Descriptions:
Bloomberg Technology explores the emerging category of orbital data centers, interviewing Star Cloud CEO and co-founder Philip Johnson, whose company launched the first Nvidia H100 GPU aboard a satellite in November 2025. The segment opens with a detailed Bloomberg-produced explainer on the physical architecture of a space-based compute node — covering radiation-tolerant chips, laser inter-satellite links, massive solar arrays, and deep-space radiators as a cooling alternative to terrestrial air and water systems.
Johnson explains that Star Cloud has pivoted from training-focused large satellite designs — including a rendered concept featuring a four-kilometer-by-four-kilometer solar array paired with a 5GW compute cluster — toward a distributed constellation of small inference nodes. The company has filed with the FCC for a constellation of 88,000 satellites, targeting approximately 20GW of deployable compute capacity. Johnson describes the shift as a direct response to the AI industry’s move toward inference workloads, which he estimates now account for 99% of AI compute demand.
The interview also addresses competitive dynamics with SpaceX, which has announced plans for its own orbital compute network. Johnson frames the market as large enough for multiple independent providers, positioning Star Cloud as a vendor for spacecraft operators who want compute infrastructure independent of SpaceX. Star Cloud currently launches on SpaceX Falcon 9 rideshare and dedicated missions.
📺 Source: Bloomberg Technology · Published June 05, 2026
🏷️ Format: Interview
