A fundamental transformation in revenue structure is underway as Alibaba Cloud begins to emulate Google's strategy, building from chips to models and selling from computing power to "intelligence." According to a comprehensive 60-page report released by Citigroup Research on May 11, Alibaba Cloud is constructing a vertically integrated AI system spanning from chips to applications—from its self-developed T-Head semiconductor chips, to cloud infrastructure and AI platforms, to its core Model-as-a-Service (MaaS) offerings centered on the Qwen model. This comprehensive approach mirrors Google's full-stack AI strategy from TPU to Gemini. Consequently, Citigroup positions Alibaba as the preferred investment target in China's AI sector. In this report, analysts Alicia Yap and Nelson Cheung introduced the first systematic forecast for Alibaba Cloud's MaaS revenue and significantly raised their cloud revenue projections for fiscal years 2028 through 2031. The analysts predict that Alibaba Cloud's AI-related revenue will surge from RMB 240 billion in FY2026 to RMB 5,855 billion by FY2031, representing a compound annual growth rate (CAGR) of 90%. This would increase its share of total cloud revenue from 15% to 70%. The fastest-growing segment is the Model-as-a-Service (MaaS) business. Analysts forecast MaaS revenue will grow at a staggering 235% CAGR, skyrocketing from approximately RMB 10 billion in FY2026 to RMB 438.6 billion (approximately $62.6 billion USD) by FY2031. At that point, MaaS is projected to constitute 53% of Alibaba Cloud's total revenue. In other words, by 2031, over half of Alibaba Cloud's income is expected to come from "selling models" rather than "selling computing power." Citigroup maintains a "Buy" rating on Alibaba, with a price target of $205 for its American Depositary Receipts (ADR) and HK$204 for its Hong Kong shares, reiterating it as the "top pick for AI investment in China."
**Major Revenue Transformation: From 15% to 70%** The analysts meticulously reconstructed Alibaba Cloud's revenue segmentation, forecasting it across two main categories—AI and non-AI—and three sub-layers: Infrastructure-as-a-Service (IaaS), Platform-as-a-Service (PaaS), and Model-as-a-Service (MaaS). Overall, they predict Alibaba Cloud's consolidated total revenue will grow at a 39% CAGR from RMB 1,590 billion in FY2026 to RMB 8,334 billion in FY2031. Revenue from external customers is forecast to grow at a 44% CAGR to RMB 6,668 billion (approximately $95.2 billion USD), nearing the management's five-year target of $100 billion. Within the three sub-layers of AI-related revenue: * **MaaS**: 235% CAGR, reaching RMB 438.6 billion by FY2031, accounting for 53% of total cloud revenue. * **AI-IaaS**: 46% CAGR, reaching RMB 128.7 billion by FY2031, accounting for 15% of total cloud revenue. * **AI-PaaS**: 40% CAGR, reaching RMB 18.2 billion by FY2031, accounting for 2% of total cloud revenue. In contrast, non-AI cloud business revenue, while still projected to grow at a 13% CAGR to RMB 247.9 billion, will see its share of total revenue decline from 85% in FY2026 to 30% by FY2031. This shift signals a clear trend: traditional services like "selling servers, storage, and bandwidth" are giving way to "selling model calls, AI computing power, and intelligent services," with the latter growing nearly seven times faster.
