Make better decisions on the cloud infrastructure market

Cloud infrastructure is growing fast again, and AI is pushing the bills up.

We've found a lot of interesting signals. For example, OpenAI signed a $38B deal in November 2025 to run its models on AWS over seven years. Then, in February 2026, Alphabet said it would spend $175B to $185B on infrastructure this year, about double the year before, as its cloud backlog of signed contracts reached $240B. The biggest buyers are locking in capacity for years, and the clouds are spending as fast as they can to deliver it.

We'll dig into why AI has restarted cloud growth, who is paying for all this capacity, and what it means for smaller providers and the companies renting from them.

Companies moved to the cloud to save money and move faster. Many now get a monthly bill nobody fully understands.

We compare six kinds of cloud infrastructure companies, from compute and storage providers to Kubernetes platforms, edge clouds and tools that manage costs across clouds. Their buyers range from startups and SaaS companies to banks and governments. Then we rank what they need most: lower bills, capacity when they need it, and control over where their data sits.

This shows which problems buyers will pay to fix first, and which ones they put up with to avoid a move.

Pay-as-you-go compute, reserved capacity, multi-year spending commitments, storage and data transfer fees, and managed platforms priced on usage: cloud infrastructure providers make money in many ways.

We compare who pays and what it costs to serve them. Hyperscalers spend billions on data centers years before the revenue arrives. Smaller clouds compete on price, but they pay as much for chips and more to borrow. Software layers like Kubernetes platforms keep higher margins, but they depend on the clouds they run on. And fees for moving data out bring in money while making customers angry.

You'll see which models can earn back huge upfront spending, and which ones get squeezed between the hyperscalers and their own suppliers.

We give you a realistic growth rate. Here, we show what has to keep happening for the market to get there.

Some drivers are plain to see: AI training and inference, more data, and companies still moving old systems out of their own server rooms. Others feed each other. Every new AI app needs compute, storage and networking around the model. Long contracts give providers the confidence to build more. And once a company runs on one cloud, it tends to add more services from the same provider.

Before building or investing, you should know what keeps customers spending more each year, and what makes them move work somewhere cheaper.

AWS, Microsoft Azure and Google Cloud take most of the money in cloud infrastructure, but they're far from the only players.

Away from the giants, a lot of smaller providers do one thing well. Some rent simple servers to developers and small companies at simpler, cheaper prices. Others run clouds that keep European data under European law, manage Kubernetes for teams that don't want to, or put computing in hundreds of cities close to users.

They're hard to compare because very different businesses use the same word. A provider that owns its data centers and a platform that runs on top of a hyperscaler can both call themselves a cloud. And many smaller providers are private, so their revenue and margins rarely show up anywhere.

Below, you'll find the market grids and ecosystem maps we've built to show who runs which layer of the cloud, and where smaller providers still win deals.

Cloudflare, DigitalOcean, OVHcloud… a few companies have built real cloud businesses next to the giants.

Cloudflare started with a network that sits in front of websites, then let developers run code on it. DigitalOcean kept its cloud simple and predictable for developers and small companies. OVHcloud builds its own servers and sells European customers a cloud that stays under European law. None of them tried to beat the hyperscalers at everything.

We'll show how these teams picked a fight they could win, and kept customers who could easily have gone to the biggest clouds.