Descriptions:
Bloomberg Technology brings in Makenzie Lystrup — former director of NASA’s Goddard Space Flight Center and now principal consultant at Paradox Services — to rigorously assess the engineering realities of putting AI data centers in orbit, motivated by SpaceX’s ambitious 2028 orbital compute plans.
Lystrup lays out three distinct tiers of orbital AI compute. The first, on-orbit edge computing, is near-term and real: satellites collect raw data and perform AI inference locally, sending only high-value processed outputs to Earth — a model already prototyped by companies like Acxiom on the International Space Station. The second tier, resilient sovereign storage, has genuine defense and continuity-of-government applications but remains underexplored. The third tier — hyperscale AI inference at orbital scale, the moonshot SpaceX is pitching — faces formidable unsolved challenges in heat rejection (space has no convective air cooling, forcing all waste heat to be radiated away by physical panels), radiation-tolerant computing, high-bandwidth optical networking, and in-orbit maintenance costs that are rarely discussed.
On power, Lystrup challenges the simple “infinite solar” narrative: Google’s Project SunCapture estimates orbital solar panels can be up to eight times more productive than ground-based equivalents, but energy access does not equal energy infrastructure — flight-qualified power storage, distribution, and thermal management must all be built and operated reliably. Asked directly whether SpaceX can pull it off, Lystrup notes the company’s track record of solving hard problems and declines to count them out.
📺 Source: Bloomberg Technology · Published June 17, 2026
🏷️ Format: Interview
