Elon Musk Unveils Plan for a 100-GW Orbital AI Data Center Network
Elon Musk has revealed an ambitious proposal to build an orbital AI data center capable of generating up to 100 gigawatts of power—nearly one quarter of the average electricity consumption of the entire United States. The initiative aims to shift a significant portion of global AI computation away from Earth and into space.
A New Frontier for Starlink V3
According to Musk, the system will rely on next-generation Starlink V3 satellites equipped with compute modules capable of throughput levels approaching one terabit per second. These satellites would form a scalable orbital mesh designed to handle AI inference and potentially training workloads in the future. The concept marks a dramatic expansion of Starlink’s original mission, turning communications satellites into a distributed supercomputing layer.The Rapid Rise of Orbital Data Centers
Musk’s announcement comes amid a wave of space-based computing initiatives. Earlier this year, Starcloud—a member of NVIDIA’s Inception program—launched the first satellite carrying an NVIDIA H100 GPU. The company aims to build a 5-gigawatt orbital AI center powered by massive 4×4 km solar panels, claiming space-generated energy could be ten times cheaper than Earth-based equivalents.Google is pursuing a similar vision through Project Suncatcher, developed in partnership with Planet Labs. The plan calls for a cluster of 81 satellites arranged in a one-kilometer-wide formation, powered by high-efficiency solar panels and interconnected with laser communication channels. The system is expected to launch by 2027.
China’s “Three-Body Computing” Constellation
China has already deployed the first twelve satellites of its “Three-Body Computing” program, part of the broader Star Computing initiative targeting a fleet of 2,800 satellites. Together, they would deliver up to 1,000 peta-operations per second, with each satellite running an onboard AI model containing eight billion parameters. With 100 Gbps laser links and real-time orbital processing, China has taken a clear lead in operational space-AI capacity.Commercial Efforts Are Accelerating
Axiom Space plans to launch the first nodes of its Orbital Data Center (ODC) before the end of 2025. These modules will support high-bandwidth optical inter-satellite links at 2.5 Gbps and are aligned with the Space Development Agency’s communication standards. The emergence of multiple competing approaches suggests that orbital computing is rapidly transitioning from concept to industry.A New Industry Beyond Earth
Taken together, these developments represent the birth of a new sector: relocating energy-intensive AI computation off the planet. Low-cost solar energy, vacuum-cooled environments, and global line-of-sight connectivity offer clear advantages. Musk’s 100-GW vision is the boldest yet, and if realized, it could reshape not only AI infrastructure but the financial and geopolitical landscape surrounding compute power.Editorial Team — CoinBotLab