What it is
A hyperscaler is a company that operates cloud computing on a scale far beyond a typical business. The term usually refers to Amazon Web Services, Microsoft Azure and Google Cloud, and sometimes to other giants such as Oracle or Meta when they build comparable infrastructure. What sets them apart is the size of their data center fleets and their ability to add capacity quickly and cheaply.
What they do in AI
Training and running AI models takes huge amounts of compute, mostly from GPUs and custom chips. Hyperscalers buy or design those chips, house them in data centers, and rent them out. Many AI labs depend on them for the machines that train their models, and many companies reach AI models through a hyperscaler's cloud rather than directly.
Hyperscalers also build their own silicon and services. Google offers the Tensor Processing Unit, for example, and the big clouds sell hosted models and tools alongside raw hardware.
Why they matter
Three things make hyperscalers central to the AI industry:
- Capital. They spend tens of billions of dollars a year on data centers, chips and networking, a pattern covered under AI capex.
- Supply. Because they buy chips in bulk, they influence who gets scarce hardware and when.
- Reach. Their customers can switch on AI services in a few clicks, which spreads new tools quickly.
Hyperscalers versus neoclouds
A neocloud is a newer company focused mainly on renting GPU capacity for AI. A hyperscaler offers a far wider catalog, including databases, storage, networking and hundreds of other services. Neoclouds often compete on specialization and availability, while hyperscalers compete on breadth and scale.
Why you should care
If you build with AI, your costs and limits are tied to hyperscaler capacity, directly or through the providers you use. Local rules about land, power and permits can shape how fast they build, and that can affect availability and prices. Keeping your app able to move between providers reduces the risk of depending on any single cloud.