Orbital data centers are drawing serious industry attention, but a panel at the AI Infra Summit in Santa Clara (held Sept. 15-17) made clear that the concept remains far from commercially viable at scale. Executives from satellite and space-infrastructure companies, alongside a US Air Force IT leader, outlined the practical barriers still standing in the way.
Thermal, Bandwidth, and Launch Costs
Sumeet Singh, chief data and AI officer at Viasat, identified three core challenges. Thermal management tops the list: in the vacuum of space, heat can only be dissipated through radiators, and bigger radiators mean more weight, a costly tradeoff for orbital systems. Limited uplink and downlink bandwidth for moving data to and from orbit is another constraint. Launch costs are the third major hurdle. Singh said current costs of roughly $2,500 to $3,000 per kilogram would need to fall to about $200 to $500 per kilogram before orbital data centers make financial sense.
Joe Yaffe, COO and chief legal officer of Cowboy Space Corp., agreed these obstacles are real and added that dependence on a single dominant launch provider is a bottleneck of its own. Cowboy Space is addressing this by developing its own launch vehicle and using the rocket’s second stage as the data center itself, while planning to rely on optical links to handle growing data transmission needs.
Power and Customer Readiness
Camille Bergin, chief marketing officer of Star Catcher, said her company is developing laser-based power beaming to supplement solar arrays, potentially boosting power delivery to orbiting systems by up to 10 times. She stressed an incremental approach, warning against designing space systems without a proven business case.
On the customer side, Deepak Sachdeva, CIO of the US Air Force, said the technology isn’t yet ready for operational military use, citing gaps in latency, resilience, and security compared to consumer-level demonstrations. He recommended pursuing a proof of concept before wider adoption.
Singh pointed to StarCloud’s November 2025 launch of a satellite carrying an Nvidia H100 GPU as an encouraging early step, but noted the industry needs intermediate milestones in the 100 to 500 kW compute range before reaching megawatt or gigawatt scale. Bergin also suggested orbital data centers could eventually help address community opposition to terrestrial facilities, though she emphasized that a real business need must justify any move to space.
Despite the unresolved issues, panelists expressed confidence that orbital compute will eventually happen, framing the debate as a matter of timing rather than feasibility.