For most of the history of the electric industry, intelligence resided in the grid.

Utilities forecast demand, dispatched generation, managed reserves, maintained reliability, and balanced the countless interactions occurring across the system every second of every day. Customers consumed electricity. The grid absorbed complexity.

That model worked because the grid possessed most of the information, most of the control, and most of the resources required to optimize the system. AI infrastructure is beginning to change that relationship.

Most of the discussion focuses on generation. How much power is required? Where will it come from? How quickly can it be delivered? Important questions, but they miss a more fundamental shift.

The traditional model treated the campus as a load. The emerging model treats the campus as a system. And systems make decisions.

What is being built around many AI campuses today is not simply a source of power. It is an integrated operating environment consisting of generation, storage, controls, communications, and software capable of coordinating both energy and compute. The objective is no longer just to secure electricity. The objective is to optimize how the entire system performs. That changes where intelligence resides.

Historically, if demand increased unexpectedly, a generator tripped offline, power prices changed, or system conditions tightened, the grid responded. Increasingly, large campuses are being designed to respond to those conditions themselves. They can determine how generation is dispatched, when storage is charged or discharged, how workloads are allocated, how ramp rates are managed, and how the facility interacts with the broader network. In effect, decisions that once occurred primarily at the grid level are beginning to occur at the edge of the system. That does not mean the grid becomes less important. In fact, I believe the opposite is true.

Most of the generation, storage, and control infrastructure being built around AI loads today will eventually become connected to the broader power system.

Once built, these assets represent capacity, flexibility, and reliability resources that can create value well beyond the campus fence line. The industry often frames the discussion as grid-connected versus behind-the-meter. I suspect that is the wrong lens.

The more important question is how intelligence is distributed across the system and how the grid, large customers, and regulators work together to optimize it.

For more than a century, intelligence largely resided at the center of the network. The next generation of power infrastructure may look very different. The future electric system may not be defined by where power is generated. It may be defined by where decisions are made.