Musk vows to put data centers in space and run them on solar power but experts have their doubts

Elon Musk is doing what he always does: coming up with a crazy idea that sounds like it came from a science fiction book. This week, the man who changed the car industry with Tesla and space travel with SpaceX said he wants to change the world of data centers by putting them in space.

Yes, you read that right: big computer centers powered by solar energy floating in space, running on the sun all the time, to meet the growing need for AI without using up Earth's power grids or making everyone's electric bills go up.

On Monday, Musk put SpaceX and his AI company, xAI, together to get things going. He is also getting ready for a big IPO to pay for everything. He wants to send up as many as a million satellites that would work as these data centers in space.

He wrote on the SpaceX website, "Space-based AI is obviously the only way to scale," and then he made a joke about how "it's always sunny in space!" What is the point? Let AI tools like chatbots and machine learning models run free without causing blackouts or turning huge areas of land into server farms that get too hot and drink up all the water they can find.

But hold on—experts are putting the brakes on hard, even though Musk has a great track record (he outsmarted Detroit to make Tesla the best electric car company). People in science, engineering, and business say that this idea has a lot of problems to deal with, such as high costs, environmental risks, and technology that doesn't work well in space.

They don't hate the idea completely; they just think it might be harder to make it happen than landing a rocket on a drone ship in the ocean.

First, let's talk about the heat problem, which is a big one. The servers and chips in data centers on Earth already make a lot of heat. We cool them off here with big fans, air conditioning, or even by putting them in liquids.

But what about the space? It's a vacuum, which means it's really cold on one side. But because there's no air, heat doesn't escape like it does in air. Like a thermos bottle, the vacuum keeps the heat in, which keeps your coffee hot.

If it wasn't cooled, a computer chip up there would fry itself in no time. Josep Jornet, a professor at Northeastern University who knows a lot about electrical engineering, said it clearly: "A computer chip in space that isn't cooled would overheat and melt much faster than one on Earth."

There is a way around it: big radiator panels that send heat into space as infrared light. They've used smaller versions on the International Space Station, and it works for some installations.

But what do you think of Musk's idea of big data farms in space? You'd need to build huge, fragile buildings that have never been built before. We're talking about weak panels that could cover football fields and move at speeds of thousands of miles per hour.

One micrometeorite hit or solar flare, and boom—big problems. Musk doesn't care because he's Musk. He said again, "You can mark my words," in a preview of his Cheeky Pint podcast episode that comes out on Thursday.

Space will be the best place to put AI in 36 months, although it will probably be closer to 30 months. And then being in space will be so much better. Confidence level: always through the roof.

Then there's the scary problem of trash in space. There are already a lot of dead satellites, rocket parts, and other random things floating around in low Earth orbit.

There are about 10,000 satellites in space right now as part of Musk's Starlink network. His filings say that in the last seven years, he has only had one small "debris-generating event."

But how do you get to a million? That's a sure way to fail. A small crash could start the Kessler syndrome, which is a chain reaction of crashes that turns orbit into a shooting gallery.

John Crassidis, a former NASA engineer who now works at the University at Buffalo, says we're getting close to a "tipping point" where the chance of a collision will be too high.

These things move through space at 17,500 miles per hour. If they hit something up there, it's not like bumping cars in a parking lot. It's explosive and violent, and it could knock out important services like GPS, weather satellites, or emergency communications.

Musk's idea makes the risk much higher, and it's hard to pick up space trash because there aren't any orbital tow trucks yet.

It can also be hard to fix things in space. You send a tech person to fix a broken server or a bad GPU chip (the powerful graphics cards that AI needs). Simple.

In space? Best of luck. No one is going to climb a ladder to fix a satellite. Cosmic rays and solar particles can kill those GPUs, so they are not safe.

Baiju Bhatt, the CEO of Aetherflux, a company that works on solar technology in space, says that you should over-provision by adding extra processors as backups.

But that's expensive; each GPU costs tens of thousands of dollars, and Starlink satellites only last about five years before they burn up when they come back down.

When you multiply that by a million, the costs go through the roof—no pun intended.

This is a big problem for your money. Putting SpaceX and xAI together for an IPO could make billions of dollars, but making and launching a million satellites? We're talking about trillions, even though Musk's reusable rockets make things cheaper.

And that's not even counting the costs of research and development for those big processors or radiators that can handle radiation.

Environmental concerns are another layer: more launches mean more rocket exhaust in the air, which some scientists think could have an effect on the climate or the ozone layer.

And all that solar power in space is great in theory, but it's not easy to send it back or use it well for AI calculations.

Musk isn't the only player in this game, though. There will be competition soon. In November of last year, Starcloud from Redmond, Washington, sent one Nvidia AI processor into space to see how it would work.

Google is working on Project Suncatcher, which studies data centers in space. Blue Origin, which is owned by Jeff Bezos, is also getting ready to launch a group of 5,000 satellites late next year.

This one will be more about talking to people. But Musk has one big advantage: his rockets.

Musk's competitors, like Starcloud and Aetherflux, are using Falcon launches, but according to Pierre Lionnet from Eurospace, Musk charges them a lot more—up to $20,000 per kilo of cargo compared to $2,000 for his own things.

Lionnet believes that this is a show of strength; Musk's words are like a warning shot to his rivals. "It's a way of telling the others that I will keep these low launch costs for myself when he says we're going to put these data centers in space."

It's a game of power. In short, he's using his launch monopoly to win the orbital AI race before it even starts.

So, what's the big deal? This could change the game if Musk can do it. AI is growing quickly, and chatbots, image generators, and self-driving cars all need a lot of computer power.

The systems on Earth are under a lot of stress. Data centers use as much power as small countries do now, and they need a lot of water to stay cool.

There is an endless amount of solar energy in space (no clouds, no night), and the cold vacuum could make it easier to cool down once you figure out how to get rid of the heat-trapping problem.

It could stop blackouts, save money in the long run, and make AI more accessible to more people without hurting the environment.

But the doubts are real. Experts like Jornet, Crassidis, and Bhatt aren't completely dismissing it; they're just pointing out the engineering Everest that lies ahead.

Musk has done things that seemed impossible before, like making reusable rockets that others thought were impossible. Those who didn't believe in Tesla had to eat crow.

But space is harsh; one single mistake may ruin the whole idea.

This is important for regular people because AI is becoming a part of everything, including your job, your phone, and your car. Musk's 30 to 36 month timeline seems ambitious, but that's how he works: he makes big promises and then delivers something close enough to wow everyone.

This is classic Musk in the end, whether he's a visionary genius or just too full of himself. He believes that space can help fix problems on Earth, and the world will be watching.

If he succeeds, it could change computers in the same way that he changed rockets and electric cars. If not, it'll still be a great story. The race to get AI into space is on, and it's going to be a crazy ride.

Jennifer Chen

Jennifer Chen is a senior business correspondent covering Wall Street, corporate America, and economic trends. A former financial analyst, she brings insider expertise to stories about markets, mergers, startups, and the intersection of business and technology.