Big Tech Is Scorching The Electric Grid

 In 1946, when ENIAC, the world’s first general-purpose computer, was first turned on, it used so much power (about 174 kilowatts) that it caused the lights in Philadelphia to dim momentarily.

Six years later, John von Neumann, the mathematician and computer pioneer, unveiled MANIAC, short for Mathematical Analyzer Numerical Integrator and Automatic Computer, the first computer to use RAM (random access memory). MANIAC was far more efficient than ENIAC, drawing about 19.5 kilowatts, or one-ninth the power needed by its predecessor. Author George Dyson has written that, “the entire digital universe can be traced directly” back to MANIAC.

As I explained in May 2024, the energy efficiency of our computers has continually improved since the days of ENIAC and MANIAC. And while the efficiency of our digital machinery has increased:

The power hungry nature of computing has not. The Computer Age has been defined by the quest for ever-more computing power and ever-increasing amounts of electricity to fuel our insatiable desire for more digital horsepower. As data centers have grown over the past two decades, concerns about power availability have surged.

 

Today, the US is facing an unprecedented power crunch. After two decades of flat electricity demand, power use is suddenly soaring as the world’s biggest tech companies race to build massive data centers running thousands of AI computers that will use stunning volumes of juice. On August 11, the Electric Power Research Institute estimated that AI’s existing power demand is approximately 5 gigawatts, but that demand could reach 50 GW by 2030. For perspective, 50 GW is larger than the total electric generation capacity in Pennsylvania, which has a population of 13 million people and is the fifth most-populous state in America. (Pennsylvania currently has about 49 GW of electric generation capacity.)

On August 27, Monitoring Analytics, the independent market monitor for PJM, the largest regional transmission organization in the US, warned that the costs associated with meeting the surging demand for AI could result in a “massive wealth transfer” from ratepayers to Big Tech. The way to avoid that, it said, was for data center operators to “bring their own new generation.” In other words, PJM should require Big Tech to build its own power plants rather than simply connecting to the existing grid. PJM has previously projected that it expects peak loads on its system, due to data centers, to jump by 30 GW by 2030.

 

On Wednesday, Wood Mackenzie, the global energy consulting firm, reported that data centers used about 6% of all US electricity in 2024, and that share could increase to 13% by 2030. As seen above, it’s not just the US. The consultancy said that data center demand is surging in Japan and China, and that Japan will be the “next major battleground for hyperscale data center expansion.”

The global surge in power demand for AI has led to soaring prices for generators, transformers, and other equipment needed to generate and distribute electricity. In short, Big Tech is scorching the grid. The rush to build dozens of gigawatts of generation capacity to fuel the insatiable power needs of AI has ignited a surge of inflation that is tearing through electric sector supply chains, generation plants, and distribution systems.

The proof is in the price increases.

Let’s take a look.

 

According to the Bureau of Labor Statistics, US electricity prices have jumped by 5.5% over the past year. That may be only a soupçon of the increases to come. In May, the Energy Information Administration estimated that electricity prices will continue rising faster than inflation through 2026. In addition, the agency expects that electricity costs in regions already plagued by high power prices, including the Pacific, Mid-Atlantic, and New England, “could increase more than the national average.”

Furthermore, that 5.5% inflation figure may understate the magnitude of the increase. Paul Krugman recently reported that retail electricity prices have increased by 9% over the past six months. In New Jersey, some consumers are seeing rate increases of 20%. In Maine, prices have jumped by a whopping 37% over the past year.

Before going further, let me state the obvious: electricity prices vary widely across the US. In the hydropower-rich Northwest, consumers pay far less for power than they do in New York. Several factors determine the cost of the power that’s delivered to our homes and businesses. That said, there’s simply no doubt that the AI frenzy is supercharging these price increases.

 

Virginia ratepayers will be among the hardest hit. As shown above, by 2030, data centers in the state could need as much as 34 GW of new power capacity. Next month, Virginia’s State Corporation Commission will consider a proposal by Dominion Energy, the state’s biggest electric utility, that could raise the average customer’s bill by about 15% over the next two years. The increase is needed, Dominion says, to cover the cost of new gas-fired plants, transmission lines, and other infrastructure to meet soaring demand from data centers.

On Wednesday, a longtime grid analyst told me that she is aware of one tech giant that wants to bring more than 200 GW of data centers online over the next six years, and the company has said, “price is no object.” To preserve relationships in the power sector, the analyst, who asked not to be named, said that Big Tech’s AI push is “destroying this market in terms of pricing and availability. They are destroying the economics of this business.” The analyst then added, “Ratepayers are going to get stuck with the bill.”

 

Big Tech can outspend all of its competitors because they have staggering amounts of cash, and its valuations border on the absurd. As shown above, the five tech giants have a combined market capitalization of $14 trillion. For comparison, the combined value of the 108 largest publicly traded utilities (electricity, gas, and water) in the US totaled $1.6 trillion. The most valuable company in that group, NextEra Energy, is worth about $152 billion. By comparison, Microsoft is worth $3.4 trillion, or about 22 times more than NextEra.

