Guosheng: KINGSOFT CLOUD (03896) is given a "buy" rating, with growth driven by both ecology and platform.

Guosheng released a research report stating that it is expected that KINGSOFT CLOUD (03896) will have operating revenues of 12.6 billion, 15.5 billion, and 19.4 billion RMB in 2026, 2027, and 2028 respectively, with year-on-year growth of 32%, 23%, and 25%. Due to the high capital expenditure and heavy asset nature of the cloud computing industry, the company is valued using the P/S method. Taking into account the high growth nature of the company's AI cloud business, its pure cloud business structure, and its neutral third-party positioning, and referencing the valuation level of comparable companies, an EPS of 2.5x is given for 2026, corresponding to a target price of 8 HKD/15 USD. Initial coverage is given a "buy" rating. Guosheng's main points are as follows: KINGSOFT CLOUD is the only strategic cloud platform in the Xiaomi-Kingsoft ecosystem Listed on the US stock market in 2020 and on the Hong Kong stock market in 2022, KINGSOFT CLOUD is positioned as a full-stack cloud service provider covering the IaaS, PaaS, and MaaS chain, focusing on vertical fields such as the internet, public services, digital health, and finance. The company is closely bound to the KINGSOFT and Xiaomi ecosystems, with affiliated cloud service ceiling amounts of 4.6 billion and 6.8 billion RMB in 2026 and 2027 respectively, forming a foundation of performance. In response to the AI industry wave, the company strategically upgraded its StarFlow platform from a resource management tool to an AI training and inference integrated platform, supporting nearly 40 mainstream models and providing MaaS services. In the first quarter of 2026, revenue from intelligent computing accounted for over 50% of public cloud revenue, and the company is transitioning from a traditional cloud provider to a full-stack AI service provider. AI has driven the global cloud market into a new era of intelligent computing, with a growing divide between internet cloud and telecom operator cloud camps 1) In terms of market size, the global cloud computing market reached 692.9 billion USD in 2024, with China reaching 828.8 billion RMB. The penetration of AI is driving the expansion of cloud services from the traditional IaaS-PaaS-SaaS boundary to the intelligent computing system (AI IaaS, AI PaaS, MaaS, AI SaaS). The structure of computing power is accelerating towards generative AI, with the proportion of inference scenarios continuing to rise. 2) In terms of market pattern, internet cloud providers, leveraging AI-native genes and full-stack productization capabilities driven by self-use, have returned to a high-growth trajectory. Leading providers such as Alibaba Cloud, Baidu Cloud, etc., have significantly enhanced their pricing power. Telecom operators rely on algorithm-network fusion and enterprise-government resource barriers to maintain the foundation, while their market share continues to be squeezed. Looking ahead, internet cloud providers will upgrade towards multi-token sales, while telecom operator clouds need to focus on the strategic functions of domestic computing power ecosystems and a national integrated computing power network, finding a balance for high-quality development in vertical industry customization and marketization reforms. The explosive demand for AI computing power within the ecosystem, combined with the strategic advancement of the StarFlow platform, constitutes the core growth drivers for KINGSOFT CLOUD On the demand side, Xiaomi Group's 60 billion RMB AI investment over the next three years, MiMo series model iterations, and Miclaw Agent landing will drive continuous high demand for computing power throughout the training, inference, and algorithm research chain. At the same time, the customer structure is shifting from ecosystem dependence to external internet, autonomous driving, AI for Science, and other fields. On the supply side, the StarFlow platform has completed its strategic upgrade from a resource management tool to an all-in-one AI training and inference platform. With the support of solid intelligence full-link simulations and differentiated capabilities such as MaaS unified management of multiple models, the company is pushing its transition from a traditional IaaS computing power provider to a full-stack AI service provider. The integrated training and inference platform and MaaS service possess higher customer stickiness and additional value, and as AI applications mature, the StarFlow platform is expected to become the core engine for sustained growth in intelligent cloud revenue. Risks: Risks of ecosystem demand falling short of expectations, risks of downstream AI applications falling short of expectations, risks of intense industry competition, risks of calculation errors.