The Global Race for Electricity: AI Is Changing the World’s Energy Map

The artificial-intelligence revolution is creating an unexpected global competition — not simply for computer chips and engineers, but for something far more fundamental: electricity.

For much of the digital age, technological power was measured in processors, software and data. But as artificial intelligence expands at extraordinary speed, another resource is becoming just as important: the electricity required to run it.

Around the world, technology companies, utilities and governments are confronting a reality that could reshape the global energy map. The massive data centers required to train and operate increasingly sophisticated AI systems consume extraordinary amounts of electricity, turning access to reliable power into a strategic economic advantage.

Goldman Sachs Research has projected that global data-center power demand could increase roughly 160 to 165 percent by 2030 compared with 2023 levels. The International Energy Agency similarly expects worldwide data-center electricity consumption to more than double by the end of the decade, reaching roughly 945 terawatt-hours annually — slightly more electricity than Japan consumes today.

The numbers are beginning to change the economics of the technology industry.

For years, companies deciding where to construct data centers primarily considered land prices, taxes, fiber-optic connections and proximity to customers. Increasingly, they must ask another question: Where can we find enough electricity?

Some proposed AI campuses require power on the scale of entire cities. S&P Global reported this summer that individual AI computing racks can consume as much electricity as 80 to 100 homes, while gigawatt-scale data-center campuses can demand electricity comparable to a city such as Orlando.

That demand is creating pressure on electrical grids that were never designed for such rapid growth.

The problem became particularly visible this week when PJM Interconnection, America's largest grid operator, proposed measures that could require data centers to rely on backup generation during emergencies when the electrical system approaches its limits. PJM serves approximately 67 million people across a region stretching from Washington to Chicago.

The implications stretch far beyond the United States.

Countries capable of producing abundant, dependable and inexpensive electricity may suddenly possess an enormous advantage in the AI economy. Regions with plentiful natural gas, nuclear generation, hydroelectric resources, solar power or other reliable energy sources could become magnets for billions of dollars in data-center investment.

The race is also reviving technologies that only recently appeared politically or economically uncertain.

Nuclear power is attracting renewed attention because reactors can provide enormous quantities of around-the-clock electricity without the intermittency associated with wind and solar. Natural gas remains attractive because new generation can sometimes be developed more quickly. Renewables are expanding as technology companies attempt to meet both electricity requirements and emissions goals.

The IEA estimates that renewables will provide nearly half of the additional electricity required by data centers through 2030, followed by natural gas and coal, with nuclear power playing a growing role toward the end of the decade and beyond.

Some technology companies are no longer willing to wait for utilities. Developers are increasingly considering private generation, microgrids, batteries and other “bring your own power” strategies that allow enormous computing campuses to operate partly outside conventional electrical systems.

That could represent a fundamental shift in the geography of technological power.

For decades, oil fields, pipelines and shipping lanes helped determine which nations possessed geopolitical leverage. In the AI era, electrical grids, power plants, transformers and data centers may acquire similar strategic importance.

The countries that dominate artificial intelligence may ultimately be determined not only by who builds the best computer chips or develops the smartest algorithms.

It may also depend on who can keep the lights on.