Google Is Testing AI Data Centers in Space With Project Suncatcher

In a time when data centers are at the center of many of the world’s debates, Google is looking for a way to take its data centers out of this world and into space. Google’s Project Suncatcher is an initiative designed to explore whether space could support large-scale machine-learning infrastructure.

According to Google officials, the first satellite isn’t a full-fledged data center. In fact, Google says that it has no plans to have a functional data center in orbit within the next few years. Instead, the mission is designed to determine what works, what fails, and what the company would need to solve before putting significantly more computing power into orbit.

An AI data center in space seems like something out of a science fiction movie, but Google’s Project Suncatcher will determine just how possible it is to make it a reality.

Google Is Starting With A Prototype

Google’s Project Suncatcher includes a single satellite that will launch on October 1, 2026, from Vandenberg Space Force Base in California. The satellite will be carried into orbit by a SpaceX Falcon 9. Instead of trying to put an entire data center into orbit, Google is launching an experimental satellite to test its AI hardware under space conditions.

Google first announced Project Suncatcher in November 2025, and said that it was focused on determining whether space could eventually host scalable machine learning. The company has described future versions as constellations of satellites carrying its Tensor Processing Units, or TPUs, and communicating with one another through high-speed optical links.

At its core, the first satellite is a test for a much bigger idea. Google wants to find out how its computing hardware behaves when it has to deal with radiation, extreme temperatures, launch vibrations, and the unusual thermal environment of space.

Space Offers One Very Big Advantage

Data centers are a hot-button topic, thanks in large part to the amount of natural resources that they require. Many communities don’t want data centers taking up large lots of local land, and even when there’s room for them, there are concerns about the amount of water that they require.

However, preserving natural resources is not the primary reason for Google’s venture into orbit. Instead, the experiment revolves around energy. Satellites operating in the right low-Earth orbit can receive sunlight for much longer periods than solar panels on the ground, potentially generating substantially more power over time. According to preliminary studies on the topic, Google believes that satellite-based AI may be able to generate up to eight times more solar power than it does on Earth because they can receive near-constant sunlight.

Google's AI Chips Have To Survive Space

AI programs rely on chips to store, retrieve, and produce information. Google has to make sure that its chips are capable of withstanding the harsh conditions of space. When it comes to Project Suncatcher, the company has been testing its Trillium TPUs to determine how they respond to the radiation environment they would encounter in orbit.

Google’s leadership says that the early results have been promising. The company claims that its Trillium TPUs have shown that they can withstand higher doses of radiation than they would face during a five-year space mission. Still, testing the TPUs in a lab is simply not the same thing as putting them on a rocket and launching them into space. What looks promising in a controlled laboratory setting may not be transferable to the unpredictable conditions that those chips would face while in orbit.

Google's Space Data Center Is Still A Long Way Off

Google’s Project Suncatcher is an intriguing example of just how far artificial intelligence is pushing the boundaries of the tech industry as we know it. The latest initiative isn’t about making better Google AI chips or finding ways to make data centers more eco-friendly. It’s about taking what we know about data centers and trying to translate that information into satellites that could do the work of a data center from Earth’s lower orbit.

The October 1 launch is only the beginning. The company first needs to prove that its hardware can survive in orbit, that it can be cooled, that satellites can exchange data quickly enough, and that the economics can work. Future missions will build on those tests, including a planned two-satellite mission in 2027 focused on high-bandwidth optical communication. For now, the AI data center in space is a hypothesis that’s being tested. But the results of this test may reshape the future of AI.

Looking for stories that inform and engage? From breaking headlines to fresh perspectives, WaveNewsToday has more to explore.