AI’s explosive growth demands vast amounts of power. Big Tech is turning to renewables, gas and nuclear, while some see the ultimate solution in space.
Gas turbines at xAI’s data center on Riverport Road in Memphis, Tennessee, reflect the AI industry’s growing reliance on quickly deployable power as rising electricity demand strains existing grids. Photo: Brandon Dill for The Washington Post via Getty Images
|
The AI boom is becoming an energy boom. The data centers needed to train and run increasingly powerful models consumed around 415 TWh of electricity in 2024, equivalent to about 1.5% of global electricity consumption. The International Energy Agency (IEA) expects that figure to more than double to around 945 TWh by 2030.
Technology companies building this infrastructure are spending hundreds of billions of dollars to secure more computing capacity. But buying more chips and constructing more data centers solves only part of the problem. The facilities also need large amounts of reliable electricity.
The IEA expects renewables to provide nearly half of the additional electricity required by data centers through 2030. Solar and wind are increasingly cheap and can be deployed relatively quickly, making them an important part of the solution. But data centers operate around the clock, while renewable generation fluctuates with weather and daylight.
That mismatch is pushing technology companies toward energy sources capable of providing power when renewables cannot.
Source: Gartner, June 2026
Gas Offers the Fastest Route
Natural gas provides an immediate answer. Gas turbines can supply large amounts of electricity around the clock and can be installed much faster than major new nuclear plants or transmission infrastructure.
Elon Musk’s xAI has already turned to gas turbines to supply electricity for its rapidly expanding computing infrastructure around Memphis, Tennessee. Generating electricity close to a data center can allow operators to add computing capacity without waiting years for utilities to expand grid infrastructure.
But that speed comes at a cost. Burning natural gas produces carbon emissions and other pollutants, creating a potential conflict between the technology industry’s climate commitments and its race to expand AI infrastructure.
The turbines used by xAI have also generated local opposition, with environmental groups alleging that their emissions worsen air pollution in communities around its facilities.
That leaves Big Tech searching for another reliable source of electricity with much lower carbon emissions: nuclear power.
Nuclear energy is increasingly attractive because reactors can provide large amounts of electricity around the clock without the direct carbon emissions associated with gas generation.
The difficulty is speed. Conventional nuclear plants can take many years to permit and construct, making them poorly matched to an AI industry in which new data centers can become operational within two to three years.
Technology companies are therefore looking beyond conventional reactors. Small modular reactors (SMRs) promise smaller facilities that could eventually be manufactured and deployed more rapidly. The IEA says technology companies have announced plans to finance more than 20 GW of SMRs, although significant deployment is unlikely before the end of the decade.
Fusion offers an even more ambitious possibility, potentially providing enormous amounts of low-carbon energy. But despite billions of dollars flowing into fusion companies, commercial deployment remains uncertain and is unlikely to solve the immediate power shortage facing AI infrastructure.
For the next several years, the industry will therefore rely on a mixture of technologies. The IEA expects renewables to provide the largest share of additional data-center electricity demand through 2030, followed by natural gas and coal, with nuclear becoming increasingly important toward the end of the decade and beyond.
Many Do Not Want Data Centers Nearby
Generating enough electricity is not the only obstacle.
Despite their growing strategic importance, data centers face a basic problem: they are noisy, energy-intensive and can have significant environmental impacts.
Seven in 10 Americans oppose the construction of AI data centers in their local area, according to Gallup. Some 56% of Democrats strongly opposed such projects, compared with 39% of Republicans.
This creates an increasingly difficult political problem. Governments want more computing infrastructure, while the communities asked to host it may resist the additional electricity infrastructure, noise, water consumption and pollution associated with large facilities.
Washington increasingly considers the issue one of national security. In May 2025, President Donald Trump ordered certain AI data centers associated with Department of Energy facilities to be designated as “critical defense facilities” where appropriate. The order specifically identified advanced nuclear reactors as a potential source of resilient electricity for AI infrastructure and national-security installations.
That connection became even more explicit in June 2026, when a National Security Presidential Memorandum called for a roadmap to ensure the US national-security apparatus has access to sufficient advanced computing resources, including high-security AI computing facilities.
The result is a growing contradiction: infrastructure Washington considers increasingly important to national security and technological competition is often the same infrastructure voters do not want built nearby.
If electricity, grids, land and local opposition limit how far data centers can expand on Earth, one radical solution is to move some computing off the planet altogether.
Space offers one particularly attractive resource: solar energy. Google’s Project Suncatcher is exploring constellations of solar-powered satellites carrying AI processors and connected through high-speed optical links.
Google estimates that, in the right orbit, a solar panel could be up to eight times more productive than on Earth and generate electricity almost continuously. Computing could therefore take place in orbit without consuming terrestrial land, water or grid capacity.
The engineering challenges, however, are formidable. Hardware is expensive to launch and difficult to repair, radiation can damage processors and cooling powerful computers in a vacuum requires large radiators.
Space debris presents another problem. Large constellations of computing satellites would add more objects to already congested orbits, increasing collision risks and requiring sophisticated avoidance and deorbiting systems.
Space-based data centers are therefore unlikely to replace terrestrial facilities anytime soon. The AI industry’s immediate future will instead be decided much closer to home.
Renewables offer the fastest-growing source of electricity for data centers. Natural gas offers speed and reliability. Nuclear promises continuous low-carbon power but requires longer development times. If those terrestrial solutions eventually run into hard limits imposed by grids, land, water and political acceptance, space represents the most radical final step.
The race for artificial intelligence is no longer simply a race for better chips. Increasingly, it is a race to secure the energy needed to keep them running.