**Full-Stack Strategy: The Rationale for the "Chinese Google" Analogy** The report references NVIDIA CEO Jensen Huang's "five-layer AI cake" theory, which outlines the complete vertical chain of the AI industry: Energy, Chips, Systems/Infrastructure, Models, and Applications. Excluding the energy layer, Alibaba has established a presence in the remaining four. * **Layer 1: Chips (T-Head)**. Founded in 2018, T-Head is Alibaba's wholly-owned semiconductor subsidiary. Its flagship AI chip, the Zhenwu 810E, is reported by media like SCMP to be "comparable to NVIDIA's H20," with cumulative shipments reaching 470,000 units. Over 60% of these serve more than 400 external enterprise clients, including the State Grid, Xpeng Motors, and the Chinese Academy of Sciences. In March 2026, T-Head launched the XuanTie C950 CPU based on RISC-V architecture, designed for Agentic AI. In April 2026, it released its first 400G smart network interface card, the Panmai 920, which is now in mass production and slated for deployment in Alibaba Cloud data centers. * **Layer 2: Infrastructure (IaaS)**. Alibaba Cloud currently operates 94 availability zones across 29 regions globally, with 13 in mainland China and 15 overseas. It plans to add new data centers in Brazil, France, and the Netherlands in 2026. According to 2025 data, Alibaba Cloud holds a 33% share of China's public cloud IaaS market (Gartner data), ranking first. Its share in the Asia-Pacific region is 22.5%, placing it fourth globally (7.7%). * **Layers 3 & 4: Platform (PaaS) & Model-as-a-Service (MaaS)**. Alibaba Cloud's Model Studio (Bailian Platform) is the core vehicle for its MaaS business. This platform integrates the full Qwen model series and over 200 third-party mainstream large models, including DeepSeek, Kimi, and MiniMax. It currently boasts over one million enterprise and individual users and has facilitated the creation of more than 800,000 AI agents. Analysts note that, as disclosed in the FY3Q26 earnings call, Alibaba's management reported a six-fold increase in token consumption on the MaaS platform over the past three months, with MaaS becoming the largest revenue category for the Cloud Intelligence Group. This complete chain from chips to models to applications forms the core logic behind the "Chinese Google" analogy—Google is also a full-stack player, moving from its self-developed TPU chips, through GCP cloud infrastructure, to the Gemini large model, and finally into applications like search, advertising, and enterprise services.
**Token Economics: The Underlying Driver of High MaaS Growth** The essence of MaaS is "AI call services charged by usage"—enterprises call large models via APIs and pay based on the number of tokens consumed. A token can be understood as the smallest unit of information processed by AI, with text, images, and audio all convertible into tokens for computation. This market is experiencing explosive growth. Data cited from China's National Data Administration in a March 30, 2026 report indicates that by March 2026, China's daily AI token call volume had exceeded 140 trillion, a more than 1,400-fold increase from 100 billion in early 2024. More notably, China's weekly AI model token call volume surpassed that of the United States for the first time in February 2026, making it the world's largest token consumer. Enterprise-side growth is equally rapid. Data from Frost & Sullivan and the China Academy of Information and Communications Technology cited in a September 2025 report shows that the average daily token consumption of enterprise large models in China reached 10.2 trillion in the first half of 2025, a 363% increase from the second half of 2024. Among these, Alibaba's Qwen model ranked first with a 17.7% market share in the general-purpose large models most chosen by Chinese enterprises, leading ByteDance's Doubao (14.1%) and DeepSeek (10.3%). The explosion in token consumption has directly contributed to the restoration of pricing power for MaaS. Alibaba Cloud raised prices for its AI computing power products by 5% to 34% on April 18, 2026. Analysts view this price hike as a positive signal of strong demand and improved pricing capability.
**Capital Expenditure: RMB 822.9 Billion Over Six Years, Betting on a "Once-in-a-Generation Opportunity"** The analysts predict that Alibaba's capital expenditure will remain at an annual level of RMB 130-140 billion. The cumulative capital expenditure from FY2026 to FY2031 is projected to reach RMB 822.9 billion, averaging 32% of cloud revenue. This scale of investment has historical context. In February 2025, Alibaba first disclosed a three-year capital expenditure plan of RMB 380 billion, labeling AI a "once-in-a-generation opportunity." Alibaba Group CEO Eddie Wu later stated at the 2025 Apsara Conference that the company would expand its global data center capacity tenfold by 2032 to provide infrastructure support for the "transition from AGI to ASI (Artificial Superintelligence)." During the FY2Q26 (November 2025) earnings call, management further indicated that, given supply chain constraints, AI demand would continue to outpace computing power supply for the next 2-3 years, potentially leading to the early completion of the original three-year RMB 380 billion capital expenditure plan. Analysts calculate that the proportion of capital expenditure to cloud revenue will gradually decline from 83% in FY2026 to 17% by FY2031. This suggests that as economies of scale take effect, marginal investment efficiency will significantly improve, creating potential for increased cloud business profitability.
