The global infrastructure-as-a-service market vaulted to $171.8 billion in 2024, a 22.5% leap that starkly outpaced the prior year’s 16.2% growth, as the three largest providers—Amazon Web Services, Microsoft Azure, and Google Cloud—tightened their stranglehold on cloud infrastructure to a combined 71% share. Behind the gaudy numbers is a fundamental rewiring of enterprise IT: generative AI workloads are reshaping demand for compute, networking, and storage, forcing cloud giants to pour unprecedented capital into data center builds and GPU-heavy hardware. Gartner’s latest data, released May 28, confirms that the big three’s dominance is not only intact but intensifying, even as the entire market expands at its fastest clip in years.

Why the Market Accelerated

Two forces collided in 2024 to produce the sharpest IaaS growth since the pandemic-era cloud rush. First, the relentless enterprise push to modernize legacy applications and migrate them to the cloud continued to generate baseline demand for virtual machines, object storage, and managed Kubernetes services. Second—and far more disruptive—was the explosion of generative AI. Training and serving large language models requires orders of magnitude more high-performance compute than traditional workloads, and most organizations lack the capital or expertise to build GPU clusters on their own. Instead, they turned to hyperscalers, which can deploy thousands of accelerators in minutes and wrap them with managed services that shrink time-to-value for model development.

“As enterprises continue to seek greater flexibility, improved resilience and optimized performance, there is sustained demand for cloud migration and modernization services,” said Gartner Principal Analyst Hardeep Singh in the report. “Enterprises want to transform their IT infrastructure by leveraging multiple platforms for AI and prioritizing modernization by migrating existing workloads to the cloud.” The result: IaaS carved out the fastest segment growth within the broader public cloud market, which Gartner already expects to surpass $675 billion in 2024 and top $700 billion in 2025.

The Big Three’s Divergent Paths

Together, AWS (38%), Microsoft Azure (24%), and Google Cloud (9%) pocketed nearly $122 billion of the $171.8 billion IaaS pie. Each is navigating the AI boom with a distinct playbook, and their market share trajectories reveal how those bets are paying off.

AWS: breadth and proprietary silicon. AWS remains the 800-pound gorilla, with a 37.7%-38% slice that still dwarfs its rivals. Its strategy in the AI era leans on its broadest-by-far portfolio of instance types—now including Trainium and Inferentia chips—coupled with a global data center footprint that spans more regions than any competitor. However, because AWS’s revenue base is so large, its year-over-year growth rate in IaaS can appear slower than smaller foes, and Gartner’s data shows a modest share contraction from 2023. That erosion is relative; in absolute dollars, AWS added more new revenue than any other provider.

Microsoft Azure: enterprise muscle and hybrid lock-in. Azure’s 23.9%-24% share marks another year of gains, fueled by its unique ability to convert existing enterprise relationships into cloud consumption. For any organization running Windows Server, Active Directory, or Microsoft 365, Azure offers a path of least resistance: native integrations, unified identity and compliance tools, and hybrid services like Azure Arc that extend management policies to on-premises data centers. AI has only amplified that advantage, as Microsoft layers Copilot and AI services into productivity stacks that enterprises already use. “Cloud providers are investing heavily in AI infrastructure and capabilities to become leaders in the rapidly evolving AI-optimized IaaS market,” Singh noted. “They expect that AI will become a much larger revenue contributor in the future, even though it currently remains a relatively small slice of their overall revenue within the IaaS space.”

Google Cloud: the AI-first challenger. Google Cloud’s 9% share may look small, but its growth rate has been the fastest among the big three in recent quarters. Its bet is straightforward: use its research pedigree to attract AI-native companies and data-centric workloads with pre-built models, high-performance infrastructure, and open-source ecosystems. While Google lacks the enterprise installed base of Microsoft or the pure scale of AWS, it has carved out mindshare among startups and data scientists who value its AI tooling.

The Infrastructure Building Boom

Feeding this AI appetite is requiring hyperscalers to rebuild their data centers from the ground up. Dell’Oro Group recorded a more than 50% year-over-year spike in data center capital expenditures during the first quarter of 2025, pushing quarterly spending to $134 billion. For all of 2024, data center investments topped $450 billion, a 51% jump over the prior year. Those figures reflect a supply-side scramble to provision the racks, power distribution, and cooling systems that modern AI hardware demands.

Key investments unfolding across the industry include:
- GPU clusters and custom accelerators: New instances packing NVIDIA H100s, AMD MI300X, and proprietary chips (AWS Trainium2, Google TPU v5) are becoming the norm. These accelerators consume exponentially more power than general-purpose CPUs, forcing hyperscalers to standardize on dense rack designs that can host 4–8 GPUs per node.
- Liquid cooling adoption: Air cooling, long the default in data halls, is no longer sufficient for many AI racks. Direct-to-chip liquid cooling and rear-door heat exchangers are being deployed rapidly, with some providers testing full immersion cooling for their highest-density clusters.
- Power infrastructure upgrades: AI data centers are drawing electricity at the scale of small towns. Hyperscalers are investing in upgraded substations, busway systems, and on-site battery storage to ensure reliability. In some regions, grid capacity constraints are already delaying new data center construction.
- Network fabric evolution: To minimize inference latency, cloud providers are distributing model-serving clusters closer to population centers and upgrading interconnect fabric with 400 Gbps and 800 Gbps links.

