North America Is Entering a Power Reality Check
For most of my career, when people talked about growth, they talked about capital, land, or connectivity. Power was always assumed. It was there, it would show up, and it would scale when we needed it.
That assumption is breaking.
For the first time in decades, power, not capital, not land, not fiber, is becoming the real limiting factor for economic growth in North America. And AI is the reason the math changed.
AI Changed the Equation
The last 20 years were all about efficiency. We optimized servers, compressed workloads, virtualized everything, and figured out how to squeeze more performance out of the same electrical footprint.
That era is over.
Modern AI workloads are power dense, always on, and intolerant of interruption. These systems do not idle gracefully and they do not wait patiently. A single hyperscale AI campus can now consume as much electricity as a mid sized city. That is not a projection, it is already happening.
At the same time, the grid was never built for this moment.
A Grid Built for a Different World
Most of North America’s transmission infrastructure was designed for predictable, linear load growth. Incremental increases. Long planning cycles. Known demand.
What we are seeing now is the opposite.
AI data centers, electrified transportation, advanced manufacturing, and domestic reshoring are all arriving at once. Demand is compounding, not stepping up. Interconnection queues stretch years into the future, and in many regions approved power does not actually arrive until 2029 to 2032.
This is not a policy issue. It is not a permitting issue alone.
It is a physics problem.
Why the Grid Alone Cannot Keep Up
Even where generation technically exists, the bottleneck is often transmission. Congested substations, transformer shortages, and long regulatory timelines mean that “available power” on paper is not usable power in reality.
For AI operators, waiting five to seven years is simply not an option.
Uptime matters. Latency matters. Delivery timelines matter. That reality is forcing a fundamental shift in how next generation data centers are planned, financed, and built.
The Rise of Off Grid and Behind the Meter AI Campuses
The most serious AI infrastructure players are no longer asking, “How fast can we interconnect?”
They are asking something very different.
“How do we control our own power destiny?”
Behind the meter and off grid generation, including natural gas, hybrid gas and renewables, fuel cells, and on site storage, is becoming core infrastructure. Not a temporary bridge. Not a stopgap.
These systems allow operators to deliver power on their own timeline, bypass congested transmission nodes, guarantee reliability for mission critical workloads, scale in phases as compute demand grows, and hedge long term power price volatility.
In many cases, these campuses are still designed to integrate with the grid later. The difference is that they are already live, operational, and generating revenue long before that happens.
Power Is Now the Moat
Over the next decade, competitive advantage in AI infrastructure will not come from who has the best GPUs alone.
It will come from who controls power, land, and delivery timing.
Sites with secured generation, fuel access, water, and expansion optionality will trade at premiums. Locations without power certainty, no matter how attractive they look in a spreadsheet, will stall.
This is why power first site selection is no longer a niche strategy. It is quickly becoming the dominant approach across hyperscalers, AI labs, and infrastructure investors.
The Bottom Line
North America does not have an AI problem.
It has a power delivery problem.
The winners will be the ones who stop waiting for the grid to catch up and instead build intelligently alongside it.
Power is no longer just an input.
It is the foundation.