Talk of megawatt-scale data centers in Earth’s orbit is now migrating from a dream and a vision to boardroom pitch decks and media soundbites. Space executives speak about relocating the digital economy to low Earth orbit, as if it were just a matter of launch vehicle capacity. But there’s a less glamorous challenge lurking behind this new paradigm: It’s not the launch that keeps engineers up at night.

It’s thermal management.

Back on Earth, every data center relies on the ability to effectively reject waste heat dissipated by the enormous and insatiable computing capacity: Be it using huge noisy fans, liquid cooling, or other innovative techniques, each redirects excess thermal energy that would otherwise result in a total and literal meltdown into Earth’s atmosphere, extensive water resources, or other solutions.

In space, this changes significantly.

The cosmic background temperature of deep space—dating to the Big Bang origin of the universe—hovers around 4 K, or –269°C. And in the vacuum of space, there’s no air to use convection to reject heat as we do on Earth. Without air to convect heat or the ground to absorb it, one is left with radiating the heat into the coldness of deep space.

But there’s a catch.

Radiation heat is less efficient than convection. Its challenges also grow as data centers scale. The heat a surface can radiate depends on its radiating area, temperature, and material properties, which are all governed by the Stefan-Boltzmann law of physics. A data center in space trying to get rid of 1 MW of waste heat into deep space needs a radiating surface of about 1,200 square meters (35 × 35 meters) for the system to remain in a narrow band of allowable operating temperature range. In orbit, that translates into a significantly large radiator structure that is within technological reach with today’s state-of-the-art for large, deployable structures. For example, AST Mobile is planning to deploy a 20 × 20-meter phased-array antenna in lower Earth orbit for direct satellite to cellphone communications. A similar structure can be used for the purpose of building a megawatt-sized free-flying data center module in a constellation of modules comprising an orbiting data center.

An effective thermal management approach to building orbiting data centers in lower Earth orbit therefore would have to rely on the following factors:

• Data center free-flying modules are placed in a sun-synchronous orbit, with almost constant sun illumination (barring short solar eclipses) on its solar arrays and constant heat rejection into deep space.

• Between the hot solar arrays and the coldness of deep space, computing elements (which act as heaters) passively conduct heat to radiators that reject that heat into deep space, keeping electronics in the allowable operating temperature range.

• Heat generated internally is used to keep other components warm and thus passively manage the thermal equilibrium within the operating system.

• By using passive cooling, with no pumps, fluids, or active thermal loops, conductive spreading allows heat to be radiated directly into space with no moving parts.

Telemetry readings must also inform the operating system of the thermal performance of the system in real time so as to inform load-balancing workload assignments that take temperature into account.

Ultimately in orbit, however, physics itself is the systems’ architect. That means thermal management must be an integral part of design from the very start.