Ethos Space

AI compute demand is compounding toward a terawatt per year of new capacity, but Earth's power grid limits how much new data center capacity can be built. Space based data centers are emerging as a response, but rocket launch capacity becomes the next bottleneck, with Starship at full cadence capping out at only a few hundred gigawatts per year of deliverable mass. Reaching terawatt scale AI infrastructure in space requires using mass that is already outside Earth's gravity well, meaning it has to come from the Moon rather than be launched from Earth.

Ethos develops a process to convert lunar regolith, which is primarily silicon, aluminum, and oxygen, into usable materials through melting, electrolysis, and purification. This single process yields construction materials, rocket propellant, and solar grade silicon at different levels of refinement. The company tests this process the way it would actually run on the Moon, using molten regolith in hard vacuum inside their own lunar environment chamber. Their stated sequence starts with first heavy landings, followed by mining and construction work to build landing pads, produce propellant for return trips, and manufacture silicon for solar power, with later stages including a mass driver to launch AI satellites over an estimated ten year buildout.

Demand for AI compute is compounding toward a terawatt per year, and both Earth's grid and rocket launch cadence are described as near term limits, pushing the case for lunar sourced infrastructure now.