AI is reshaping grid demand. Efficiency must lead the response
Artificial intelligence is advancing faster than our energy systems were built to handle. I’ve watched this shift accelerate over the past few years, and the numbers are starting to reflect what utilities are already feeling on the ground. Across North America, electricity demand is rising in ways that challenge traditional planning models. In the U.S. alone, Deloitte projects peak electricity demand could grow by about 26% by 2035, while data center demand could reach 176 gigawatts — roughly five times 2024 levels.
That kind of growth would have been hard to imagine just a few years ago. Today, it’s changing how utilities plan, invest, and operate.
It’s tempting to frame this as a crisis. It’s more useful to see it as a stress test: Can the energy industry respond quickly enough, and with the right tools, to manage a very different load profile than the one the grid was designed around?
I believe it can, but only if efficiency and demand flexibility move to the center of the response.
The timing problem
Building new generation and transmission has long been the traditional answer to rising demand. Those investments still matter, but they take time.
Large-scale infrastructure can take a decade or more to permit, build, and connect. AI-related load is arriving in years, not decades. Data centers can be developed quickly, and once online, they create continuous, high-intensity demand that doesn’t follow traditional consumption patterns.
This mismatch is already showing up across the system. Utilities are managing growing interconnection queues, regional reliability pressure, and rising uncertainty around where the next large load will appear. Regulators and policymakers are also taking a harder look at how large new loads connect to the grid, who bears the infrastructure costs, and what kind of flexibility those customers may need to provide in return.
The system is adapting in real time. But new infrastructure and regulatory reform, while necessary, will not move fast enough on their own.
Demand is part of the solution
What’s emerging is a different way of thinking about demand — and the industry needs to move faster to make it a standard practice.
Some of the world’s largest technology companies are already working with utilities to make their energy use more flexible, adjusting consumption during periods of grid stress rather than simply drawing maximum power at all times. That matters because it points to a broader shift: large new loads do not have to be managed only through more supply.
Demand flexibility is becoming a strategic tool for managing that growth. By changing when electricity is used, not just increasing how much is generated, utilities can better balance supply and demand, especially during peak periods. And unlike major infrastructure projects, these approaches can often be deployed on much shorter timelines.
Efficiency is the fastest resource we have
Demand flexibility changes when electricity is used. Efficiency changes how much is needed in the first place. For too long, that's made efficiency feel incremental — important, but not central in moments of rapid growth. The AI era changes that entirely.
Efficiency and demand-side flexibility are the fastest, most cost-effective ways to support the grid in the near term. They help defer or better target infrastructure investment, give utilities and grid operators breathing room while longer-term solutions come online, and work within the system we already have.
That matters in an environment where timing is everything. The industry does not have the luxury of waiting a decade for every solution to arrive. It needs resources that can be deployed sooner, scaled efficiently, and integrated into current operations.
What this means for utilities
AI is not a passing trend. It’s a structural shift in how energy is consumed, and it will shape grid planning for years to come.
Utilities that treat this moment as purely a supply challenge will find themselves behind and more exposed to cost, timing, and reliability pressure. The ones that integrate demand-side solutions into their core strategy — alongside new generation and transmission — will be better positioned to manage load growth, protect reliability, and serve customers through the transition.
Hydro-Québec’s $10 billion, decade-long efficiency investment strategy targets a 10% reduction in total electricity consumption — equivalent to the power used by one in four Quebec homes —even as the province expects data center peak demand to increase fivefold by 2035. The strategy plans for rising electricity demand alongside electrification and other system needs, treating efficiency as a resource rather than a side program.
Similar thinking is emerging in the U.S. AEP Ohio’s approved Data Center Tariff requires data center customers to pay for a minimum of 85% of the energy they subscribe to use — even if actual usage is lower, ensuring they help fund the infrastructure needed to serve them. It’s an early step toward treating large new loads as financially accountable for the grid capacity they require, rather than unconstrained demand.
Together, these examples reflect a broader shift toward integrated planning that treats demand-side resources and cost-accountable infrastructure planning as essential to maintaining reliability and managing growth. That kind of planning is where the industry needs to go — and it’s the work CLEAResult has been doing for more than two decades.
As AI-driven load growth accelerates, the capabilities CLEAResult has built — helping utilities and large energy users reduce demand, improve efficiency, and build programs that deliver measurable results at scale — are becoming more important to how utilities respond. Program design, demand-side management, customer engagement, and large-load integration are no longer supporting strategies; they’re part of the solution.
The path forward
The idea that AI will overwhelm the grid is understandable. It's also incomplete.
Demand is rising and new supply must be built. But framing this as a choice between growth and reliability misses the opportunity. We already have tools that reduce demand, shift load, and make the entire system more responsive.
The pressure on the grid is real — and so is the opportunity. Utilities that move now to integrate efficiency and demand flexibility into their core strategy won't just manage this transition. They'll be better positioned for every wave of load growth that follows.
That's not a distant possibility. The tools exist. The models are working. The question is whether the industry moves fast enough to use them.