Sometimes the boldest ideas start small — in this case, roughly the size of a kitchen appliance. Google just put a refrigerator-sized satellite into space, part of a research initiative the company hopes could eventually pave the way for AI data centers orbiting Earth, powered entirely by the sun.

Chips Built for Earth, Now Tested in Space

The prototype satellite carries four specialty chips called Tensor Processing Units, or TPUs, designed by Google for machine learning and already deployed widely in the company's terrestrial data centers. In orbit, these chips will run a version of Google's Gemma AI model for 15-minute stretches at a time, limited by heat management constraints, using the open-weight model to answer simple queries. In a blog post, Google said the mission aims to measure how well the TPUs withstand "the physical stress of spaceflight and the radiation and thermal extremes of space."

An Industry Racing to Keep Up With Demand

AI usage has surged dramatically over the past few years, fueling an insatiable appetite for data center capacity. At Google's annual developer summit in May, CEO Sundar Pichai said demand for AI services currently exceeds available supply, and he expects the company's capital expenditures this year to reach between $180 billion and $190 billion — more than six times larger than spending back in 2022 — as Google races to build out computing infrastructure and develop new chips.

Escaping Earthbound Opposition

As resistance to power-hungry data centers continues growing here on Earth, Google finds itself among a small handful of companies exploring an entirely different frontier: putting these facilities into orbit, where energy from the sun is essentially limitless and free, and where, notably, there are no protesters lining up outside.

Tapping Into the Solar System's Real Power Source

"The sun puts out almost all of the power in our solar system. All of the other power sources that humanity has tapped into are just a tiny fraction of a percent," said Travis Beals, senior director and lead of Project Suncatcher. "So in some sense, this project is about tapping into the best way to use solar power to run AI compute."

Designed to Chase the Sun Forever

The satellite will orbit in a sun-synchronous path, positioned so its solar panels will almost never sit in shadow — a design choice that eliminates the need for heavy batteries or backup power systems aboard the spacecraft entirely.

Google Isn't Alone in This Race

Last November, venture-backed startup Starcloud launched a spacecraft carrying an Nvidia H100 chip, successfully demonstrating a version of Google's Gemini AI running from orbit. Elon Musk's SpaceX is pursuing similar ambitions, working on space-based data centers and stating it expects to begin deploying "orbital AI compute satellites" as early as 2028.

Building Toward Satellite Clusters

Google, partnering with aerospace and satellite imagery company Planet, envisions clusters of satellites eventually orbiting together to form fully functioning space-based data centers. Each satellite would carry dozens of TPU chips, communicating with one another — and with Earth — via laser links. Next year, the company plans to launch two additional satellites specifically to test that inter-satellite connection.

A Modest First Step With Big Implications

The satellite launched today represents a prototype Beals described as "a very minimal test," designed simply to confirm the chips can function properly in space at all. A SpaceX rocket carried it into orbit, with Planet handling the process of getting it operational. From there, Beals said, Google will begin powering up and testing the TPUs, with the mission expected to remain operational for roughly a year.

A Concept That Sounds Like Science Fiction

Brandon Lucia, a professor of electrical and computer engineering at Carnegie Mellon University, said the idea sounds "very sci-fi," though he noted much of the underlying technology has already been proven at a research level.

The Heat Problem That Won't Go Away

Still, enormous challenges remain. Heat ranks among the biggest. Computer chips generate significant heat, and on Earth, that heat typically dissipates through wind or water cooling — options that simply don't exist in the vacuum of space.

"In a satellite in particular, using more power to do the computations means dumping more heat into the confined environment inside of the satellite," Lucia explained — a problem with nowhere for that excess heat to actually go.

Google's Approach to Keeping Things Cool

In its blog post, Google called cooling the chips "a crucial research challenge," outlining several approaches currently under development, including a combination of pipes and radiators designed to wick heat away from the chips. Beals acknowledged that these radiators currently rank among the heaviest components on the mission — a notable drawback, since heavier payloads cost considerably more to launch. His team, he said, is actively working on more efficient cooling methods.

Who Fixes a Satellite When Something Breaks?

Lucia also raised significant questions about long-term maintenance. When a chip malfunctions inside a terrestrial data center, a technician can simply walk in and fix it. That option disappears entirely when the data center in question is orbiting hundreds of miles above Earth.

"Those all have their cost and complexity amplified by a factor of 10, maybe a factor of 100. And so there has to be a big payoff," Lucia said — a reminder that even minor technical hiccups become exponentially more complicated once gravity is no longer an option.

Environmental Questions Linger Too

Beyond technical hurdles, environmental concerns also loom over the project. While space-based data centers would run entirely on solar power, the rockets launching them still burn carbon-based fuel. Eventually, these satellites will re-enter Earth's atmosphere, potentially generating additional pollution as they burn up on reentry.

A "Moonshot" With an Uncertain Timeline

Elon Musk has said space represents the "only way to scale" data centers to meet future AI demand. Yet the underlying economics remain a genuine question mark — launches are expensive, repairs are enormously difficult, and laser-based communication across hundreds of miles through Earth's atmosphere is inherently slower than terrestrial alternatives. Achieving computing capacity equivalent to a single Earth-based data center would require entire constellations of satellites working in tandem.

Google's Own Team Weighs the Trade-offs

Beals said he and his small team grapple with these exact questions regularly. "It doesn't help a lot if this is technically possible, if it's always going to be too expensive to be practical," he said. He added that predicting when space-based data centers might become economically viable remains genuinely difficult. "I don't see this being something where it's cheaper to do this in the next five years. I think it will take longer than that," he said.

Coexistence, Not Replacement

Beals doesn't anticipate space-based data centers replacing their Earthbound counterparts anytime soon. "I think that there are going to be use cases that make more sense to do on Earth and use cases that make more sense to do in space," he said. "In the really-long run, perhaps the set of things you do in space gets larger and larger until that's where most of the compute is happening. But I think for a long time you'll have both coexisting."

He described the entire undertaking plainly as a "moonshot." "It's going to take a long time," Beals said. "It's not going to be a completely straightforward journey" — a fitting way to describe a project that, quite literally, is reaching beyond Earth itself in search of AI's next frontier.