Space-based power for AI compute is closer to economic parity than many think. In this week’s analysis, we break down the true $/W of orbital power systems, compare them directly to terrestrial datacenters, and show why mass-optimized HEO architectures could already compete on cost at Starship-era launch prices.

The premium model applies real-world constellation physics to the $37B DTC TAM to show how much revenue SpaceX and ASTS can actually serve. A full supply-side Monte Carlo reveals the potential performance of the two architectures, and projects 2030 DTC coverage gap revenue.

Mach33’s open-source model estimates a $37B consumer Direct-to-Cell market in 2030, driven primarily by large middle-income regions. Modelling on a per-country basis, the analysis shows that dropout prevalence, not income, is the dominant driver of consumer DTC demand.

A breakdown of SpaceX’s 100-gigawatt per year orbital compute goal in ~5 years: the power, thermal, and hardware breakthroughs needed to make it real.