When AI Starts to Look Like Infrastructure

There is a shift starting to appear in how AI is positioned. Rather than being something you buy, like software or a standalone tool, it is increasingly starting to look like something you access. Used when needed, paid for over time, and quietly embedded into how other systems operate.

Sam Altman recently suggested that AI could eventually be sold more like electricity or water. Not as a product, but as infrastructure.

That framing changes the conversation. Infrastructure is different from software. It sits underneath other systems. People stop thinking about it directly because it becomes part of how everything else functions. Electricity is rarely the focus of attention until it fails. Internet access feels invisible right up until it disappears.

If AI moves in the same direction, the implications become broader than features, subscriptions, or productivity tools.

When something becomes infrastructure, it stops feeling optional in the same way. Dependence increases. Businesses, services, and workflows begin building around it. Other systems start assuming its presence. And with that comes a different set of questions.

Pricing is no longer just about cost, but about access.

Reliability matters differently because outages or disruption affect everything built on top of it. Control becomes more significant because the organisations providing that infrastructure gain influence over how it is used, integrated, and distributed. It also changes how AI is perceived socially. Less as a tool people actively choose to use, and more as something that is simply there. Embedded into platforms, workflows, search, communication, administration, decision-making, and everyday digital processes in ways that are not always obvious or visible.

The shift is subtle, but important. Because once something starts to feel like infrastructure, it stops being a feature. It becomes something everything else quietly depends on.