CMSC International: Z.AI (02513) completes approximately US$5 billion strategic financing; maintains "Overweight" rating and target price of HK$1,600

date
17:04 17/09/2026
avatar
GMT Eight
At the September 16 investor call, company management stated that the year-end ARR guidance was raised from US$2.4 billion to US$3 billion, and that current all-business ARR (calculated on a monthly basis) has reached US$1.8 billion, up further from US$1.6 billion two weeks earlier.
CMSC International released a research report stating that Z.AI (02513) recently announced the completion of approximately US$5 billion in strategic financing, with proceeds mainly directed toward next-generation GLM foundation model R&D, a fully self-trained system, and computing infrastructure. The bank maintains its "Overweight" rating with a target price of HK$1,600. The bank cited the company's management exchange meeting on September 16, noting that following this capacity expansion, computing power has begun to reach scale, and short-term computing supply is no longer the primary constraint on revenue growth; the company has signed revenue-sharing agreements with leading domestic and international cloud service providers, with related revenue to be recognized starting in October; year-end ARR guidance was raised from US$2.4 billion to US$3 billion, with current all-business ARR (on a monthly basis) already reaching US$1.8 billion, up further from US$1.6 billion two weeks ago; cumulative order value for co-work's first scenario exceeded RMB 1 billion within one month of launch. The bank noted that this round of financing anchors computing infrastructure and a fully self-trained system, and combined with approximately RMB 28.5 billion in net cash at the end of the first half of this year, the computing reserve has increased significantly. The newly added computing power could theoretically support over US$6 billion in inference revenue, with the actual outcome depending on utilization and discount rates. In the first half of this year, the computing multiplier increased 14-fold year-over-year; since incorporating domestic chips into primary inference computing power at the beginning of the year, per-token inference cost has fallen 80% compared with the start of the year.