In all, the combined value of the five tech behemoths is nine times greater than the combined value of 108 utilities. My friend, the grid analyst, sneered when discussing Big Tech’s dominance. “The utility guys,” the analyst said, “are bringing a butter knife to a gun fight.”

 

Share

The surge in Big Tech’s electricity use is shown in the graphic above. According to Alphabet/Google’s environmental reports, the company’s electricity use has more than doubled since 2020. That power demand will keep rising. In April, the company announced it would spend $75 billion on new data center capacity this year. In July, it announced a $6 billion data center deal in India. On Wednesday, it announced a $9 billion expansion in Virginia.

 

Microsoft is one of Google’s main rivals in the AI race. In January, Microsoft said it would spend $80 billion in fiscal 2025 on AI-enabled data centers. As seen above, Google, Microsoft, Amazon, and Meta, spent nearly $100 billion on new capital projects during the second quarter of this year. All that spending is fueling unprecedented inflation in electric sector supply chains.

That inflation is easily seen when it comes to new gas-fired power plants. In 2023, New Orleans-based Entergy announced it would build a 1,200-megawatt gas-fired plant in Bridge City, Texas, at a cost of $1.2 billion, which works out to $1 million per megawatt of capacity.

On August 26, JEA, the municipally owned electricity and water utility in Jacksonville, Florida, announced that it will sign a contract for the construction of a 675-MW combined-cycle gas-fired power plant, which will cost nearly $1.6 billion, or about $2.3 million per megawatt. (The plant will use a gas turbine built by GE Vernova.) Thus, in just two years, the cost of a new gas-fired power plant in the US has more than doubled.

In addition, wait times for new gas turbines now stand at six to seven years. That’s according to Ryan Long, an executive at Xcel Energy, who discussed the wait times during a meeting of the Minnesota Rural Electric Association earlier this month. Long confirmed the wait time to me in person after his presentation. (In May, S&P Global reported on the longer lead times for gas turbines as well as the dramatic increase in prices to build new capacity.)

High-power transformers – that is, ones that are 345kV and above -- are also in short supply, and prices are soaring. A new friend of mine who has been selling transformers for many years said, “It’s a seller’s market. I spent most of my career groveling for customers. And now, I’m telling them to take a number.” In an interview earlier this month, he stated that the wait time for new high-power units now stands at about four years, and prices have increased by more than 30% over the past two years. Wait times for high-power breakers are also in the three-to four-year range. “If a buyer balks,” he told me, “we don’t worry because we have other customers right behind them.”

Earlier this week, Mike Steffes, the CEO of ACES, a cooperative-owned management company that serves electric utilities, told me that the growth in demand for AI “is unprecedented. We’re seeing five-fold and ten-fold increases in new demand that’s coming from Big Tech. We’ve never seen anything like it before.” Steffes continued, saying the AI building boom is happening “at the same time that municipal utilities, coops, and investor-owned utilities are all working to update old equipment and do routine maintenance. This massive slug of new demand from Big Tech is increasing the cost of generation, the cost of transmission, and utility equipment, and all of it is hitting at the same time.”

 

In addition to the cost of generators and other utility equipment, consumers will face higher rates due to AI’s need for new high-voltage transmission. In April, Dominion Energy announced plans to build a 7-mile transmission line in Chesterfield County at a cost of $121 million. The project is being built to serve a “planned 900 MW hyperscale data center.” The tenant for the data center was not named. But it’s almost certain that the cost of the new power lines will be borne by Dominion’s ratepayers, not Big Tech.

Earlier this month, residents of Ashburn, Virginia, gathered to protest a proposed high-voltage transmission line that would go through their town. The proposed line is designed to serve nearby data centers. Credit: Shannon Heckt/Virginia Mercury

 

A similar move is underway in Maryland, where Potomac Edison has requested financial incentives to spend $1.1 billion on new transmission lines that will serve data centers in Virginia. On Friday, the Maryland Office of People’s Counsel filed a motion with the Federal Energy Regulatory Commission to stop the utility’s plan because it would “unfairly burden Maryland ratepayers.”

Shortages of skilled workers are also fueling inflation in the electric sector. A friend of mine who works for a large construction firm told me last week that Big Tech’s AI push has stretched his firm and others like it to the limit. The entire construction sector is “constrained by labor.” That includes “welders, dirt workers, and construction workers. Where’s that labor going to come from?” he asked. The labor crunch has forced his firm to increase its prices as a type of insurance policy. “With this volume of work,” he said, his firm needs more skilled labor. If it can’t attract enough workers, there’s a greater risk that his firm won’t be able to deliver projects on time. To counter that risk, his firm has raised its prices.

There’s more to write about the stress that the AI boom is putting on the electric sector. I’ll conclude with another quote from Steffes, who has been in the energy sector for over 40 years. “Electricity defines our country,” he told me. The global competition to build and deploy AI is a modern version of the Space Race, and the US has to win that contest. But, he added, “we have to do it in a thoughtful way.”