Zhipu stated that on September 11, it completed a $5 billion equity‑debt financing round and raised its ARR guidance for the end of 2026 from $2.4 billion to $3.0 billion, an increase of approximately 25%. The proceeds will primarily fund capacity expansion, and the company has already signed revenue‑sharing agreements with several leading domestic and international cloud service providers. Its GLM series models are deployed as hosted APIs, while CoWork orders have exceeded RMB 1 billion, with cybersecurity applications being rolled out first.
$Z.AI (02513.HK)$The company is accelerating the expansion of its computing capacity and advancing commercialization. On September 16, according to information from an iZhiPu analyst briefing, the company closed a $5 billion financing round—combining equity and debt—on September 11, and raised its full-year Annual Recurring Revenue (ARR) guidance for 2026 from $2.4 billion to $3.0 billion, representing an increase of approximately 25%. As of the first half of September, the company's monthly ARR across all business lines had reached $1.8 billion.
This round of financing will be primarily used to expand computing power. The company stated that, following the completion of the funding, its total computing‑power capacity has risen to among the top ranks of independent large‑model providers in China, meaning that, in the short term, computing‑power supply will no longer be the primary constraint on revenue growth. Previously, constrained by computing resources, Coding Plan had temporarily imposed sales restrictions; after fully reopening from late July to early August, its sales surged by more than 15 times.
On the commercialization front, the company has signed revenue-sharing agreements with several leading domestic and international cloud service providers, with related revenues expected to begin recognition starting in October 2026. Meanwhile, businesses such as Coding, cloud services, and CoWork are all progressing; notably, Z Code has surpassed 2 million users, and within one month of the GLM‑5.3 release, industry‑specific orders for work‑related use cases exceeded RMB 1 billion.
At the briefing, the company also disclosed updates on its computing‑power investments, partnerships with cloud service providers, and progress in its CoWork business, while providing an overview of preparations for an A‑share listing. It is important to note that the $3 billion ARR represents management's annualized order‑revenue guidance and does not fully align with reported revenue; actual revenue recognition will depend on execution and applicable accounting standards.
Computing power capacity expansion is advancing, and the Coding business has resumed growth.
Over the past year, computing power capacity had been a major bottleneck to the company's business expansion. Following the release of GLM‑5.5 in February 2026, demand for the model surged nearly tenfold compared to the beginning of the year, rapidly depleting the company's existing compute resources. As a result, the flagship product, Coding Plan, was temporarily subject to sales restrictions, with service provided exclusively through a daily quota of 500 slots.
Between February and March, the company enhanced its computing‑power utilization efficiency through an "all‑in‑infra" strategy, boosting overall efficiency to roughly 2.3 times its previous level. However, as demand for model inference continued to grow, supply remained insufficient to fully meet needs. Following the release of GLM‑5.2 in June, overseas inference demand surged further; at the time, the company could only handle about 10% of the traffic, with the remaining demand largely being accommodated by third‑party deployments of open‑source models.
Following a $4 billion refinancing in July, the company began expanding its computing‑power capacity. From late July to early August, the Coding Plan, which had been suspended for six months, resumed full availability, with sales surging by more than 15-fold after its relaunch. The company stated that, once this $5 billion financing is completed, computing‑power supply will no longer be the primary constraint on revenue growth in the near term.
According to estimates disclosed at the company's investor conference, a $5 billion funding round corresponds to approximately RMB 30 billion in computing‑power investment, enabling the procurement of nearly 100,000 AI accelerator cards—40% for training and R&D, and 60% for inference workloads. The company expects that, as its computing‑power capacity expands, previously supply‑constrained businesses such as Coding will be able to further scale their services.
Revenue from cloud vendor revenue sharing will be recognized starting in October.
In terms of commercialization,$Z.AI (02513.HK)$The company has entered into revenue-sharing agreements with several leading domestic and international cloud service providers, deploying the GLM series of open-source models as hosted APIs on their cloud platforms. Revenue is shared between the parties in accordance with the agreed-upon ratio, with related revenues expected to begin recognition starting in October 2026.
The aforementioned collaboration is aligned with the commercial arrangements following the open-sourcing of GLM‑5.3. The company previously disclosed that cloud service platforms operating MaaS businesses with annual revenues exceeding USD 10 billion must obtain security clearance and commercial authorization to use its cutting-edge models. At present, the company has not yet disclosed the names of specific partners, the revenue scale of individual platforms, or the revenue-sharing ratios.
The company stated that partnering with cloud providers will, on the one hand, broaden model distribution channels, and on the other, lower the barrier to entry for users while reducing the direct consumption of its own computing resources for certain API calls. As these services are gradually rolled out, revenue sharing with cloud providers is expected to become one of the new sources of income in the fourth quarter.
CoWork's order value has exceeded RMB 1 billion, with cybersecurity being the first to be implemented.
CoWork is$Z.AI (02513.HK)$Another key business initiative unveiled at this exchange event: the company stated that within one month of the GLM‑5.3 launch, order value in the "to work" use case has exceeded RMB 1 billion, and the business has entered the production‑scale deployment phase.
Cybersecurity is one of the earliest vertical domains to achieve large-scale deployment. To date, more than 100 cybersecurity companies have integrated the GLM model into their security products and business workflows. At the product level, GLM‑5.3 primarily handles tasks such as vulnerability discovery and complex code auditing, while GLM‑5.3 Flash covers high-frequency use cases like batch scanning and alert analysis.
The company believes that cybersecurity scenarios are well-suited for integrating AI models into real-world business workflows, as they typically feature easily verifiable outcomes and relatively straightforward value quantification. During the briefing, the company disclosed that GLM‑5.3 achieved a performance score of approximately 85% on Cyber Games, and during red-team testing, it identified over 1,500 high‑severity vulnerabilities.
The company stated that CoWork's business model will be further expanded to cover specialized sectors such as finance, law, industrial manufacturing, life sciences, and mathematics. Meanwhile, the company completed the guidance and acceptance process for its A-share listing in June and is currently in the pre‑filing stage. It has also announced that, from September 30 onward and until its listing on the STAR Market, it will no longer raise equity‑based capital from new investors.
Edited by melody