The field of artificial intelligence is evolving towards a system in which it is considered a basic service rather than a product. At the BlackRock Infrastructure Summit in Washington, DC, on 11 March 2026, OpenAI CEO Sam Altman outlined a model in which AI can be delivered “on demand” and billed according to usage.
In essence, this model considers AI to be similar to electricity and water in that users are charged based on consumption. This is a new way in which digital service structures are evolving, particularly in a world in which more computing power is constantly in demand.
Sam Altman explains how AI token usage could replace fixed subscription fees
Sam Altman further added that AI systems already function in measurable units, which are called “tokens.” Tokens are defined as the amount of data processed when a user asks a query and receives a response.
In a utility model, these units are what define the pricing.
The model is based on a system where users are charged according to the amount of AI they use. In this model, a simple query would use fewer tokens, but more complex queries would involve more computation and hence more usage. The model thus provides a form of proportional pricing.
Sam Altman explains the role of compute in AI pricing and access
The availability of compute remains central to this system. Compute refers to the processing capacity needed to run AI models, supported by specialised chips, servers, and large-scale data centres.
Altman indicated that limited compute supply can influence both price and access. If demand exceeds available capacity, services may become expensive or restricted.
Sam Altman on how energy availability will become a key factor in AI as a utility
The availability of energy has come to be linked to the development of AI. This is because the capacity to produce and utilize energy has a direct link to the pace at which infrastructure to support computers can be rolled out. The pace at which countries are developing their energy infrastructure could determine their position in the development of AI capabilities.
The faster pace at which countries are developing their energy infrastructure would allow them to roll out their data centre infrastructure faster. This would mean that there is a direct link between the energy capacity of a country and its technological capabilities in AI.
The utility-based model represents a paradigm shift in the provision of artificial intelligence. This is because, instead of paying for a product, users would be using a utility.
