The global cloud-infrastructure market is one of the most concentrated and consequential markets in modern technology. Three companies — Amazon Web Services, Microsoft Azure, and Google Cloud Platform — hold roughly two-thirds of the market, and the concentration has only tightened over the past five years. Understanding how the three compare, and where each fits in the broader ecosystem, is the entry point to any serious view of the sector.

The scale

Global cloud infrastructure spending crossed $300 billion in annualised run-rate in 2026, up roughly 25% year-over-year. Of that, AWS holds approximately 32% share, Azure approximately 24%, and GCP approximately 11%. The remaining share is spread across Alibaba Cloud, Oracle Cloud, IBM Cloud, and dozens of smaller regional and specialty providers.

The three hyperscalers each generate revenue in the tens of billions of dollars per quarter from their cloud businesses. Their operating margins on these businesses have expanded from mid-single-digits a decade ago to above 30% today — a scale-driven margin expansion that mirrors the trajectory of the software-as-a-service business a decade earlier.

AWS: the incumbent

Amazon Web Services is the oldest of the three, launched publicly in 2006. Its early lead was enormous — for the entirety of the 2010s, AWS held more market share than the next two competitors combined. Its lead has narrowed materially over the past three years as Azure and GCP have grown faster off smaller bases, but it remains the largest single provider by a wide margin.

AWS's competitive strengths are the accumulated depth of its service catalog (over 200 discrete services), the maturity of its enterprise support and compliance certifications, and the network effect of its developer ecosystem. Its weaknesses are more subtle: some legacy pricing complexity, a growing perception among some enterprise buyers that its aggressive expansion into vertical software increasingly competes with its own customers, and less integration with mainstream enterprise productivity software than Azure enjoys.

Azure: the enterprise incumbent

Microsoft Azure has grown into the second-largest hyperscaler by exploiting a specific structural advantage: Microsoft's incumbency in the enterprise software stack. Companies running Windows Server, Office 365, SQL Server, and Active Directory have a natural gravitational pull toward Azure for infrastructure that integrates with those platforms.

The AI acceleration since 2023 has strongly favoured Azure. Its exclusive partnership with OpenAI, the enormous capital deployment into GPU capacity, and the deep integration of Copilot across the Microsoft product suite have combined to produce the strongest growth rates of any hyperscaler over the past two years. Whether this advantage is durable depends on questions about the AI ecosystem that are still open — competing model providers, cost pressures on compute, and the structural evolution of AI workloads.

Azure's weaknesses are the perception among some technical audiences of legacy engineering compromises (some inherited from Microsoft's server infrastructure roots) and the concentration risk of tying so much of its growth story to the OpenAI relationship.

GCP: the technical incumbent

Google Cloud Platform is the smallest of the three by market share but by no means the smallest in capability. It benefits from Google's decades of internal infrastructure investment — the search index infrastructure, YouTube's video serving, Google's own internal data platforms — that translate into cloud services with distinctive technical characteristics. BigQuery, Google's cloud data warehouse, is widely regarded as best-in-class for large-scale analytical workloads.

GCP's structural challenges are commercial rather than technical. Google's enterprise sales motion is younger and less mature than AWS's or Microsoft's. Its ecosystem of independent software vendors and system integrators is smaller. Its historical roots in serving Google's own consumer-web workloads mean it has historically been less optimised for the mainframe-and-Windows-server workloads that dominate legacy enterprise IT.

The AI moment has been more mixed for GCP than for Azure. Google's own AI models (Gemini) are competitive on published benchmarks, and Google's TPU chips are a genuine alternative to Nvidia GPUs, but the commercial translation of these advantages into cloud revenue has been slower than the technical potential would suggest.

The pricing dynamic

Cloud pricing has been broadly stable for standard compute and storage services for over a decade. The stability masks continuous price improvement on a performance-per-dollar basis — the same money buys much more compute or storage than it did five years ago. This "silent" price decline has been one of the strongest structural tailwinds for cloud adoption; customers migrating to cloud infrastructure have consistently found their all-in costs lower than the on-premises alternatives they replaced.

The pricing dynamic has begun to shift for AI-specific workloads. GPU compute is scarce, expensive, and pricing power has moved to the providers. The three hyperscalers have all announced multi-year committed capacity purchase agreements with Nvidia in the tens of billions of dollars, and the resulting supply is being sold at premium margins to customers with AI workloads. Whether these margins are sustainable depends on how quickly GPU supply expands, how quickly custom silicon alternatives (TPUs, Trainium, MI accelerators) capture share, and how quickly the workloads themselves migrate to more efficient architectures.

The regulatory context

Cloud infrastructure has become a topic of active regulatory attention in multiple jurisdictions. The UK's Competition and Markets Authority has been examining AWS and Azure specifically on questions of switching costs, licensing practices, and vertical integration. The EU has looked at similar questions through the Digital Markets Act framework. In the US, the FTC has raised concerns about the market structure but has not taken major enforcement action.

None of these processes is close to imposing structural remedies. They do, however, represent an ongoing regulatory overhang that could affect the sector's long-run margin trajectory — particularly on questions of switching costs and egress pricing that have become industry-standard practices.

What this frame means for an analyst

Do not treat "the cloud sector" as a single thing. The three hyperscalers have distinct competitive positions, distinct growth drivers, distinct exposure to the AI cycle, and distinct regulatory profiles. The correlation between their stock prices on daily returns is high; the underlying business dynamics are meaningfully different, and the divergence between them over multi-year horizons has been substantial.

Educational content only. Not investment advice.