The mass budget is the governing constraint in orbital compute: every kilogram of hardware lifted to space carries a launch cost that terrestrial infrastructure never pays, and that cost multiplies across every satellite in a constellation. SpaceX, Elon Musk's rocket company, is pressing against that constraint at a scale the industry has not attempted, with plans to deploy millions of satellites into orbit to deliver AI compute from space. The project has drawn the reaction "it seems crazy," and the challenges are described as formidable on both the physics and the financial side.
The physics compound the financial case rather than separating from it. Orbital infrastructure cannot be built to the densities that make ground-based data centres economical. Each satellite must generate its own power within the limits of what its solar panels can supply. Heat exits only through radiation into vacuum. A failed node in orbit cannot be replaced the way a failed server rack on the ground can be. Scaling to millions of units does not smooth those constraints; it multiplies each of them.
The financial pressure runs in parallel. Millions of satellites require a launch cadence with no clear precedent in existing constellation programmes, and the capital required accumulates before a single workload is delivered. Whether SpaceX can compress the cost structure of orbital deployment far enough to compete with ground-based AI infrastructure is the question the project is now forcing into view, and the "it seems crazy" reaction reflects how wide that gap currently appears to be.