Morgan Stanley: Z.AI (02513) raises year-end ARR guidance to US$3 billion; reiterates "Overweight" rating

Morgan Stanley issued a research report stating that Z.AI (02513) has reached US$1.8 billion in annualized recurring revenue (ARR) as of mid-September, and has raised its year-end ARR guidance from US$2.4 billion to US$3 billion, reiterating an "Overweight" rating with a target price of HK$1,800. The bank noted that management's framework indicates approximately US$5 billion in financing can support about 94,000 GPUs, of which roughly 40% is for training and about 60% for inference; approximately 56,000 inference GPUs imply a theoretical annual revenue ceiling of about RMB 41 billion, without yet factoring in utilization rates and actual pricing realization, with theoretical inference gross margins reaching up to about 80%. Easing compute bottlenecks is expected to unlock the next phase of growth, while cloud partnerships can also improve capital efficiency and global distribution, with revenue sharing expected to contribute starting from October. On the R&D side, the training allocation can support an annualized training expenditure of approximately RMB 3 billion on a four-year cumulative basis, sufficient to support multiple rounds of GLM-5.3-level or larger-scale training. The bank also noted that Cowork is expanding beyond coding, with cybersecurity as the first scaled vertical, where over 100 cybersecurity companies have integrated GLM models, and orders exceeded RMB 1 billion within one month of GLM-5.3's launch. The ARR upgrade is mainly driven by three major forces: coding, cloud partnerships, and Cowork. The key discussion has shifted from whether demand exists to how quickly incremental compute, partner distribution, and vertical workflow monetization can recognized revenue and sustainable gross margins.