Schmidt’s Vision: Orbital Data Centers and Cloud Infrastructure

In the wake of his surprise acquisition of Relativity Space in early 2025, former Google CEO and chairman Eric Schmidt is charting an ambitious course: deploying large-scale data centers into low-Earth orbit to meet the exploding energy and computing demands of AI. Schmidt’s rationale became evident during his April testimony before the House Committee on Energy and Commerce, where he highlighted projections showing AI-driven data centers could require up to 67 GW of additional power by 2030—equivalent to nearly 70 new average-sized nuclear power plants.
Schmidt’s Congressional Testimony and the Energy Crunch
During the hearing, Schmidt underscored the inadequacy of terrestrial power grids to support next-generation AI workloads. “People are planning 10 GW data centers,” he said, noting that today’s global data-center load hovers around 3–5% of total electricity generation but could surge to as much as 99% without new solutions. He warned that U.S. power consumption has only grown ~0.5% per year over the last decade—far below the projected ramp for AI.
- Average U.S. nuclear plant: ~1 GW output
- Projected 2027 demand: +29 GW for AI centers
- Projected 2030 demand: +67 GW, driving major grid upgrades
Beyond electricity, water usage for cooling data centers is another constraint that orbital facilities could sidestep.
Why Relativity Space?
Relativity Space stands out as one of the few U.S. launch providers developing a truly heavy-lift, partially reusable rocket: Terran R. Designed to loft 33.5 metric tons to LEO in expendable mode (23.5 t with first-stage recovery), Terran R eclipses both ULA’s Vulcan Centaur and Rocket Lab’s upcoming Neutron in payload capacity. With a projected maiden flight in late 2025 and a recent $650 million Series E funding round, Relativity is rapidly maturing its 3D-printed engines and composite structures.
“Control over launch cadence and manifest is crucial if you want to build orbital infrastructure,” says aerospace analyst Dr. Mira Thompson of the Center for Space Policy. “Relativity’s factory in Long Beach can ramp to dozens of Terran R rockets per year once the next printer farms come online.”
Technical Challenges of Orbital Data Centers
Placing data centers in space entails solving several engineering hurdles:
- Power Generation & Storage: Massive deployable photovoltaic arrays tailored to sun-synchronous orbits can generate 1–3 kW/m2. High-efficiency multi-junction cells coupled with radiation-hardened lithium-ion batteries provide continuous operation through eclipse periods.
- Thermal Management: In vacuum, heat rejection relies on large-area radiators. Assuming 1 kW of waste heat per kW of IT load, a 10 MW orbital module would require ~1,000 m2 of radiator surface coated with high-emissivity materials like Optical Solar Reflectors (OSRs).
- Radiation Shielding: Servers and electronics need protection against cosmic rays and solar particle events. Shielding strategies include spot shielding critical components with high-Z materials and using error-correcting code (ECC) memory for system resilience.
Orbital Infrastructure & Logistics
Operating a constellation of data centers in LEO or medium Earth orbit (MEO) requires a robust on-orbit logistics chain:
- Launch and Resupply: Terran R’s payload fairing (up to 7 m diameter) can deliver pre-integrated compute racks. Regular cargo flights—potentially via new commercial tug services—will carry spare parts and coolant fluids.
- On-Orbit Assembly: Robotic arms and autonomous grappling satellites, part of NASA’s recent OSAM (On-Orbit Servicing, Assembly, and Manufacturing) program, could streamline module integration without costly human EVA missions.
- End-of-Life & Debris Mitigation: Deorbiting large structures will rely on drag sails or controlled propulsion burns. Compliance with U.S. Space Force orbital debris guidelines is mandatory to avoid traffic jams in popular sun-synchronous lanes.
Economic Feasibility & Market Outlook
While orbital data centers promise abundant solar power and free‐air thermal dissipation, the business case hinges on:
- Cost per kW: Even with economies of scale, launch costs of ~$5,000 per kilogram for Terran R translate into a significant capital outlay. Schmidt is reportedly courting sovereign wealth funds and defense contractors to share risk.
- Latency & Connectivity: LEO-based compute must integrate seamlessly with terrestrial networks. Partnerships with satellite internet constellations like Starlink or Amazon Kuiper can minimize round‐trip delays to <10 ms.
- Regulatory Environment: Export control (ITAR) for advanced computing hardware and FCC licensing for high-bandwidth Ka-/V-band links are critical path items for orbital data center operators.
Expert Perspectives
“This is a natural evolution of cloud computing,” says Dr. Sara Nguyen, CTO at CloudOrbit Inc. “Edge servers in space combine the best of high‐density compute, renewable power, and global reach—ideal for AI inference at scale.” NASA’s Deputy Director for Commercial Low Earth Orbit Utilization, Dominic Cervelli, notes that such efforts align with the agency’s goal to foster a sustainable commercial space economy by 2030.
Looking Ahead: Timeline and Next Steps
Relativity Space aims to complete Terran R’s static-fire campaign by Q3 2025, with orbital launches targeted in early 2026. Meanwhile, Schmidt’s team is evaluating prototype micro-data centers—each hosting 50 AI-optimized GPUs—for a demonstration flight aboard Terran R. If successful, a full-scale orbital cloud could follow by 2028, reshaping how we power and deploy AI workloads.
While formidable technical, economic, and regulatory obstacles remain, Eric Schmidt’s gambit underscores a fundamental truth: as AI scales beyond the capacity of Earth’s grids, the final frontier may become the next data-center megahub.