The moonshot idea of putting AI data centers into outer space just moved one step closer to reality, despite doubts about its feasibility.
Last week, Google’s Project Suncatcher launched a refrigerator-sized satellite carrying four AI chips into orbit on a SpaceX rocket, in an early test of whether the technology behind today’s power-hungry data centers could someday operate in space.
The satellite, developed in partnership with Planet Labs, took off on Oct. 1 and is operating as expected, according to a Google blog post. The point of the mission is to see how Google’s custom TPU chips handle the intense vibrations of launch as well as the “radiation and thermal extremes of space,” the company said. The satellite will conduct experiments in low Earth orbit for about a year, the New York Times reported.
Google’s satellite is powered by solar panels that harness energy from the sun. Because space makes it difficult to dissipate heat using water or fans, Google’s researchers designed a system of heat pipes and radiators to keep the chips from overheating. Still, the satellite’s cooling capabilities mean Google will run their chips for 15 minutes at a time before shutting them down to cool off, NPR reported.
However early the experiment might be, the launch is a milestone for an emerging technology that has the potential to help meet AI’s growing demand for compute while avoiding the constraints on resources like power, water, and land on Earth. As companies pour billions of dollars into building out terrestrial data centers, and have faced backlash from community members and politicians in the process, building AI infrastructure in space seems increasingly attractive.
The idea of orbital data centers gained mainstream attention in early 2026 when SpaceX, led by perennial tech optimist Elon Musk, filed plans with the Federal Communications Commission to launch a network of up to 1 million data-center satellites in space.
“The lowest-cost place to put AI will be space, and that will be true within two years, maybe three, three at the latest,” Musk said at the World Economic Forum in January.
Jeff Bezos’s space company Blue Origin in March also filed a proposal to deploy 51,600 satellites in a satellite network dubbed Project Sunrise.
Yet, before the concept went mainstream, Google was already working to make it a reality. In November 2025, it launched Project Suncatcher, to explore “how an interconnected network of solar-powered satellites, equipped with our Tensor Processing Unit (TPU) AI chips, could harness the full power of the Sun.”
With its recent launch, it has taken a significant step toward proving out the orbital data center concept even as experts have warned of major obstacles ahead.
For one, operating data centers in space would require massive solar arrays to capture the sun’s rays and turn it into energy. These large and heavy arrays, along with other equipment, would be expensive to transport to space, Jeff Thornburg, the CEO of Portal Space Systems, previously told Fortune. The expensive price tag remains an obstacle, despite SpaceX, which transports the majority of payloads into space, having cut its estimated launch costs by 95% since 2008 to about $1,000 per kilogram, according to research from Cathie Wood’s ARK Invest.
Another factor is the effect of radiation on the chips powering the data centers which can cause errors. This is something Google is studying with its Project Suncatcher satellite. Although Google said it tested its chips on Earth and found they “can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission,” the company acknowledged that solar events and cosmic rays pose a risk.
There’s also the question of repairing equipment, which is much more difficult in space.
Still, Google isn’t the only company already testing AI chips in orbit. Last November, startup Starcloud launched a satellite carrying one of Nvidia’s high-powered H100 chips into space and showed it could use the chip to run Google’s lightweight AI model, Gemma.
Another startup, Cowboy Space Corp., founded by Robinhood cofounder Baiju Bhatt, also launched a satellite aboard the same SpaceX rocket as Google last week.
The satellite, called Reason-1, plans to test sending solar energy down to Earth using a powerful laser. The company has established contact with the satellite and is preparing to demonstrate its technology for the first time in the coming weeks, Bhatt said in an interview with Bloomberg published Thursday. Bhatt also said Cowboy Space’s ultimate ambition is to build a vehicle to put data center payloads into space.
The Oct. 1 launch is just the beginning of Google’s ambitions with orbital data centers.
