The platforms renting out intelligence.
Cloud computing is being reshaped by AI workloads that require large-scale accelerator clusters, data centers, power and long-duration capacity commitments. AWS, Microsoft Azure and Google Cloud remain the central platforms, but specialized AI clouds, major model providers and infrastructure-rich technology companies are broadening the market. The resulting competition is as much about physical capacity, financing and customer lock-in as it is about cloud software.
Hyperscalers provide the shared computing, storage and networking infrastructure on which enterprises and developers run digital workloads. AI has increased the strategic importance of those platforms because training and serving advanced models can require data-center-scale computation. The cloud contest therefore extends beyond conventional IT services to access to accelerators, large clusters, data-center capacity and the software layers used to deploy AI.
AWS, Microsoft Azure and Google Cloud are the main public-cloud competitors identified in the coverage. In Q1 2024, Altimeter put AWS at 31% cloud market share, Azure at 25% and Google Cloud at 11%. Those positions reflect an established market, while AI demand creates an opening for shifts in workload placement and provider differentiation.
AI infrastructure is increasingly procured as capacity rather than solely as on-demand cloud usage. GPU clusters, data-center space, power and networked infrastructure can be secured through advance reservations and multi-year commitments, making the duration and reliability of supply central commercial terms. This changes cloud competition: providers must assemble physical capacity before demand can be served, while customers seek assurance that capacity will be available when needed.
Large compute agreements illustrate this shift. OpenAI has planned to add Google Cloud for growing computing needs, while it also signed a seven-year AWS agreement for AI compute and began running workloads on AWS infrastructure. Such arrangements show that major AI customers may use more than one cloud provider when capacity requirements are large, even as long-term commitments deepen relationships between providers and customers.
Meeting AI demand requires continuing investment in data centers and other infrastructure. Microsoft, Meta and Alphabet disclosed more than $32 billion in combined data-center and capital spending in one quarter during their AI spending acceleration, while hyperscalers have continued to raise capital expenditures for AI. Morgan Stanley forecast that hyperscalers would fund $1.4 trillion of a projected $2.9 trillion in future AI infrastructure through 2028, with other financing sources supplying the remainder.
This buildout is an infrastructure challenge, not simply a chip-purchasing exercise. Power, land, grid interconnection, construction finance, permitting and community acceptance can all constrain usable AI capacity. The growing electricity demand associated with Amazon, Microsoft and Google’s expansion also places cloud investment in a broader energy and regulatory context.
The AI cloud market includes both established hyperscalers and specialized providers. AI neoclouds such as Crusoe, Lambda Labs and CoreWeave have been described as providers built around GPU rentals, offering another route to obtain AI compute. On-premises hardware suppliers, including Dell and Qualcomm, have also seen an opportunity as cloud providers face pressure to meet AI demand.
Platformization is widening the set of potential sellers. Meta has been reported to be planning a cloud infrastructure business that would offer AI compute and models in competition with AWS, Azure and Google Cloud. Companies that build enough internal compute for their own products can potentially commercialize spare or expanded capacity by hosting outside developers and model providers.
Scale can create an advantage in cloud infrastructure, but it can also sharpen concerns over customer choice. Ofcom found that hyperscalers including AWS and Microsoft Azure were limiting competition in the UK cloud market by making it difficult for businesses to switch. EU preliminary findings likewise identified Azure and AWS as the largest and second-largest cloud services in the bloc, respectively, as tougher oversight was being considered.
The central questions are whether capital-intensive AI expansion produces sufficient usable capacity, whether demand and cloud revenue support the level of investment, and how readily customers can distribute workloads across providers. Competitive outcomes will also depend on each provider’s ability to combine hardware access, AI platforms, contractual capacity and energy supply. As cloud infrastructure becomes more utility-like, financing discipline, power availability and regulatory treatment may matter as much as traditional cloud market share.
Grounded in the archive and knowledge graph. Browse all topic guides, the concept reference, or the posts.