What happens when artificial intelligence needs more electricity than Earth’s infrastructure can easily provide? Google is experimenting with an extraordinary answer.
What happened?
On October 1, Google took an unusual step toward the future of artificial intelligence: sending AI chips into space.
Through Project Suncatcher, Google launched an experimental satellite built with Planet Labs aboard a SpaceX Falcon 9 rocket, as part of the Transporter-18 rideshare mission. Google · Space.com
According to The New York Times, the satellite carries four of Google’s Tensor Processing Units (TPUs), specialized chips designed to power artificial intelligence. The New York Times
The goal isn’t to build a working orbital data center overnight. It’s to determine whether AI hardware can survive and operate in space.
If successful, the experiment could eventually lead to networks of solar-powered AI satellites orbiting Earth.
Why does it matter?
Artificial intelligence requires enormous computing power. And computing power requires electricity.
As AI adoption accelerates, technology companies face growing challenges involving electricity supply, cooling infrastructure and the construction of new data centers.
Google is exploring whether space could offer an alternative.
Satellites in certain orbits can receive nearly continuous sunlight, potentially generating substantial solar electricity without depending on terrestrial power grids. Google
However, there’s a catch.
Space is cold, but cooling computers in space is surprisingly difficult.
Without air to carry heat away, satellites must rely on specialized radiators.
According to The New York Times, Google’s initial experiment can run its AI chips for only about 15 minutes at a time before they must shut down to let the cooling system catch up. The New York Times · Google
That’s a reminder of how experimental this technology remains.
The bigger picture: moving infrastructure beyond Earth
Here’s where things get particularly interesting.
If orbital computing eventually becomes practical, we could begin moving certain computing workloads away from Earth.
Imagine satellites processing satellite imagery directly in orbit rather than transmitting enormous quantities of raw information to terrestrial data centers.
Future applications could include environmental monitoring, disaster detection and scientific research.
However, moving large-scale AI computing into space introduces new challenges, including launch costs, radiation, communications, orbital debris and equipment maintenance.
The opportunity is substantial, but the engineering obstacles are equally real.
The investment connection
The obvious company to watch is Google’s parent, Alphabet.
But the potential implications extend beyond AI developers.
Space infrastructure: Companies such as Planet Labs and Rocket Lab could benefit from growing demand for satellites, launches and related services.
Advanced hardware: More demanding space applications could create opportunities for specialized semiconductors, radiation-resistant electronics and thermal-management technology.
Communications: Large networks of computing satellites would require fast, reliable connections between spacecraft.
These are potential long-term opportunities, not established revenue streams. Orbital data centers must demonstrate both technical reliability and commercial viability before investors can reasonably assess their economic significance.
What we’re watching next
Google plans its next experiment in 2027, when it intends to put two satellites in orbit to test high-speed laser communications between them. Google
Three questions will matter:
- Can AI chips operate reliably for extended periods in orbit?
- Can orbital cooling systems handle increasingly powerful hardware?
- Can launch and operating costs eventually compete with terrestrial data centers?
The bigger question isn’t whether we can put AI in space. It’s whether doing so makes economic sense.
And that answer could determine whether Project Suncatcher remains an ambitious experiment or becomes an entirely new computing industry.
Stay sharp, stay organized.
Educational commentary, not personalized investment advice. Companies named are examples, not recommendations. Source: Google’s October 1 announcement.