Delian Asparouhov and Philip Johnston on making the case for orbital data centers
Summary
A SpaceNews event explored the future of orbital data centers, featuring insights from Founders Fund's Delian Asparouhov and Starcloud CEO Philip Johnston. Asparouhov, initially skeptical, now sees potential by the 2030s, driven by projected lower launch costs, cheaper space solar, and terrestrial data center moratoriums. He notes Elon Musk's efforts in chip and solar panel production will initially benefit Starlink's direct-to-cellular business, with early orbital data center applications focusing on edge compute for Earth observation, orbital factories, and lunar operations. Johnston detailed Starcloud's progress, including launching its first spacecraft with an Nvidia H100 GPU, achieving in-space model training and high-powered SAR inference. Starcloud plans an 88,000-satellite constellation to deploy 20 gigawatts in Low Earth Orbit, aiming to compete with terrestrial data centers on energy cost for inference workloads by 2035, contingent on Starship's frequent flights by 2029-2030. Key challenges include developing low-mass, low-cost deployable radiators and radiation-hardened chips.
Key takeaway
For CTOs and AI architects planning long-term compute strategies, orbital data centers offer a future solution for specific inference and edge workloads, particularly by the 2030s. You should monitor Starship's operational cadence and advancements in space-hardened hardware, as these will dictate the feasibility and cost-effectiveness of deploying non-latency-critical AI inference in orbit, potentially offering a competitive alternative to terrestrial hyperscalers.
Key insights
Orbital data centers are emerging as a viable, long-term solution for edge compute and inference, driven by space technology advancements.
Principles
- Terrestrial data center moratoriums increase demand for space-based alternatives.
- Frequent, reusable heavy-lift launch is essential for scaling orbital compute.
- Edge inference and specialized processing are early, high-value orbital compute applications.
Method
Starcloud's strategy involves deploying large constellations of 3-ton, 200kW spacecraft, utilizing advanced deployable radiators and radiation-hardened chips, with Starship as the primary launch vehicle.
In practice
- Conduct extensive ground testing for chips in high-radiation environments.
- Develop highly efficient, low-mass thermal dissipation systems for space hardware.
Topics
- Orbital Data Centers
- Space Infrastructure
- AI Inference
- Starship
- Satellite Constellations
- Edge Computing
- NVIDIA H100
Best for: Investor, CTO, AI Architect
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by SpaceNews.