Datacenters could blow up your electric bill thanks to AI

Americans could face a 70 percent hike in their electricity bills by 2030 unless action is taken to boost generation and transmission capacity to satisfy an AI-driven surge in demand from datacenters.

This latest warning comes in a report from Washington, D.C.-based think tank the Jack Kemp Foundation (JKF), which says that billions need to be invested in extra grid capacity to keep pace with the growth in consumption by datacenters, with many more under construction.

Failure to do so could force regulators to allow price increases of as high as 70 percent as they ensure the grid functions properly and delivers energy to all users, with regular brownouts and blackouts possible across the country, but particularly in bit barn hotspots such as Northern Virginia.

An accompanying infographic shows the percentage of power that datacenters are estimated to account for in each US state by 2030.

The report, "The Impact of Datacenters on Energy Demand and Market Prices," points out that the US had seen two decades of relatively static energy consumption, but demand is now growing rapidly thanks in no small part to server farms chugging through AI workloads.

This echoes warnings from management consultancy Bain & Company, which forecast that energy use could outstrip supply in just a couple of years, and advised US utility operators to boost their annual energy generation by up to 26 percent by 2028 to fulfill the expected demand.

Gartner also recently estimated that datacenter energy demand could baloon 160 percent over the next two years, leading to 40 percent of existing facilities being operationally constrained by the availability of power from 2027.

The JKF report, however, is the first we've seen which lays out in stark terms what the consequences of all this booming AI-driven demand for energy could mean for US consumers and businesses.

The problem is that AI use is driving a wave of new datacenter construction, as well as pushing up the energy consumed by the IT infrastructure therein. This growth is outpacing new energy supplies, so spare capacity in the US power grid is shrinking. This slack in the system provides a margin of safety during hot summers, winter storms, and temporary outages by power plants, the report says, and it is disappearing.

The report singles out Northern Virginia – the datacenter capital of the world – where utility Dominion Energy indicated that bit barns represented 24 percent of its revenue in 2023, up from 21 percent the previous year.

According to the JKF report authors, nine new facilities have already been connected to the grid in the state in 2024, with 15 more expected to come online by the end of the year. In Loudoun County, energy demand from all those datacenters is expected to exceed seven gigawatts by 2030, effectively doubling in just seven years.

• Japan looks to nuclear energy to power AI-powered datacenter boom
• Datacenters line up for 750MW of Oklo's nuclear-waste-powered small reactors
• AI's power trip will leave energy grids begging for mercy by 2027
• Datacenter developer says power issues holding up new builds

But it isn't just a case of building additional power generation. In many cases, transmission lines will need upgrading to carry the extra load. Yet as the report highlights, creating this infrastructure faces local opposition.

"Building new high-voltage transmission is not an easy task, especially through scenic areas that encompass many of the wealthiest counties in the country – including Loudoun County, whose residents have organized to oppose the construction of additional transmission lines in their community," it notes. The local firm responsible, PJM, warned that energy reliability will degrade if upgrades are unable to meet the completion deadline.

JKF argues that an increase in energy demand need not necessarily increase the price of energy over the long term, if suppliers are able to bring more output onto the market at prices close to current ones. In other words, market forces should ensure more natural gas, wind, and solar energy is made available to meet demand.

Its authors recommend policies to prioritize accountability and transparency for datacenters, so that policymakers can understand their impact on the energy grid in order to take action.

Perhaps controversially, it says that AI companies should be required to bear the additional costs of the energy they consume to ease the burden on ordinary Americans and small businesses. This could be implemented by charging datacenter operators higher fees to reflect their disproportionate impact on electricity markets.

It also recommends that US states should deter local governments from subsidizing the construction of datacenters, which create few local jobs and deliver little economic benefit for the areas they are sited in.

But most importantly, steps must be taken to increase the capacity and stability of the grid.

"The average American household could pay over a thousand dollars more each year for electricity by the end of the decade if we don't take urgent steps to resolve these energy shortages," claimed Ike Brannon, co-author of the study and a senior fellow at the Jack Kemp Foundation. ®