infrastructure//cloud computing//hyperscaler

A hyperscaler is a cloud provider that operates computing infrastructure at the largest scale, dozens of regions of its own data centers worldwide, and sells hundreds of services on top of it, and in practice the word names a handful of companies: Amazon Web Services, Microsoft Azure and Google Cloud, with others such as Oracle and Alibaba Cloud in some counts. It is what most people mean by cloud computing: rent a virtual machine, a database, a queue or a model API by the hour or the request, and let the provider run everything underneath.

A hyperscaler is a cloud provider that operates computing infrastructure at the largest scale, dozens of regions of its own data centers worldwide, and sells hundreds of services on top of it, and in practice the word names a handful of companies: Amazon Web Services, Microsoft Azure and Google Cloud, with others such as Oracle and Alibaba Cloud in some counts. It is what most people mean by cloud computing: rent a virtual machine, a database, a queue or a model API by the hour or the request, and let the provider run everything underneath.

What sets a hyperscaler apart is breadth and integration more than raw size. Beyond compute and storage it offers managed databases, identity, networking, analytics, machine-learning platforms and compliance certifications for regulated industries, all wired into one account, one bill and one permission system (IAM). Its own campuses are hyperscale data centers of tens to hundreds of megawatts (colocation uses the same word for the building class), and it designs its own servers, networks and increasingly its own chips.

The breadth is both the value and the trap. A team can assemble a product from managed services in days, but each proprietary service it adopts raises the cost of leaving later (vendor lock-in), and egress fees for moving data out add a toll on the way.

For GPU-heavy AI work the comparison is now with the neocloud, providers specialised in renting accelerators. A hyperscaler offers GPUs inside the whole ecosystem of services and enterprise contracts; a neocloud offers mostly the GPUs, often sooner or in larger blocks. Neither is always cheaper or faster: the price depends on the hardware generation, the commitment, the region and the terms.

The word hyperscaler describes scale and breadth, and says nothing about fit. The same provider is the right choice for an enterprise that wants one contract and certified services, and an expensive one for a team that needs a thousand GPUs for six months and nothing else (FinOps).