MW Nvidia's Jensen Huang is the latest tech CEO to cheer the potential of data centers in outer space

By Emily Bary

CEO acknowledges that 'the economics are poor today' but should 'improve over time'

"Artificial intelligence in space will have very good, very interesting applications," according to Nvidia CEO Jensen Huang.

Chips aren't the only thing in short supply as companies engage in a mad dash to develop artificial-intelligence applications. Major AI players are also struggling to find enough space and power for their gigantic data centers - without encountering backlash from local communities.

The dynamic has become enough of a stumbling block that some tech executives, including Tesla's $(TSLA)$ Elon Musk, Amazon.com (AMZN) founder Jeff Bezos and Alphabet's $(GOOG)$ $(GOOGL)$ Sundar Pichai, have floated the possibility of putting data centers in outer space, where temperatures are cooler and the environment is, suffice to say, much less crowded.

More from MarketWatch: Why Musk and other tech leaders think outer space can help solve one of AI's biggest challenges

And on Wednesday, Nvidia (NVDA) CEO Jensen Huang lent his voice to the chorus. "Artificial intelligence in space will have very good, very interesting applications," he said in response to analyst's question on the company's earnings call, according to a transcript provided by AlphaSense.

Of course, there are plenty of practical considerations to take into account. "Well, the economics are poor today," Huang conceded, but they're "going to improve over time."

Read: Nvidia's blowout earnings guidance is met with a shrug. Here's what to know.

"The way that space works is radically different than how it works down here," he continued, noting various puts and takes. There's a lot of energy in space, and also "plenty of space in space," making it a potentially good spot for large solar panels.

While the colder climate of space holds appeal for some backers of the galactic data-center vision, "there's no airflow," Huang noted, which could present challenges when it comes to dissipating heat. The liquid-cooling infrastructure currently used on earth for that purpose is "heavy and freezes" and therefore "obviously out of the question."

But Huang seemed to see the appeal of the idea when he looked past some of the practical hurdles. He gave the example of images that have extremely high resolutions and consume "petabytes and petabytes" of data. That data could perhaps be processed in outer space more efficiently, he offered, allowing the more important analysis to take place on Earth.

Starcloud, a startup that focuses on putting data centers in space, sent one of Nvidia's older, H100 graphics processing units into orbit late last year. "The 60-kilogram Starcloud-1 satellite, about the size of a small fridge, is expected to offer 100x more powerful GPU compute than any previous space-based operation," Nvidia said in a blog post at the time.

For Tesla's Musk, space-based data centers appear to be well more than just a thought experiment. SpaceX, his rocket-launch company, wants government permission for a constellation consisting of a million satellites that could be used for orbital data centers.

Don't miss: These space stocks could benefit from Elon Musk's latest AI obsession

"The company pitches it as a way for power-intensive AI compute to benefit from near-constant solar power with little or no post-launch maintenance costs while reducing the environmental impact of AI on Earth," MoffettNathanson analyst Nick Del Deo wrote earlier this month.

Achieving Musk's vision is likely easier said than done. "The capital needs would be simply enormous," wrote Del Deo, who estimated that the company might have to devote something like $5 trillion in capital expenditures annually toward the endeavor. That would be good news for Nvidia, he suggested, as the budget might fall to a slightly less outrageous $2 trillion or so if SpaceX were to outfit these data centers with cheaper custom chips.

See more: Elon Musk's plans for SpaceX data centers in orbit could cost $5 trillion a year, analysts say

The goal isn't impossible, he concluded: "But there are, to state the obvious, substantial challenges to building a constellation along the lines of what SpaceX laid out. At a bare minimum, one can safely conclude that a full-fledged build is not happening anytime soon, given the requisite operational maturity, supply-chain development and financial requirements."

-Emily Bary

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

February 25, 2026 21:01 ET (02:01 GMT)

Copyright (c) 2026 Dow Jones & Company, Inc.