Z.AI raises about $5 billion through Hong Kong shares and yuan convertible bonds
The AMW Read
Same ~$5B HK equity-plus-convertible package already noted in prior coverage; peer MiniMax/Moonshot framing is incremental, but the explicit $5B public raise is a structural capital-cycle signal for CN foundation-model labs.
Z.AI raises about $5 billion through Hong Kong shares and yuan convertible bonds
Beijing-based foundation-model company Z.AI Co Ltd filed with the Hong Kong Exchanges to place about 21.97 million new shares at HK$714 ($91.05) each — a 10 percent discount to Friday’s close of HK$793 — for roughly $2 billion, alongside about 20.14 billion yuan ($3 billion) of zero-coupon convertible bonds due September 2027 that settle in US dollars. The company said about 60 percent of net proceeds will go to R&D on next-generation GLM foundation models and its Fully Self Training system, plus deployment and upgrades of large-scale training and production-inference compute. Per the AI Market Watch index, Z.ai was previously tracked at $1.5B in total funding (coverage of ~5,000 companies, not a census), so this public-markets package sharply expands its capital base beyond prior private raises.
The filing sits inside a broader Chinese frontier-lab financing wave: MiniMax is seeking about HK$16.04 billion via a similar share-and-bond structure, and Moonshot AI recently closed a larger-than-expected $3.5 billion round at a $35 billion valuation after Kimi K3. Z.AI’s dual equity-and-bond raise on a Hong Kong listing shows Chinese LLM labs are using listed-market capital — not only private rounds — to underwrite the shift from model invocation toward longer-horizon and end-to-end task delivery, where training clusters, inference systems, data, and talent remain the principal cost items.
For builders and investors, the concrete takeaway is funding durability through at least the September 2027 bond maturity: proceeds are explicitly earmarked for GLM-class model R&D and compute scale-up, which sustains price and capability pressure on anyone competing in open-weight reasoning, coding, or agentic workloads against well-capitalized Chinese foundation-model labs.
