Satellite constellations optimized as AI training data sources represent a distinct third market category in the AI-space intersection that is viable at current launch costs

The AI-space intersection has three distinct market categories with different technical requirements and commercial viability timelines: (1) Orbital edge inference processes satellite sensor data in orbit for operational efficiency (Axiom/Kepler, Planet Labs) - already operational; (2) Orbital AI training attempts to compete with terrestrial data centers by training models in space (Starcloud model) - speculative, requires sub-$500/kg launch costs; (3) Satellite-as-AI-training-data uses space as continuous multi-modal sensing infrastructure feeding ground-based AI training (Xoople model) - viable today at current launch costs. Xoople's $225M funding (including $130M Series B) and L3Harris partnership demonstrate investor confidence in category 3 as commercially mature. The distinction matters because category 3 doesn't face the thermal management, bandwidth, or radiation hardening constraints of orbital computing - it leverages space's unique vantage point for continuous Earth observation (optical, infrared, SAR, SIGINT) while performing compute terrestrially. L3Harris involvement signals defense/intelligence community interest as anchor customer, parallel to the national security demand floor pattern in commercial LEO computing. This represents a viable business model today rather than a speculative future dependent on launch cost breakthroughs.