Cloud Giants Undergo Major Role Shift: From Compute Power Providers to AI Distribution Hubs

AI computing power demand is fundamentally reshaping the growth logic of hyperscale cloud computing vendors. According to Wind Chase Trading Desk, a UBS report points out that the market still regards the three major cloud giants as simple AI computing power providers, but their roles have quietly transformed—from selling GPU computing power to upgrading into core distribution platforms for cutting-edge models such as OpenAI and Anthropic, reaching enterprise customers. The incremental revenue brought by this structural change has not yet been fully priced by the market. In the first quarter of 2026, the combined revenue of the three cloud giants reached $84.8 billion, up 39% year-on-year, with order backlogs at $2.1 trillion, surging 184% year-on-year. This growth rate and order volume have surpassed what can be explained by simple "computing power resale," suggesting that enterprise clients are paying higher premiums as "platform access fees" to connect to advanced AI models. The market previously worried that the cloud giants’ skyrocketing capital expenditure (expected to reach $673 billion in 2026) to support computing power expansion would erode profits. However, over the past six quarters, their combined operating profit margins have remained stable at 36%-37%. This indicates that high-margin platform service revenue is becoming the main profit pillar, rather than low-margin computing power resale. Yet the market still prices them under the "computing power supplier" framework, with current price-to-earnings ratios for 2027 only at 19-27 times. For a sector with overall revenue growth around 40%, and income structure shifting towards high-margin platform services, this valuation clearly underestimates the value revaluation potential brought by the role transformation. **Three-Layer Revenue Model: How Cloud Giants Transition from “Selling Computing Power” to “Selling Platforms”** Wall Street is accustomed to seeing hyperscale cloud providers as “resale channels” for Nvidia GPUs—that is, buying chips and then selling computing power to clients for a margin. UBS notes that this view overlooks the massive incremental revenue they generate as core distribution platforms for OpenAI and Anthropic. Google Cloud’s performance is a typical example. It is clearly insufficient to explain its 35-percentage-point leap in growth over the past year solely by "computing capacity expansion." Cloud giants are replicating the successful path from 10 to 15 years ago—extending from underlying infrastructure upwards to become "model distribution hubs" of the AI era. The distribution platform role has three main revenue sources: **First layer: Model API services.** Enterprises do not need to build their own AI models but can directly invoke them through API interfaces provided by cloud vendors. Microsoft's Azure OpenAI API service has grown into a multi-billion-dollar annual business. Amazon AWS Bedrock resembles a "model marketplace," where clients can choose and invoke various models; it now has 125,000 customers, covering 80% of Fortune 100 companies, with client spending up 170% quarter-on-quarter. **Second layer: Distribution channel for third-party AI products.** OpenAI and Anthropic are promoting their own AI coding tools, but large enterprises prefer to purchase via existing Microsoft Azure or AWS procurement agreements for better security compliance and bargaining power. Cloud vendors collect platform service fees or drive additional consumption of underlying computing resources in this process. **Third layer: Sales of self-developed AI products.** Google’s Gemini API and enterprise subscriptions are estimated to have annual revenues around $2 billion, covering more than 11 million seats. Microsoft GitHub Copilot’s enterprise subscriptions doubled year-on-year. All this revenue is directly accounted into their cloud business segments. In addition, Google has taken a differentiated strategy—directly selling TPU chips (an alternative to Nvidia GPUs) to specific external clients. The related revenue is recognized upon delivery, most of which is expected to be booked in 2027. This is likely to be one of the key drivers behind Google Cloud’s $220 billion quarter-on-quarter surge in order backlog in Q1. **Valuation Expansion Far From Peaking** UBS believes that the current round of valuation correction for hyperscale cloud providers is far from over. The market has **two obvious cognitive biases**: firstly, **underestimating the growth slope of AI computing power demand**; secondly, and more crucially, **severely underestimating the upside revenue potential of AWS and Azure as full-platform distribution channels for OpenAI and Anthropic’s entire suite of AI products**. In terms of valuation, based on expected earnings in 2027, Amazon, Google, Microsoft, and Oracle correspond to price-to-earnings ratios of 24x, 23x, 19x, and 27x respectively. For a sector with overall revenue growth around 40%, robust growth momentum, and resilient profit margins, a valuation range of 19 to 27 times is relatively reasonable. ~~~~~~~~~~~~~~~~~~~~~~~~ The above content is from [Wind Chase Trading Desk](https://mp.weixin.qq.com/s/uua05g5qk-N2J7h91pyqxQ). For a more detailed analysis, including real-time interpretation and frontline research, please join [Wind Chase Trading Desk ▪ Annual Membership](https://wallstreetcn.com/shop/item/1000309).  Risk Warning & Disclaimer The market carries risks, and investment must be prudent. This article does not constitute personal investment advice and does not take into account specific investment goals, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investment decisions are solely at your own responsibility.