04 · RESEARCH TOPIC
Orbital Computing
Computers in orbit can process data where it is collected and serve people far from any ground station. They also cannot be repaired, every kilogram has to be launched, and all their heat must be radiated into space. I measure where orbital computing pays off and what it costs.
Satellites have carried computers for decades. What is new is the scale: in 2026, companies filed plans for constellations of data centres in orbit, from tens of thousands of satellites to as many as a million, powered by sunlight around the clock. The image above shows 13,020 of them, a sample I modelled on one of those filings.
The idea has real strengths, and some costs that are easy to leave out of a pitch. With colleagues at Saarland University, Nanjing University and the University of Genoa, I put numbers on both: latency, energy, mass and carbon over the whole life of the hardware, from the factory to re-entry.
Where it can work, and what it costs
Where it can work
- Distance to the ground. A cloud in orbit answers faster than one on the ground when the user is more than about 5,000 km from the nearest gateway. Under 2,000 km, the ground wins. ASMS/SPSC 2025, §V-A
- Shrinking data before sending it. Processing in orbit saves carbon when it cuts the data enough: with one inter-satellite hop, down to 50% of the raw size on Falcon 9, or 70% on Starship. Dirty Bits, §4.3
- Against an average data centre. Within its first one or two years in orbit, a node emits less carbon per GPU-hour than the global-average data centre on Earth (about 477 g of CO₂e). Dark Clouds, LEO-NET 2026, Fig. 1
What it costs
- Heat. Every watt a computer draws ends up as heat, and in vacuum it can only be radiated away. For a 1 kW node, the radiator weighs 30.5 kg of the 44.3 kg modelled; the computer itself, 3.2 kg. Dark Clouds, §4 Scaled to the 250 kW that SpaceX quotes for each of its Star Mind spacecraft, the same ISS-based density gives about 760 m² and 7.6 tonnes of radiator, the heat load of some 18 ISS radiator panels. Dark Clouds, §3
- Launch and re-entry. Each kilogram costs about 52 kg of CO₂e on Falcon 9 and 34 on Starship. Re-entry adds about half again on top of the launch. Dirty Bits, §3.1
- No repair. Spares have to fly from day one. One cold spare of the computer adds 40% to the carbon per GPU-hour on Starship, and 34% on Falcon 9, for a three-year mission. Dark Clouds, abstract
- Against a clean data centre. Matching a data centre on a clean grid (about 107 g per GPU-hour) takes at least six years in orbit with one spare on Starship, or nine years with two spares on Falcon 9. Dark Clouds, §4
Carbon figures count manufacturing, launch and re-entry of the modelled components; once in orbit the node runs on sunlight. The satellite bus (structure, attitude control, avionics) is left out, as in the papers; including it would add mass.
Size a computing node
One node: an H100-class GPU with its controller, storage and laser link, 1 kW in all, in a dawn-dusk orbit at 510 km. Choose how long it flies, how many copies of the computer it carries, and which rocket launches it.
- Carbon per GPU-hour 147 g CO₂e
- Modelled mass44.3 kg
- Radiator share
The numbers behind this: carbon per GPU-hour (g CO₂e)
From Fig. 1 of Dark Clouds Rising in Low-Earth Orbit (LEO-NET 2026), computed with the open-source ESpaS-ODC model. Rows: years in orbit. Columns: copies of the computer (1 = no spare). For comparison, a global-average data centre emits about 477 g per GPU-hour and one on a clean grid about 107 g.
| Years | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| 1 | 441 | 616 | 791 | 966 | 1141 |
| 2 | 221 | 308 | 396 | 484 | 571 |
| 3 | 147 | 206 | 264 | 323 | 381 |
| 4 | 111 | 155 | 198 | 242 | 286 |
| 5 | 89 | 124 | 159 | 194 | 229 |
| 6 | 74 | 103 | 132 | 162 | 191 |
| 7 | 64 | 89 | 114 | 139 | 164 |
| 8 | 56 | 78 | 100 | 121 | 143 |
| 9 | 50 | 69 | 89 | 108 | 127 |
| Years | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| 1 | 529 | 710 | 892 | 1073 | 1255 |
| 2 | 265 | 355 | 446 | 537 | 628 |
| 3 | 177 | 237 | 298 | 358 | 419 |
| 4 | 133 | 178 | 223 | 269 | 314 |
| 5 | 106 | 143 | 179 | 215 | 252 |
| 6 | 89 | 119 | 149 | 180 | 210 |
| 7 | 76 | 102 | 128 | 154 | 180 |
| 8 | 67 | 89 | 112 | 135 | 158 |
| 9 | 59 | 80 | 100 | 120 | 140 |
The radiator is sized by power, at 3.03 m² per kilowatt (calibrated on the International Space Station), so mission length and spares leave it unchanged. A GPU in orbit is expected to last about three years, so a mission of T years needs at least ⌈T/3⌉ copies. A fuller calculator built on ESpaS is planned.
Tools and projects
-
ESpaS
Open-source estimator of the carbon footprint of computing in orbit, from manufacturing through launch to re-entry, built at Saarland University. MIT and Apache-2.0.
-
MeteorNet
A continuous-time emulator for edge computing on low-Earth-orbit constellations, used to test where servers should sit and how tasks should move between orbit and ground.
-
FOUNDATION
Computing without repair: what it takes to keep hardware dependable for years when no one can reach it. A 30-second teaser.
Selected papers
- Dark Clouds Rising in Low-Earth Orbit: On Environmental Limits to Massive Orbital AI ACM Workshop on LEO Networking and Communication (LEO-NET), co-located with ACM SIGCOMM, 2026
- Optimal Design of Space-Based Data Center Constellations for In-Orbit Computing Services IEEE Communications Letters, 2026
- Space Cloud Networks: Technologies, Architectures, and Key Enablers IEEE Communications Magazine, 2026
- Dirty Bits in Low-Earth Orbit: The Carbon Footprint of Launching Computers ACM SIGENERGY Energy Informatics Review, 2025
- On the Latency Trade-off Between Space and Terrestrial Clouds in Non-Terrestrial Networks Advanced Satellite Multimedia Systems Conference and Signal Processing for Space Communications Workshop (ASMS/SPSC), 2025
People and partners
- Robin Ohs, PhD student at Saarland University, co-supervised with Holger Hermanns: the lifecycle sustainability of computing in orbit.
- Gregory F. Stock, PhD student at Saarland University, co-supervised with Holger Hermanns.
- Wenxiao Ge, PhD student at Nanjing University, co-supervised with Kanglian Zhao: the architecture and design of space clouds.
- Camilo Rojas Milla, PhD at the University of Genoa (2026), co-supervised with Fabio Patrone: space and ground clouds compared.
Nanjing University · University of Genoa · TU Munich
Funding
DFG (TRR 248 CPEC)
Writing about orbital data centres?
I can explain the trade-offs, check numbers, or point you to the right people. The press kit has bios, a photo and these results with their sources.