A DC Atlas Frontier

Data Centres in Orbit: The Frontier That Cooks Itself

Solar that never sets, cooling from the cold of space, no land, no water, no grid queue. The orbital-data-centre pitch is seductive. So we did the physics and the dollars. It doesn't close.

Built on published aerospace engineering figures and DC Atlas's facility data.

By DC AtlasPublished Last updated 7 modules · 77 min

The constellation, and what it's really made of

A ring of data-centre satellites in low orbit, laser-meshed into one network. Drag to rotate — then toggle the reality to see the radiator field each one has to launch.

The constellation, and what it's really made of

A ring of data-centre satellites in low orbit, laser-meshed into one network. Drag to rotate — then toggle the reality to see the radiator field each one has to launch.

OrbitAltitudePeriodRound-trip latency
Low Earth orbit~550 km~95 min~5 ms
Geostationary35,786 km24 hours~240 ms
Radiator, 40 MW
60–80k m²
Mass to launch
1,500–3,000 t
Upfront — orbit
$2.3–4.5B
Upfront — ground
$1.2–1.6B
Upkeep/yr — orbit
$0.6–3.2B
Upkeep/yr — ground
$0.4–0.6B

Illustrative. The orbital regimes (ISS, LEO, MEO, GEO) are labelled with real altitudes but spaced for readability, not to true scale; the debris layer is a representative sample of the 25,000+ tracked objects from Module 5; the reality figures carry forward the 40 MW estimates from Modules 2–4.

Source: DC Atlas — orbit geometry and the Modules 2–4 mass/cost synthesis; Google Research (Project Suncatcher) for the constellation architecture

Illustrative. The orbital regimes (ISS, LEO, MEO, GEO) are labelled with real altitudes but spaced for readability, not to true scale; the debris layer is a representative sample of the 25,000+ tracked objects from Module 5; the reality figures carry forward the 40 MW estimates from Modules 2–4.

Source: DC Atlas — orbit geometry and the Modules 2–4 mass/cost synthesis; Google Research (Project Suncatcher) for the constellation architecture

The Frontier

  1. 01The Pitch, Taken SeriouslyStarcloud, Google's Project Suncatcher, the ESA-backed ASCEND study and a 20,000-satellite filing. Who is proposing orbital data centres, what they claim, and the strongest version of the case, stated fairly before the teardown.11 min
  2. 02The Heat ProblemIn vacuum there is no convection: every watt of waste heat must be radiated. The Stefan–Boltzmann floor, the real ISS radiator numbers, and what a 40 MW facility's radiator field actually masses. This is the argument.12 min
  3. 03The Launch ProblemThat radiator, plus power, plus structure, plus compute. All of it multiplied by dollars per kilogram to orbit. Falcon Heavy's demonstrated cost against Starship's aspirational target, kept honestly apart.10 min
  4. 04The Power ProblemSolar arrays sized per kilowatt, then the eclipse tax: 47% of every orbit in Earth's shadow forces battery mass and array oversizing. What 40 MW continuous actually needs in the sky.11 min
  5. 05The Operations ProblemGPUs die in one to three years and in orbit no one can swap them. Radiation-driven failure, orbital debris, and why on-orbit servicing at tens of millions per mission can't stand in for a loading-dock full of spares.11 min
  6. 06The Connectivity QuestionHow you would even connect it: LEO's tens-of-milliseconds against GEO's 240, a constellation laser-meshed into a ring, whether Starlink's optical backbone is the ready-made answer, and the ground-station bottleneck the pitch skips.10 min
  7. 07The VerdictThe whole system priced against a terrestrial build per megawatt, then the strongest case against our own position, and the exact conditions under which orbit wins. Plus the assumptions table, published in full.12 min

How we work. Every figure in a Frontier is either cited to a primary source or labelled as a DC Atlas estimate with its assumption published. Charts render from the numbers, never as images. Read our methodology.