AI Is Driving a New Era of Energy Demand The Time to Act Is Now

MIT Technology Review recently released a powerful series on the future of AI and energy. The numbers are telling, and the implications for data centers are urgent.

By 2028, data centers could account for up to 12% of total U.S. electricity use, according to projections from Lawrence Berkeley National Laboratory. That's a dramatic shift in a short window.

Looking out over the next decade, U.S. electricity demand is expected to increase by 1,000 to 1,200 billion kilowatt-hours, a 24% to 29% jump. Here's how that increase breaks down:

~50% from the electrification of vehicles
~30% from electrified buildings and industry
~22% from AI and data centers

While AI's share may seem modest, it's the most urgent part of the equation. Unlike vehicles or buildings, AI's energy demand is surging now: spiky, unpredictable, and geographically concentrated. It's already straining grids and prompting calls to restart fossil-fired power plants.

And this is the beginning.

As the MIT team notes, today's AI workloads are relatively lightweight. But tomorrow's systems (agentic, multimodal, always-on) will push infrastructure further. The challenge isn't only scale. It's volatility. And existing monitoring infrastructure needs a new detection layer to keep pace with this new paradigm.

Verdigris: Built for the New Game

In the old game, energy was expendable. In the new game, energy is the constraint.

Data centers were built to contain and control. Keep the lights on. Plan for the worst. Overprovision everything. That playbook, and the monitoring infrastructure that supports it, works well for predictable loads.

AI has redefined the rules. Now, watts are the limiting factor, not space or compute. And the electrical events AI workloads create require a detection layer that existing metering wasn't designed to provide.

Verdigris adds that layer. We complement existing infrastructure with high-frequency intelligence, turning steady-state monitoring into an aware, adaptive, and orchestrated system.

Working alongside your existing panels, DCIM, and BAS, Verdigris enables discovery of stranded capacity, prediction of power bottlenecks, simulation of future constraints, and dynamic alignment between electrical infrastructure and business logic.

The AI era doesn't require replacing what works. It requires adding the validated layer that makes existing infrastructure ready for what's next.