Research Feed

This is a bottom-up revenue model for orbital compute infrastructure, calibrated against CoreWeave's published contract economics. Two independent approaches converge on a revenue range of $10.7 billion to $16.0 billion per gigawatt of input power per year. The lower bound captures the advantage of higher Power Usage Effectiveness (PUE). The upper bound prices in custom inference silicon performance of operations per delivered watt. Silicon efficiency is the dominant variable, producing a $14.5 billion/GW swing across its feasible range. Revenue per GW is location-independent; orbital deployment wins on the cost side, where Wright's Law learning curves compress launch and subsystem costs with each replacement cycle while revenue holds.

Over $2 billion of venture capital has flowed into in-space mobility in recent years, every dollar into chemical or electric architectures with known physics ceilings. We make the case for a third propulsion class, solar-thermal, and for Portal as the company to bring the technology to production.