AI at a Crossroads: The Data Energy Surge Ignored by the U.S. Action Plan

By Ben Banerjee 

A Framing Thought:

The U.S. AI Action Plan promotes rapid infrastructure build-out, but neglects the physics of energy use. If AI is to be scalable and sustainable, policy must integrate efficiency. Designing smarter systems and demanding higher performance standards is not optional, it’s essential to balancing AI ambition with climate reality.

Two weeks ago, a sweeping new vision for American AI leadership was unveiled through the America’s AI Action Plan, backed by three Presidential executive orders. In bold strokes, the plan lays out how the United States intends to build faster, deregulate harder, and dominate smarter in the global AI race.

But beneath the visionary language lies a less glamorous reality. As much as the plan accelerates AI growth, it pays scant attention to the physical cost of intelligence at scale, namely, energy.

Let us follow the current to where it's really flowing: into the growing energy appetite of AI infrastructure, a story of rising electricity bills, strained grids, and invisible emissions with planetary consequences.

The Quiet Power Behind Your Query

Every AI interaction, from a chatbot answer to a recommendation algorithm, triggers computations inside data centers. These centers are the beating hearts of the digital age, stacked with high-performance servers cooled by industrial-scale air systems, drawing power day and night.

In 2023 alone, U.S. data centers consumed an estimated 176 terawatt-hours of electricity, roughly 4.4% of national electricity use. That’s more than the entire annual consumption of some mid-sized European countries. Worse, much of this energy still comes from fossil fuels.

The real alarm isn’t where we are, but where we’re going. Projections suggest this consumption could triple within five years, driven largely by AI’s heavy computational demands. Without intervention, data centers could soon consume up to 12% of U.S. electricity, rivaling the steel and chemical industries.

What the Plan Gets Wrong

Rather than acknowledging this surge, the U.S. AI Action Plan leans into deregulation: streamlining permits, loosening environmental protections, and fast-tracking infrastructure approvals. Nowhere in the document is there a meaningful reference to efficiency standards, let alone incentives for green infrastructure.

Utilities, meanwhile, are sounding alarms. States like Virginia and Ohio have started imposing special tariffs on hyperscale data centers to recover the costs of grid expansion. In some communities, pollution from data center–linked fossil energy has triggered health costs surpassing $5 billion annually, disproportionately affecting low-income neighborhoods.

We are, in effect, building a faster machine without asking how we’ll fuel it.

A Better Blueprint: Innovation in the Shadows

Not all is bleak. Some innovators have been quietly engineering a different path forward.

Take Swiss Vault , a Swiss company developing ultra-efficient data storage systems. Their proprietary design enables 5 Petabytes of storage to run on just 5 kilowatts of power, compared to up to 150 kilowatts in standard setups. This isn’t a minor improvement; it’s a 30x leap in efficiency.

Solutions like these don’t just save energy. They open the door to scaling AI without locking in higher emissions or infrastructure strain. They prove that the equation—more AI equals more carbon—doesn’t have to hold.

In Conclusion: Powering AI Without Powering Collapse

We often talk about artificial intelligence as though it’s made of code and clouds, but in truth, it’s made of circuits, servers, and electricity. The smarter our machines become, the harder they press on our grids, our policies, and our climate commitments.

If we’re serious about leading in AI, we must look beyond speed and scale, and begin designing for resilience, responsibility, and real-world physics. The U.S. AI Action Plan may offer a blueprint for ambition—but without energy efficiency, it risks building castles on sand.

It’s time we treat infrastructure not just as a backdrop, but as a lever for change. And in doing so, we can steer AI not only toward progress, but toward sustainability that lasts.

By Ben Banerjee