These investments are not just building for today’s demand; they’re a bet that AI adoption will accelerate further. And early 2025 data center physical infrastructure growth of 17% year-over-year, as reported by Dell’Oro, suggests that bet is only getting larger.

What It Means for Windows Shops

For IT teams steeped in the Microsoft ecosystem, the current market dynamic presents a clear, if nuanced, opportunity. Azure’s deep integrations with Windows Server, SQL Server, and Microsoft 365 mean that migrating on-premises workloads to the cloud is often a matter of lifting and shifting—then gradually modernizing. Tools like Azure Migrate, Azure Arc, and the Azure Hybrid Benefit can cut migration costs and extend unified management across hybrid environments. For organizations that have standardized on Windows-based development stacks, Azure’s AI services—from Cognitive Services to the Azure OpenAI Service—plug directly into familiar DevOps pipelines.

Yet Windows-centric enterprises should not conflate convenience with inevitability. Lock-in risk is real: as Azure layers more proprietary AI services onto its IaaS foundation, the complexity and cost of later moving those workloads to another cloud can spike. Savvy IT leaders are already using containerization (Windows containers on Kubernetes) and infrastructure-as-code practices to maintain portability, even while they take advantage of Azure’s tight integrations. A hybrid strategy—keeping some sensitive or latency-critical workloads on-premises via Azure Stack HCI—can also serve as a hedge against vendor lock-in and provide cost predictability.

Risks Looming on the Horizon

For all its momentum, the cloud infrastructure market is navigating a field of potential landmines.

Power and supply constraints: The electricity demands of AI data centers are colliding with finite grid capacity. Several hyperscalers have quietly warned about power-related delays in new region rollouts, and small to mid-sized data center operators are being squeezed out of some markets altogether. If grid upgrades can’t keep pace, the supply of AI-ready capacity could tighten, driving up prices.

Margin pressure and capital intensity: AI hardware is ferociously expensive, and the break-even threshold for a new data center has risen sharply. While the big three can amortize these costs across millions of customers, smaller providers—and even some tier-two clouds—face a tougher road. A sustained price war over GPU instances could compress margins across the board.

Regulatory and antitrust scrutiny: When three companies control 71% of a market as critical as cloud infrastructure, regulators take notice. The European Union and the U.S. Federal Trade Commission have both sharpened their focus on cloud market concentration, data portability, and vendor neutrality. New rules could force hyperscalers to unbundle services, reduce data egress fees, or adopt interoperability standards—any of which could reshape competitive dynamics.

Technical lock-in: The more enterprises adopt a hyperscaler’s managed AI services—custom APIs, proprietary model formats, integrated data lakes—the harder and costlier it becomes to leave. Portability remains more aspiration than reality for many AI workloads, making it essential for CIOs to architect for flexibility from day one.

Strategic Playbook for IT Leaders

Given the market’s velocity and complexity, IT decision-makers should anchor their strategies in a few foundational moves.

1. Classify workloads ruthlessly. Not every application needs to run on the latest GPU instance. Separate workloads into tiers: AI training and inference, latency-sensitive customer-facing services, general enterprise apps, and archival/storage. Use this taxonomy to decide which cloud (or on-prem) environment each tier belongs in.

2. Embrace multi-cloud by design, not by accident. Instead of defaulting to a single provider, build a containerized, infrastructure-as-code foundation that can target AWS, Azure, or Google Cloud with minimal rework. Tools like Terraform, Pulumi, and Kubernetes abstract away many provider-specific wrinkles—provided teams invest in the upfront engineering.

3. Optimize cloud economics relentlessly. Implement real-time cost observability, set budget guardrails for AI experimentation, and leverage reserved capacity where it makes sense. For variable AI workloads, consider GPU-as-a-service alternatives from specialized providers (like CoreWeave or Lambda Labs) that offer flexible pricing and faster provisioning for certain use cases.

4. Build an AI infrastructure plan that includes on-premises. For predictable, long-running training jobs or highly sensitive data, a co-located GPU cluster or on-premises AI appliance can be cheaper and more compliant than cloud bursting. Treat the cloud as an elastic extension, not the sole home for AI.

5. Negotiate contracts with exit in mind. Demand transparency on data egress costs, model portability, and committed-use discounts. If a provider won’t support open model formats or standard API specifications, factor that into your risk assessment.

Conclusion

Cloud infrastructure in 2024 is no longer just about virtual machines and object storage; it’s the foundation for an AI-driven economy. With IaaS spending hitting $171.8 billion and showing no signs of cooling, the hyperscalers are fortifying their positions through massive physical investments and deeply integrated AI services. For Windows-centric enterprises, Azure offers an express lane to modernization—but the broader lesson is one of strategic balance. The coming quarters will be decisive: those who marry the speed of hyperscaler innovation with deliberate portability and cost governance will extract the most value, while those who race in blind risk being hemmed in by the very platforms that promised them freedom.