also a financing-culture point here. Chinese issuers have learned to work the convertible bond market aggressively, and zero-coupon structures with forced-conversion clauses have become the default for growth companies that want cash now and dilution later. Z.AI’s terms mirror deals done by other Chinese tech names this year, and the speed with which both tranches priced suggests investors see the sector’s momentum, not just one company’s story.