**Qwen: The Moat from Open Source to Commercialization** The Qwen (Tongyi Qianwen) series is the core product pillar of Alibaba Cloud's MaaS business. Since releasing its first open-source model in August 2023, Qwen has evolved to its third generation and beyond, covering multiple sub-series including text, reasoning, code, vision-language, audio, video, and mathematics. By the end of 2025, the cumulative downloads of the Qwen series models on Hugging Face had exceeded 600 million, with over 170,000 derivative models based on Qwen, establishing it as one of the world's most popular open-source models. Following its launch on April 2, 2026, the latest Qwen3.6-Plus rapidly attracted a large user base with an initial "free" strategy and, according to a Towards AI report, set a historical record on the OpenRouter platform by processing over one trillion tokens in a single day. According to Artificial Analysis data, Qwen3.6 Max Preview ranks 6th on the AA Intelligence Index with a score of 52, gradually narrowing the gap with OpenAI's GPT-5.5 (score 60). At the application layer, Alibaba has integrated Qwen with its ecosystem apps like Taobao, Alipay, and Amap to advance commercial implementations such as "AI shopping." A May 10, 2026 report indicated that Alibaba is deeply integrating Qwen into Taobao, allowing users to browse, compare prices, and shop through conversations with AI. Qwen will have access to over 4 billion products on Taobao and Tmall.
**Competitive Landscape: Alibaba Cloud's Position in the "War of the Gods"** While Alibaba Cloud leads in overall share of China's cloud market, it faces strong competition from ByteDance in the MaaS segment specifically. According to IDC data, in China's public cloud LLM service market (measured by call volume), ByteDance's Volcano Engine ranks first with a 49.2% share, followed by Alibaba Cloud at 27%, and Baidu AI Cloud at 17%. However, analysts point out that Alibaba Cloud's competitive advantage lies in its "full-stack" capability rather than any single component. Its strengths include cost reduction through self-developed chips, a vast Alibaba ecosystem (where internal users from Taobao, Alipay, etc., provide a natural "training set" and early adopters), and the external developer stickiness fostered by the Qwen open-source ecosystem. This comprehensive competitive barrier is difficult for any single competitor to replicate. Eddie Wu has previously stated that there may only be 5 to 6 global super cloud platforms capable of providing large-scale AI computing services. Analysts believe that "Alibaba Cloud, as a world-leading full-stack AI service provider, is well-positioned to capture this long-term growth opportunity."
**Alibaba Cloud's Development Trajectory and Current Phase** Founded in 2009, Alibaba Cloud initially served as infrastructure for Alibaba's internal e-commerce system. After a period of "rapid growth" (FY2015-2018) with annual growth rates of 64%-138%, it entered a "difficult period" (FY2022-2024, with growth slowing to 3%-4%) around 2021 due to factors like ByteDance migrating its business and intensified domestic price competition. Since FY2025, Alibaba Cloud has entered a "re-acceleration phase." Analyst data shows that cloud revenue grew 36% year-over-year in FY3Q26, with AI-related revenue achieving triple-digit growth for ten consecutive quarters. Currently, AI-related revenue accounts for approximately 20% of total cloud revenue (as disclosed by management), a proportion analysts forecast will rapidly increase to 70% within five years. Based on management statements during the FY3Q26 earnings call, Alibaba has set a goal of "achieving combined cloud and AI external revenue exceeding $100 billion over the next five years." MaaS is expected to be the primary driver of this target and will become the largest revenue category by that time. Analysts define the current stage as the "AI high-growth period (FY2027-2028)" and note that the market remains skeptical of a revenue CAGR exceeding 40%. However, through top-down and bottom-up comprehensive analysis, they believe this target is achievable.
Comments