The company is planning to send two more satellites into space in 2027 to test how they can communicate via lasers, which it said in a blog post requires precision “similar to hitting a coin-size target from miles away while both points are in motion.”
Still, the ultimate dream of putting data centers into space may not come to fruition even in the next few years, James Manyika, Google’s senior vice president for research, told the New York Times.
“Remember how Google was researching for like 15 years, before anything showed up? I think this is going to look like that,” he said.
The concept of placing AI data centers in space has taken a significant step forward with the recent launch of Google’s Project Suncatcher, a satellite designed to test the viability of AI technology in an outer space environment. On October 1, 2026, a SpaceX rocket successfully launched a refrigerator-sized satellite equipped with four custom Tensor Processing Unit (TPU) chips into low Earth orbit. This mission, developed in collaboration with Planet Labs, aims to assess how these chips perform under the extreme conditions of space, including intense vibrations during launch, radiation exposure, and thermal variations.
The satellite is powered by solar panels that collect energy from the sun. Due to the absence of conventional cooling methods such as water or fans in the space environment, Google has implemented a system of heat pipes and radiators to manage the heat generated by the chips. However, the satellite can only run its chips for 15 minutes at a time before needing to shut down for cooling, indicating the challenges of thermal management in space.
This launch represents a milestone for the concept of orbital data centers, which could help address the increasing demand for computational power in AI while alleviating some of the resource constraints faced by terrestrial data centers. As companies invest heavily in building data centers on Earth, they often encounter community resistance and political challenges. The idea of moving this infrastructure to space offers a promising alternative.
The notion of orbital data centers gained traction in early 2026 when SpaceX, led by Elon Musk, proposed a network of up to 1 million satellites aimed at supporting data center operations in orbit. Musk predicted that within two to three years, space would become the most cost-effective location for AI infrastructure. Following this, Jeff Bezos’s Blue Origin also filed a proposal for a satellite network named Project Sunrise.
Before these mainstream proposals, Google had already initiated Project Suncatcher in November 2025. This project investigates how a network of solar-powered satellites with TPU chips could harness solar energy for computational tasks. The recent launch marks a significant step toward realizing this futuristic vision, despite the numerous challenges that lie ahead.
Operating data centers in space would necessitate large solar arrays to capture sunlight and convert it into usable energy. These systems can be costly to transport due to their size and weight. Although SpaceX has reduced its launch costs significantly in recent years, the expenses associated with deploying the required infrastructure remain a significant barrier.
Another critical challenge is the impact of space radiation on the chips. While Google has performed tests indicating that their chips can withstand radiation levels equivalent to a five-year space mission, the company acknowledges that solar flares and cosmic rays still pose a risk. Furthermore, repairing and maintaining equipment in space presents its own set of difficulties.
Google is not alone in exploring AI chips in orbit. Last November, the startup Starcloud successfully launched a satellite that carried one of Nvidia’s high-powered H100 chips, demonstrating that it could run Google’s lightweight AI model. Additionally, Cowboy Space Corp., founded by Robinhood co-founder Baiju Bhatt, launched its satellite, Reason-1, on the same SpaceX mission. This satellite aims to test a technology for transmitting solar energy back to Earth using a powerful laser.
Google’s October launch is just the beginning of its plans for orbital data centers. The company intends to send two additional satellites into space in 2027 to explore laser communication between them, a task that requires precise targeting over considerable distances while both satellites are in motion.
Despite these advancements, the realization of a fully operational space-based data center may still be years away. James Manyika, Google’s senior vice president for research, cautioned that the development process could resemble Google’s own lengthy research timeline, which took 15 years before yielding significant outcomes in other areas.
In summary, Google’s Project Suncatcher represents a pivotal moment for the potential of orbital data centers, showcasing both the promise and the challenges of implementing AI infrastructure in space. As companies like Google, SpaceX, and Blue Origin push the boundaries of technology, the dream of harnessing the vast resources of outer space for computational power could reshape the future of AI and data processing.
