Over the course of my career, I have participated in countless discussions about generation, transmission, substations, transformers, and other forms of physical infrastructure. While the conversations were often difficult, they shared one common characteristic: everyone understood what was being built and roughly how it would perform.

•   A transmission line has a rating.
•   A generator has a rating.
•   A transformer has a rating.

The industry knows how to model these assets, finance them, regulate them, and incorporate them into long-term planning.

What strikes me about today’s discussions around AI infrastructure is that some of the most important innovation is no longer occurring at the level of individual assets.

It is occurring in how assets work together.

Generation, storage, cooling systems, controls, and compute are increasingly being designed as integrated systems rather than independent components. Increasingly, value comes not only from the assets themselves, but from how intelligently they operate together.

That distinction matters.

Historically, when utilities needed greater reliability, capacity, or operating flexibility, the answer was often another physical asset. Today, some of those same outcomes can be achieved through coordination, controls, optimization, and intelligent operation across existing assets.

The challenge is that our planning, regulatory, and commercial frameworks were built to value and reward physical infrastructure.

They were not built to value and reward intelligence operating across physical infrastructure.

If we fail to make that transition, the industry will continue to solve every problem with more steel.

Additional infrastructure will always be needed. But if intelligent operation can deliver reliability, flexibility, and capacity at lower cost, failing to recognize that value will force the system to rely on additional generation, transmission, substations, and reserve margins to achieve the same outcome.

That means more capital invested than may ultimately be necessary. And because those costs are ultimately recovered through electricity rates, customers will bear the burden.

In other words, the inability to value and reward intelligence does not simply affect project economics. It risks increasing the long-term cost of electricity itself.

As AI infrastructure scales, learning how to value and reward intelligence may become as important as learning where to build the next generator, transmission line, or substation.