The case for putting data centers in orbit has always been a cooling and resource argument before it was a compute argument. In a June 2026 analysis, the World Economic Forum walked through why cooling is the real obstacle to space-based data centers, and the framing matters as much as the conclusion. When an institution like the WEF treats orbital compute as a serious response to terrestrial water and power constraints, it is signaling how severe those constraints have become for operators on the ground.
The pitch is clean on paper. In orbit there is continuous solar power above the atmosphere, no neighbors fighting a rezoning, no aquifer to draw down, and a heat sink that costs nothing to access: the roughly 2.7 kelvin background of deep space. A data center in orbit consumes no municipal water for cooling because it does not evaporate anything. It radiates heat away as infrared energy. For a terrestrial industry now rationing water and fighting community opposition over both power and land, that pitch lands.
The physics counterweight is unforgiving. Radiative cooling rejects heat far more slowly per unit of surface than the convective and evaporative loops that terrestrial facilities rely on. A two-sided radiator held near 20 degrees Celsius emits only about 633 watts per square meter, governed by the Stefan-Boltzmann relationship between temperature and radiated power. By the WEF's own math, a one-megawatt orbital data center would need on the order of 1,600 square meters of radiator surface to stay cool, roughly the footprint of four tennis courts for a single megawatt of compute. Running the radiators hotter, near 60 degrees Celsius, can cut the area by about half, but it pushes the silicon toward its thermal ceiling.
Surface area in orbit is mass, and mass is launch cost. Radiator panels run on the order of five to nine kilograms per square meter, and under pessimistic assumptions radiators can account for as much as 60 percent of a system's total mass. Every kilogram has to be lifted out of the gravity well, which is why the entire economic case hinges on launch prices falling toward a few hundred dollars per kilogram rather than the thousands that defined the prior era. Low Earth orbit adds a second tax: a satellite there cycles between sunlight and shadow roughly every 90 minutes, subjecting radiators and structures to relentless thermal expansion and contraction that stresses reliability over time.
The concept is moving past slideware, which is part of why the WEF is engaging with it now. Starcloud put an NVIDIA H100 in orbit aboard Starcloud-1 in late 2025 and has signaled far larger clusters to follow. Axiom Space began deploying orbital data center nodes in January 2026 and is testing thermal tiles with a hardware partner, a program we covered in Axiom and Spacebilt's vacuum-radiator work on the ISS. SpaceX has shown its own orbital compute satellite design built around a large solar array and a deployable liquid radiator. The broader investment thesis, including the gigawatt-scale ambitions described in Jeff Bezos's pitch for gigawatt space data centers, treats radiative thermal management as the load-bearing engineering problem rather than an afterthought.
Maintenance and latency remain genuinely unsolved. A failed pump or a micrometeorite strike on a radiator panel cannot be serviced by a technician walking a hot aisle, and round-trip light delay rules orbital sites out for latency-sensitive inference even before the cooling question is settled. These are not reasons to dismiss the idea so much as reasons the WEF frames it as serious-but-early: a plausible destination for specific workloads, contingent on launch economics and on radiator engineering that does not yet exist at scale.
The signal for the terrestrial cooling industry runs in two directions. First, radiative and two-phase thermal management for space is becoming a real engineering discipline with venture funding behind it, and the work on high-emissivity surfaces, deployable radiators, and dielectric loops in vacuum will feed back into how heat rejection is studied on the ground. Several of the same ambitions show up in national programs, including the buildout described in China's five-year plan for space-based data centers.
Second, the orbital concept is a barometer. Operators do not seriously price out launching racks to space because it is convenient. They do it because terrestrial water and power constraints have tightened to the point where a 1,600-square-meter radiator in vacuum starts to pencil against a water permit that will never be granted. The honest reading of the WEF piece is that physics still sets the terms in orbit, and the more interesting story is what the rising seriousness of that conversation says about the limits closing in on cooling here at home.