Hyperscaler Glossary
A hyperscaler is a company that operates computing infrastructure at enormous scale: data centers, cloud regions, networking, power contracts, storage, GPUs, custom chips, and the operational staff to keep all of it running.
Amazon, Microsoft, Google, Meta, and Oracle are common examples, although analysts draw the boundary differently. The term describes an operating scale and architecture rather than a tidy industry membership list.
Frontier-AI labs depend on this layer through several arrangements. Some buy cloud capacity, some form investment and infrastructure partnerships, and some build or control large clusters themselves. OpenAI has used Microsoft Azure and other providers; Anthropic has announced capacity agreements with Amazon, Google, Microsoft, and SpaceX; xAI developed the Colossus cluster later acquired by SpaceX. Calling every lab a hyperscaler would blur the useful distinction between making models and operating the infrastructure on which they run.
Hyperscalers are powerful because scale creates advantages: cheaper hardware procurement, higher utilization, specialized cooling, redundancy, security teams, custom inference stacks, and the ability to finance enormous training runs.
The sovereignty problem is not that hyperscalers are useless. The problem is that they are useful enough to become the default substrate for everyone else’s cognition.
Sources
- Anthropic, “Higher usage limits for Claude and a compute deal with SpaceX”, 2026.
- Microsoft, “Microsoft and OpenAI evolve partnership to drive the next phase of AI”, 2025.