Orbital data centers are no longer theoretical. Google and SpaceX are in active talks to launch data centers in space, according to reporting by The Wall Street Journal citing sources familiar with the matter. The discussions center on a rocket launch deal to deploy compute infrastructure in orbit, a move driven by the relentless energy demands of AI training and inference. Google CEO Sundar Pichai has already sketched the timeline: “A decade or so away, we’ll be viewing it as a more normal way to build data centers.”
Key Takeaways
- Google and SpaceX are negotiating to deploy orbital data centers to address AI compute bottlenecks.
- SpaceX filed with the FCC for up to 1 million satellites at 500–2,000 km altitudes, projecting 100 gigawatts of AI compute capacity.
- Google’s Project Suncatcher aims to launch prototype satellites by 2027 equipped with Tensor Processing Units for testing.
- Starcloud trained the first large language model in orbit in late 2025, proving early feasibility.
- Terrestrial data centers remain cheaper today when factoring satellite construction and launch costs.
Why Orbital Data Centers Matter Right Now
The timing is critical. AI’s compute demands are doubling faster than terrestrial infrastructure can expand. Ground-based data centers face siting challenges—local opposition, power grid constraints, water availability—that slow deployment. Orbital infrastructure sidesteps these bottlenecks entirely. No neighbors to object. No local utility politics. Just physics and engineering.
SpaceX is positioning orbital compute as the next frontier, particularly as it prepares for a $1.75 trillion IPO later in 2026. The company has already acquired xAI and is leveraging that foothold to pitch orbital data centers as the cheapest place for AI compute within a few years. Meanwhile, Anthropic recently partnered with SpaceX to access xAI’s Memphis, Tennessee data center resources, with hints of future orbital collaboration.
Google’s Dual-Track Approach to Space Compute
Google isn’t betting solely on SpaceX. The company is talking to other rocket-launch providers while simultaneously advancing Project Suncatcher, its own orbital compute initiative. Project Suncatcher calls for launching prototype satellites by 2027—two spacecraft equipped with Tensor Processing Units designed to test thermal management and reliability in the harsh space environment.
This dual strategy mirrors Google’s broader infrastructure philosophy: diversify suppliers, reduce single-vendor dependency, and maintain optionality. Investing $900 million in SpaceX back in 2015 gave Google early exposure to the company’s ambitions. Now, as launch costs plummet and orbital economics improve, that early bet is paying strategic dividends.
The Technical and Economic Barriers Remain Real
Don’t mistake momentum for inevitability. Orbital data centers face substantial hurdles. Radiation-hardened hardware costs more than terrestrial equivalents. Laser-based inter-satellite and Earth communications systems are still maturing. Launch economics, while improving, remain expensive at scale. Orbital debris is a growing hazard that could threaten constellations.
Current analyses show terrestrial data centers are still cheaper when you factor in satellite construction, launch costs, and operational complexity. The orbital advantage exists primarily on paper—in projections of what happens when launch costs fall further and satellite manufacturing reaches true scale. That future may arrive within a decade. It hasn’t arrived yet.
Blue Origin is also pursuing orbital data centers, ensuring this isn’t a Google-SpaceX duopoly. Starcloud, the early mover, already trained a large language model in orbit in late 2025, proving the concept works at small scale. The competition is real, and it’s accelerating.
What Changes When Orbital Data Centers Become Normal
If Pichai is right and orbital compute becomes routine within a decade, the implications ripple across the entire tech stack. AI training workloads could shift to space-based infrastructure, reducing pressure on terrestrial power grids. New companies could emerge to specialize in orbital operations, maintenance, and data transmission. Satellite manufacturing would become as industrialized as semiconductor fabs.
The geopolitical angle matters too. Space-based infrastructure exists in a regulatory gray zone. International treaties govern satellite operations, but orbital data centers don’t fit neatly into existing frameworks. Expect regulatory clarification—and tension—as these projects move from talks to hardware.
Is orbital compute cheaper than ground-based data centers today?
No. Terrestrial data centers remain more cost-effective when accounting for satellite construction, launch, and operational complexity. The orbital advantage exists in future projections, assuming launch costs continue falling and manufacturing scales significantly.
When will Google’s orbital data center prototypes launch?
Google’s Project Suncatcher targets prototype satellite launches by 2027. These two test satellites will carry Tensor Processing Units to validate thermal management and reliability in orbit before larger-scale deployment.
Are other companies building orbital data centers?
Yes. Blue Origin is pursuing orbital compute infrastructure, and Starcloud already trained a large language model in orbit in late 2025, demonstrating early feasibility at smaller scale.
The shift to orbital data centers represents a fundamental rethinking of infrastructure placement. For decades, data centers stayed on the ground because physics and economics demanded it. Now, as AI’s hunger for compute outpaces terrestrial expansion, space becomes the logical next frontier. Google and SpaceX’s talks are not fringe speculation—they’re the leading edge of a transition that will reshape how the world’s most powerful computing happens. Within a decade, launching a data center into orbit may feel as routine as opening a new facility in a different city.
Edited by the All Things Geek team.
Source: TechRadar
