Utilities Question the Reality of AI Data Center Power Demand

Electricity providers across the United States are grappling with the burgeoning demand projections fueled by the artificial intelligence boom. The stock market is already pricing in substantial investments in the infrastructure required to support a massive data center expansion, but the actual power consumption remains a significant question mark.

“There is a question about whether or not all of the projections, if they’re real,” Willie Phillips, former chairman of the Federal Energy Regulatory Commission (FERC), told CNBC. “Some regions initially projected huge increases, only to later readjust those figures.”

The crux of the issue lies in the ambitious plans of AI companies to establish server farms, each potentially consuming as much electricity as an entire city. However, these tech firms are simultaneously courting multiple utilities with the same large-scale project proposals in a bid to secure the quickest access to power. This tactic, while understandable from a business perspective, is creating significant forecasting challenges for utilities. This is amplified by the complex regulatory and planning landscapes differing from state to state.

“We’re starting to see similar projects that look exactly to have the same footprint being requested in different regions across the country,” Brian Fitzsimons, CEO of GridUnity, explained to CNBC. GridUnity’s software provides utilities and transmission operators with a comprehensive view of power project connection requests across the fragmented U.S. electric grid. This visibility is crucial for informed decision-making, but the ‘shopping around’ behavior of data center developers introduces a layer of uncertainty that is hard to quantify.

This competitive pressure among utilities, coupled with inflated or duplicated demand forecasts, creates a precarious situation. Utilities are struggling to accurately determine the necessary power generation capacity to maintain grid reliability. Concurrently, electricity prices are climbing for consumers as power supply struggles to keep pace with the existing demand, potentially exacerbating the issue for rate payers as utilities are forced to invest in additional infrastructure. Further complicating matters is the current regulatory environment, which is putting pressure on utilities to decarbonize their energy generation, so they may need to replace more reliable fossil sources with renewables, which are less consistent with energy generation.

Former FERC Chairman David Rosner cautioned that even a small discrepancy in electricity load forecasts can have far-reaching economic consequences. “Put simply, we cannot efficiently plan the electric generation and transmission needed to serve new customers if we don’t forecast how much energy they will need as accurately as possible,” Rosner stated.

Constellation Energy CEO Joe Dominguez echoed these concerns during the company’s May earnings call, stating, “I just have to tell you, folks, I think the load is being overstated. We need to pump the brakes here.”

Despite the caution from some industry players, the stock market has largely ignored calls for restraint. The steady stream of data center announcements is driving a major rally in power company stocks, reminiscent of the dot-com boom. Some analysts suggest that the market is over-valuing energy companies based on the projections; however, the companies are seeing the market enthusiasm and, as such, are continuing to move forward.

The utility sector has gained approximately 21% this year, following a more than 19% rally in 2024. This surge in value has added nearly $500 billion to the collective market capitalization of U.S. electricity providers over these past two years.

Even OpenAI CEO Sam Altman has voiced concerns about a potential AI bubble, cautioning investors against excessive exuberance. However, as AI infrastructure investments continue Altman and others may be proven wrong.

Despite the uncertainty surrounding the precise scale of future demand, experts largely agree that the U.S. is on the cusp of a significant increase in electricity consumption after a prolonged period of stagnant growth, which in and of itself will drive the prices up due to the current levels of inflation. The existing data center landscape provides a tangible preview of what is to come, according to Rob Gramlich, president of Grid Strategies.

“We can see it. Data centers exist,” Gramlich stated. “They’re operating day in and day out, using a lot of electricity. It used to be the case a 50 megawatt data center was pretty big. Now, it’s very common to have data centers that are 20 times that size — that are a gigawatt.”

Grid Strategies estimates an additional 120 gigawatts of electricity demand by 2030, including 60 gigawatts from data centers based on utility forecasts. To illustrate the scale, 60 gigawatts is roughly equivalent to Italy’s peak hourly power demand in 2024, the world’s eighth-largest economy. The firm concludes that this level of demand could have a destabilizing effect on the grid.

“This is not a bubble,” Fitzsimons argued. “It’s going to transform our nation completely. It’s going to continue to grow. We need a 50-year energy policy.”

However, Gramlich stresses the importance of securing firm financial commitments from data centers to refine demand forecasts. “That is going to help us rationalize all these requests and get a better handle on the total estimate,” he said. “But the industry’s got to plan based on the best information we have at the moment.”

The uncertainty surrounding demand forecasts has raised concerns about potential overbuilding, or even underbuilding, by utilities, leading to billions of dollars in wasted investment or insufficient capacity. Utilities invested $178 billion in grid upgrades last year and are projecting $1.1 trillion in capital investments through 2029, according to the Edison Electric Institute. If these projects cannot be put online fast enough, the US will face energy challenges in the next 5-10 years.

Fitzsimons suggests that the risk of utilities overbuilding is lower than it was two decades ago due to prevailing market constraints. “They’re in a very different environment where we have massive supply chain problems,” Fitzsimons said. “We have inflation running off the hook for quite a long time now. They can’t afford to overbuild. It’s going to come down to better planning.”

Despite previous bubble warnings, Altman recently forged a partnership with Nvidia , wherein OpenAI would build 10 gigawatts of data centers powered by the chipmaker’s graphics processing units. This plan requires approximately the same amount of electricity as New York City during peak summer days, raising questions about the feasibility of securing adequate power supply. The infrastructure that is needed to support that amount of data is significant.

While securing such a substantial amount of electricity may be technically feasible, the AI industry faces growing constraints as its plans expand, according to Gramlich. These constraints cause the price to expand, which is a risk factor for investors. The companies are competing for limited infrastructure, which in turn is driving up prices for essential electrical equipment such as transformers, switches, and breakers.

“We really don’t have the electrical infrastructure to meet the aggressive targets,” Gramlich argued. “We don’t have enough generation or transmission infrastructure to meet even the modest midpoint targets.”

The critical question is the pace at which new generation capacity can be built. Natural gas turbines are facing capacity through the remainder of the decade. The tech industry is also investing in advanced nuclear power, which are sold out or unable to be deployed fast enough, but these technologies are not expected to achieve commercial-scale deployment until the 2030s at the earliest. The cost of these deployments is also uncertain, which is a destabilizing factor in the utility industry.

Renewable energy sources, especially solar and battery storage, offer the potential for the quickest deployment. According to August data from energy consulting firm Enverus, over 90% of the power projects currently awaiting grid connection are solar, battery storage, or wind projects.

“For the past 10 years, our interconnection cues have been filled with a massive percentage of renewables,” Fitzsimons noted. “The renewables are the fastest way to build out new capacity. There’s no doubt in that because of the supply chain issues we have around natural gas turbines.”

However, the potential shift in energy policy, particularly with opposing viewpoints on solar and wind power, introduces uncertainty about whether sufficient new generation capacity can be built to meet the surging demand. Utilities will likely turn away data centers if they lack sufficient power capacity, according to Gramlich.

“If they literally do not have the power to serve a customer, they’re not going to sacrifice reliability,” Gramlich emphasized. “That’s their core job.”

Some AI companies are exploring alternative solutions like “behind the meter” power generation, where power is generated onsite at the data center, disconnected from the main grid.

As Nvidia CEO Jensen Huang stated, “We should invest in just about every possible way of generating energy… Data center self-generated power could move a lot faster than putting it on the grid and we have to do that.”