Intelligence Per Hour: The Next AI Pricing Model

I think OpenAI and Anthropic will eventually move beyond per-token pricing and introduce an “intelligence per hour” consumption model.

It will look similar to the shift toward serverless infrastructure: customers stop paying for the underlying compute primitive and start paying for an abstracted unit of work.

The result will be higher value, higher margins, and much stronger vendor lock-in.

AI platform engineering will become one of the most important skills in the enterprise: optimizing workloads, routing tasks across models, controlling costs, and preserving the ability to migrate away from proprietary runtimes.

The more intelligent the runtime becomes, the more expensive it will be to leave.