Inside the tech industry’s most costly quarter: The rise of power-hungry clouds

Based on the latest financial reports, it is evident that the demand for hyperscale clouds is not the issue; instead, they are facing challenges related to the limitations imposed by the laws of physics. To tap into what appears to be an insatiable demand, access to megawatts, liquid-cooled racks, and a reliable supply chain of graphics processing units, mainly sourced from Nvidia Corp. and Taiwan Semiconductor Manufacturing Co., are crucial.

In the last quarter, industry giants Amazon Web Services Inc., Microsoft Corp., and Google LLC collectively spent approximately $63 billion on capital expenditures, setting new records. However, even this substantial investment cannot fully meet the increasing demands driven by AI models aiming for trillion-parameter scales. The crucial factor influencing the growth of hyperscalers is no longer the revenue potential but the availability of power. The ability to secure sustainable energy sources, manage GPU allocation efficiently, and translate these infrastructure investments into lucrative AI services will determine the success of providers in the upcoming cycle.

We are now witnessing a new era where market share is directly tied to the availability of megawatts. Recent earnings reports from Microsoft, Amazon, and Alphabet provide valuable insights into the current state of the public cloud sector. Despite the challenging macroeconomic environment, the three U.S. hyperscalers achieved a remarkable 26% year-on-year growth, generating over $59 billion in revenue from infrastructure-as-a-service and platform-as-a-service offerings.

Three key themes emerge from the latest data:

1. Power and GPU scarcity remain critical factors: Executives acknowledge the significance of securing reliable sources of energy, liquid-cooling capabilities, and a stable supply of Nvidia GPUs. The focus is not on meeting demand but on addressing the limitations posed by power availability.

2. Margins are a point of interest: While AWS continues to lead with an operating margin of around 40%, Google is gradually improving from a negative base. Microsoft’s Intelligent Cloud business reported a low 40% operating margin, indicating potential challenges related to Azure.

3. Accelerated CapEx spending: The hyperscalers invested a combined $63 billion in infrastructure last quarter, signaling a renewed focus on enhancing capabilities. However, the race to upgrade infrastructure must align with AI revenue growth to ensure sustainable success in the long run.

Normalized comparisons across Microsoft, AWS, and Google reveal varying definitions of cloud revenue and core services, making it challenging to evaluate performance accurately. Microsoft’s all-in public cloud revenue stands at approximately $42 billion, encompassing Azure, Office 365, Dynamics, LinkedIn, and other industry-specific services. AWS reported $29.3 billion in revenue, while Google’s figure of $12.3 billion includes Workspace offerings along with infrastructure services.

For essential core services, AWS’s revenue remains consistent at $29.3 billion, Azure’s figure is estimated at around $23 billion, and GCP’s core revenue is projected to reach $6.1-6.2 billion for the quarter. Operating margins for AWS, Microsoft’s Intelligent Cloud, and Google Cloud are approximately 39.2%, low 40%, and high 17%, respectively. The trio’s CapEx spending underscores their commitment to infrastructure expansion and technological advancements to meet growing demands in the cloud sector.