In brief
What if our data ended up orbiting the Earth? Between SpaceX’s ambitions, Google’s projects and the European ASCEND study, the race for space data centers is on. Unlimited solar energy, zero cooling water, freedom from terrestrial constraints… the promise is seductive.
But the environmental balance sheet is far more nuanced. So, sustainable revolution or technological mirage? We take stock.
Lifecycle comparison between a terrestrial and an orbital datacenter
The idea is fascinating. Move some of our computing power into space to escape terrestrial limitations: carbon-intensive electricity mix, water consumption for cooling, scarce land, taking advantage of an “always-on” sun. But between the promise and the physical, environmental and industrial reality, the gap remains considerable.
For companies committed to digital responsibility understanding this debate beyond the media buzz is essential. That’s the aim of this article, which compares the life cycle of terrestrial and orbital data centers.
Why dreams are made of orbit
Computing demand soars with AI. The International Energy Agency (IEA) estimates that electricity consumed by datacenters could exceed 1,000 TWh as early as 2026, equivalent to Japan’s annual consumption, under the combined effect of AI and cryptos ( IEA, 2026″; DatacenterDynamics, Jan. 26. 2024).
On Earth, the cooling of large campuses can swallow up several million gallons of water a day during heat peaks, which is already fuelling local tensions and usage conflicts (TIME, 2026).
The promise of space
It is against this backdrop that SpaceX/xAI argue that low-Earth orbit would enable computing clusters to be powered by near-continuous solar energy, while doing away with the need for cooling water. Elon Musk is even talking about a constellation of up to a million “satellite-datacenters” in the medium term.
Europe, for its part, has conducted ASCEND (Advanced Space Cloud for European Net zero emission and Data sovereignty), a technical, economic and environmental feasibility study coordinated by Thales Alenia Space.
In conclusion, the concept could become viable, provided we innovate on launchers and in-orbit assembly.
How can you make an honest comparison?
To avoid “cosmic greening”, the comparison must cover the entire life cycle:
1. Server and infrastructure manufacturing
2. Transport and launches
3. Operations (energy, cooling, network, maintenance)
4. End-of-life (e-waste vs. de-orbiting and debris)
This is the framework adopted by ASCEND, complemented by scientific work on orbital thermics and launcher emissions.
From manufacturing to operation
Manufacturing and construction
A terrestrial datacenter requires concrete, metals, cooling units and kilometers of fiber; it’s heavy on materials, but well controlled industrially, with mitigation levers (low-carbon localization, heat recovery, eco-design).
An orbital datacenter, on the other hand, requires additional weight, which is very penalizing :
- Large-area solar panels to capture energy
- Radiators capable of rejecting tens or hundreds of kilowatts without convection
- Shielding against cosmic radiation
- Inter-satellite laser links for communications
- In orbit, every kilogram costs energy, carbon and euros: it’s the enemy of the balance sheet.
Launches and assembly: the true carbon cost
Rockets inject CO₂, water vapor and soot (black carbon) into high layers of the atmosphere where these particles have a disproportionate radiative effect.
Reference studies show that soot emissions from launchers, if they increase with mega-constellations, will warm the stratosphere and partially degrade the ozone layer.
In terms of “pure” CO₂, a Falcon 9 typically burns 100 to 120 tonnes of kerosene (RP-1) at lift-off, corresponding to a few hundred tonnes of CO₂ per launch. French company Latitude plans to reach a rate of 50 launches a year by 2030 to power its constellations of space datacenters.
The ASCEND study adds that for orbit to become advantageous overall, launchers would have to be ×10 less emissive over their life cycle, and robotized assembly in orbit would reduce the total number of flights.
Operations: energy, cooling, network, maintenance
On Earth, energy use depends on the electricity mix and long-term contracts.
Water remains a key issue: some campuses exceed one million liters per day under normal conditions, and rise to much higher levels during heat waves.
In orbit, solar energy is almost continuous in sun-synchronous orbit, which can lighten the carbon footprint of the “use” phase. But cooling uses only radiation (no air, no water), so large-surface radiators are needed, oriented towards the “cold of the cosmos”.
The ISS, for example, rejects heat via imposing ammonia radiator wings sized for a few dozen kilowatts. Transposing this principle to a high-density GPU farm represents a considerable leap in scale, and one that has yet to be demonstrated industrially.
There are also major constraints:
- Network latency: complex inter-satellite laser links
- Maintenance: on Earth, a card or server is replaced “on the fly”; in orbit, every GPU failure that cannot be repaired degrades overall performance, until a modular exchange or robotic operation is required, technologies that are still in their infancy.
Space debris and end of life: Kessler syndrome
On Earth, the end-of-life of IT equipment is becoming a CSR issue in its own right: e-waste, recovery of critical metals, circular economy. The channels exist and are making progress, even if challenges remain.
