- Google plans to deploy AI data centers in orbit using solar-powered satellite constellations
- Each Suncatcher satellite would operate in a sun-synchronous low Earth orbit for continuous solar exposure
- Bench tests achieved 1.6 terabits per second between transceivers in controlled conditions
Google’s “Project Suncatcher” introduces an ambitious idea: to place fully functional AI data centers in orbit.
These orbital platforms would consist of a constellation of compact satellites operating in a dawn–dusk sun-synchronous low Earth orbit, designed to capture near-continuous sunlight.
Each unit would house machine learning hardware, including TPUs, powered by solar energy collected more efficiently than on Earth.
A radical concept for orbital computing
The configuration aims to reduce dependency on heavy energy storage and to test whether computation beyond Earth’s atmosphere can be both scalable and sustainable.
The research team proposes inter-satellite communication at bandwidths comparable to ground-based data centers.
Using multi-channel dense wavelength-division multiplexing and spatial multiplexing, the satellites could theoretically achieve tens of terabits per second.
To close the signal power gap, the satellites would fly within just hundreds of meters of each other, enabling data transfer rates that a bench-scale test has already demonstrated at 1.6 Tbps.
Maintaining such close formations, however, requires complex orbital control, modeled using Hill-Clohessy-Wiltshire equations and refined numerical simulations to counter gravitational and atmospheric effects.
According to Google, its Trillium Cloud TPU v6e under 67 MeV proton exposure revealed no critical damage even at doses far exceeding expected orbital levels.
The most sensitive components, high-bandwidth memory subsystems, showed only minor irregularities.
This finding suggest existing TPU architectures could, with limited modification, endure low-Earth orbit conditions for extended missions.
However, economic viability remains uncertain, though projected reductions in launch costs may make deployment plausible.
If prices fall below $200 per kilogram by the mid-2030s, the expense of launching and maintaining space-based data centers could approach parity with terrestrial facilities when measured per kilowatt-year.
Yet this assumes long-term reliability and minimal servicing requirements, both of which remain untested at scale.
Despite the promising signals, many aspects of Project Suncatcher rest on theoretical modeling rather than field validation.
The upcoming partnership with Planet, set to deploy two prototype satellites by 2027, will test optical interlinks and TPU performance under real orbital conditions.
Whether these orbiting facilities can transition from research experiment to operational infrastructure depends on sustained advances in energy management, communication stability, and cost efficiency.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
https://cdn.mos.cms.futurecdn.net/9Km7ky9AU7NcrTGcdP5ZTe-1920-80.jpg
Source link
