The promise and Folly of Orbital Data Centers

This page explores the thermodynamic and logistical challenges of moving AI infrastructure into orbit, specifically critiquing the feasibility of hyperscale space datacenters. The text argues that while centralized training faces extreme heat-rejection hurdles in a vacuum, distributed inference using localized models offers a more realistic path forward. To support this vision, the materials propose a transition from monolithic designs to hybrid architectures and specialized orbital edge nodes. Central to this infrastructure is the concept of an uncrewed “wingman” cargo drone, a modernized, Shuttle-inspired vehicle designed for persistent orbital construction and servicing. By decoupling crew safety from heavy industrial tasks, this system aims to restore the assembly capabilities lost after the Space Shuttle’s retirement. Ultimately, the sources suggest that true space industrialization requires reusable logistics and architectural efficiency rather than simply scaling traditional terrestrial hardware.