An orbital-cloud roadmap exposes its hardware assumptions
Market
Cloud-to-orbit architecture and economics / hyperscalers, constellation operators, sovereign compute buyers and spacecraft-platform teams.
Trend
This architecture proposal stages space AI from onboard inference to edge clusters and finally formation-flying orbital cloud data centers, with communications-centric free flyers for access and compute-centric clusters in dawn-dusk sun-synchronous orbit. Its illustrative sizing reaches millions of concurrent users, but that is modeled demand rather than a benchmark or deployment.
Tech Highlight
The decisive weakness is the hardware baseline: each compute satellite is treated like an NVIDIA GB300 NVL72 rack with continuous sunlight, without a sensitivity analysis for radiation derating, thermal rejection, power conversion, launch mass, FPGA or radiation-tolerant alternatives, or link outages. Transfer requires space-qualified compute benchmarks, hardware/software co-design, graceful partitioning across ground, edge and orbit, and cost per accepted workload.
6-Month Outlook
Watch for a thermal-power-mass trade study and a prototype running the same inference harness on commercial and radiation-tolerant hardware. A measured workload-per-watt and workload-per-kilogram curve would turn the roadmap into an actionable architecture.