Now tech moguls want to build data centres in outer space

Energy constraints in the artificial-intelligence race are causing high-technology entrepreneurs to start thinking out of this world.

Astronaut Edwin ‘Buzz’ Aldrin, the second man to walk on the moon during the 1969 Apollo 11 mission.

If things weren’t already frothy enough around artificial intelligence, now the excitement is headed towards the moon – literally.

The world’s richest men are earnestly talking about travelling to outer space to build gigantic data centres to run AI models among the stars. They argue such missions make the most sense for powering energy-hungry operations.

“The moon is a gift from the universe,” Jeff Bezos recently said when talking about the benefits of lunar development and using it as a base for launching projects in space.

Such talk comes as his Blue Origin and Elon Musk’s SpaceX rocket companies are working to make space travel cheaper and routine.

It isn’t clear what’s closer to being real: moon bases or superintelligent AI? But the two seem to be converging in an era of investor enthusiasm that has some worried we are in an AI bubble.

To be clear, the present economics of space-based data centres don’t make sense. But they could in the future, perhaps as soon as a decade or so from now, according to an analysis by Phil Metzger, a research professor at the University of Central Florida and formerly of the NASA.

Electricity demand is surging, driven by artificial intelligence and a boom in data centre construction. Some warn the US grid will not be able to keep up. But Scott Strazik, chief executive of GE Vernova, says his company can deliver.

“Space enthusiasts have long sought a business case to enable human migration beyond our home world,” he posted on X amid the new hype. “I think AI servers in space is the first real business case that will lead to many more.”

Back on Earth, President Donald Trump has already declared an energy emergency and his administration says we need to add massive amounts of additional capacity to handle the expected power demands coming in the next few years.

Meanwhile, AI companies are building out their own energy-generating capabilities as they wait for the power grid to catch up. Musk’s xAI, for example, has been using gas turbines as temporary power sources. And OpenAI is pushing for the US government to partner with companies to add 100 gigawatts a year.

Let’s put 100GW into perspective. Forty years ago, when Back to the Future came out, a key plot point in the movie was the need to generate such an incredible amount of energy that time travel seemed plausible to an audience.

The story revolved around needing to find 1.21GW of power, or the equivalent of a lighting strike, to send Doc Brown’s famed DeLorean time machine back home.

Now, 1GW, which the US Department of Energy once noted was roughly half the power generated by the Hoover Dam, seems paltry. And Musk, Bezos and even Alphabet chief executive Sundar Pichai are saying stuff that sounds like pure science fiction for a new generation.

The argument essentially boils down to the belief that AI’s needs are eventually going to grow so great that we need to move to outer space. There, the sun’s power can be harvested more efficiently.

In space, the sun’s rays can be direct and constant for solar panels to collect – no clouds, no rainstorms, no night-time. Demands for cooling also could be cut because of the vacuum of space.

Plus, there aren’t those pesky regulations that executives like to complain about, slowing construction of new power plants to meet the data-centre needs. In space, no one can hear the Nimbys scream.

“We will be able to beat the cost of terrestrial data centres in space in the next couple of decades,” Bezos said at a tech conference last month. “Space will end up being one of the places that keeps making Earth better.”

It’s still early days. At Alphabet, Google’s plans sound almost conservative. The search engine company in recent days announced Project Suncatcher, which it describes as a moonshot project to scale machine learning in space. It plans to launch two prototype satellites by early 2027 to test its hardware in orbit.

“Like any moonshot, it’s going to require us to solve a lot of complex engineering challenges,” Pichai posted on social media.

Nvidia, too, has announced a partnership with start-up Starcloud to work on space-based data centres.

Not to be outdone, Musk has been painting his own updated vision for the heavens.

He has long reached for Mars, a main driver of SpaceX. But in recent weeks he has been talking more about how he can use his spaceships to deploy new versions of his solar-powered Starlink satellites equipped with high-speed lasers to build out in-space data centres.

Last week, Musk further reiterated how those AI satellites would be able to generate 100GW of annual solar power – or what he said would be roughly a quarter of what the US consumes in average in a year.

“We have a plan mapped out to do it,” he told investor Ron Baron during an event. “It gets crazy.”

Previously, he has suggested he was four to five years away from that ability. He also has touted even wilder ideas, saying on X that 100 terawatts a year “is possible from a lunar base producing solar-powered AI satellites locally and accelerating them to escape velocity with a mass driver”.

Simply put, he’s suggesting a moon base will crank out satellites and throw them into orbit with a catapult. And those satellites’ solar panels would generate 100,000GW a year.

“I think we’ll see intelligence continue to scale all the way up to where … most of the power of the sun is harnessed for compute,” Musk told a tech conference in September.

Yes, it all sounds hard to believe. Yet a few years ago Musk was drawing eye rolls as he jawboned energy industry officials to boost their capabilities, warning we were headed to shortages with the rise of demand for AI, electric cars and other tech. Now we’re looking for solutions in outer space.

As they say in Back to the Future: If you put your mind to it, you can accomplish anything. Maybe.

The Wall Street Journal