A hyperscaler is a cloud giant — AWS, Microsoft Azure, or Google Cloud — operating global datacenter fleets that host much of the world's AI compute.
The big three clouds occupy every layer of the AI stack: they build and lease the datacenters, buy accelerators at the largest scale (increasingly alongside chips of their own — Trainium, TPUs, Maia), fund frontier labs through multi-billion-dollar partnerships, and sell both raw GPU capacity and managed model APIs.
Their model platforms — Amazon Bedrock, Azure AI Foundry, and Google Vertex AI — resell frontier and open models with enterprise wrapping: unified billing, private networking, regional residency, and compliance certifications. Per-token prices generally track the model creators' own rates, so the draw is procurement and governance rather than discounts, though committed-spend agreements can effectively subsidize usage.
For enterprises already inside one cloud, the hyperscaler route is often the path of least resistance to production AI; the strategic caution is concentration — compute, models, and infrastructure increasingly settle into a handful of vertically integrated stacks.
Last revised 2026-07-05 · All glossary terms → · Live AI model pricing →