infrastructure

Infrastructure is the compute, storage and networking on which software runs, and cloud computing rents it from a provider instead of owning the hardware. The provider manages the physical infrastructure, and the customer manages the software (or less, depending on the abstraction level).

Infrastructure is the compute, storage and networking on which software runs, and cloud computing rents it from a provider instead of owning the hardware. The provider manages the physical infrastructure, and the customer manages the software (or less, depending on the abstraction level).

The major providers are AWS, Google Cloud, Azure, Cloudflare and DigitalOcean.

The key shift is from capital expenditure (buying servers) to operational expenditure (paying per use).

It comes in three classic layers: IaaS (virtual machines), PaaS (managed runtimes) and SaaS (finished products).

Under every layer is a building. A data center turns megawatts into computation and heat, with its own power chain and redundancy, and its limits on power and cooling now decide how much AI hardware can be installed at all.

Machines are combined in two ways: made bigger or made more numerous (horizontal and vertical scaling); the second leads to the compute cluster, scheduled by an orchestrator such as Kubernetes and paid for under a discipline of its own (FinOps).