In orbit, poorly managed end-of-life feeds the cloud of debris that surrounds the planet. At the beginning of 2026, surveillance networks were tracking some 40,230 catalogued objects, while models estimate :
- 1.2 million pieces of debris between 1 and 10 cm
- 140 million fragments between 1 mm and 1 cm
These invisible but devastating projectiles travel at 7 to 15 km/s, a speed at which a simple bolt becomes as destructive as a grenade.
Even more worrying is the fact that the density of objects in certain altitude “corridors” (LEO 500-800 km, sun-synchronous orbits) is so high that the risk of collisions is increasing year after year. ESA warns of the need for active removals to avoid a Kessler-type runaway effect, where each impact creates a metallic dust that causes others to cascade.
A rapidly worsening situation
A study published in early 2026 reveals some alarming figures: in the event of operators losing control (during a major solar storm, for example), a collision was likely to occur every 121 days in 2018. This has risen to every 2.8 days in 2025, a 43-fold increase in risk in seven years.
For the Starlink constellation alone (over 9,000 satellites), 300,000 avoidance maneuvers were carried out in 2025, representing a 50% increase over 2024.
Each satellite must perform an average of 41 maneuvers a year to avoid collisions.
| Indicateur | Valeur |
|---|---|
| Électricité datacenters (projection 2026) | > 1 000 TWh/an |
| Eau consommée par certains sites | Plusieurs millions de litres/jour |
| Prix lancement Falcon 9 dédié (2026) | ~74 M$ |
| Coût rideshare | ~7 000 $/kg |
| Objets suivis en orbite basse | 40 230 |
| Débris 1-10 cm (estimation) | 1,2 million |
| Fragments < 1 cm (estimation) | 140 millions |
Review and outlook: who really wins?
But the upstream/downstream score is, for the moment, prohibitive:
- More and more launches and associated broadcasts
- Thermal untested at hyperscale GPU
- Delicate maintenance in a space environment
- Risk of debris and cascading losses
On a constant technology basis, the “life cycle” balance remains rather favorable to land-based energy over the next five years, provided that we accelerate the decarbonization of electricity, sobriety and water reduction.
The tipping point could come if three locks are broken simultaneously:
- launchers that are at least ten times less emissive over their lifecycle, and highly reusable
- high-flow-density, highly reliable heat sinks for megawatt GPU clusters
- modular robotic maintenance and a strict enough orbit law to guarantee a debris-free end-of-life. This is precisely the tripod pointed out by ASCEND in its feasibility conclusions.
And what about 2035-2050?
At that point, the market will need common benchmarks: we might as well expect to see the emergence of a “SpaceDataCenter ” label defining the obligations of :
- Measuring environmental impact
- Radiative cooling
- Responsible maintenance
- Debris-free end-of-life
SQORUS supports you in your responsible digital approach
SQORUS supports you in your responsible digital approach.
Orbiting data centers will remain a technological mirage for decades to come. But this debate has the merit of asking the right questions: how can we reduce the environmental footprint of our digital infrastructures while meeting growing computing needs?
The answer lies on Earth, in the choices we make today.
At SQORUS, we believe that every effort counts.
With over 35 years’ experience in the digital transformation of HR, Finance and IT functions, we support organizations in their digital responsibility initiatives.
Recognized Label Numérique Responsable level 1 awarded by the Institut du Numérique Responsable, and the EcoVadis 2025 Gold Medal with a score of 79/100 (96th percentile), these awards testify to our commitment to a more environmentally-friendly digital environment.
By taking action at your own level, you are helping to meet the collective challenge of a more sustainable digital future.
A project? A request?A question?
With SQORUS, you can commit to an approach that combines performance and responsibility.
Join a great place to work
FAQ
Do data centers in orbit already exist?
Yes, in prototype form. At the end of 2025, start-up Starcloud launched its first satellite with Nvidia GPUs into orbit. Google and SpaceX are also testing their first concepts. But a viable large-scale infrastructure is still a long way from becoming an industrial reality.
Are orbiting datacenters greener than terrestrial ones?
Not yet. Having solar power in orbit and no need for water cooling are considerable advantages, but the emissions produced during rocket launches and the dangers associated with space debris tarnish this picture. At present, a well-managed terrestrial datacenter is even more virtuous, and will probably remain so until at least 2030.
Is my data safe in a space datacenter?
Not without risks. Satellites face risks such as cosmic radiation, solar storms and space debris, which have no equivalent on Earth. It's important to remember that, in the event of failure, there's no possibility of immediate human action. . Added to this is a legal vacuum: what law applies to data hosted in orbit? The question remains unanswered.
