By 2030, artificial intelligence (AI) computing in orbit is projected to be less expensive than comparable services on Earth, according to a forecast from the research group 33FG. The analysis examines the cost dynamics of launching space-based equipment to utilize solar energy, contrasting it with the expenses associated with terrestrial resources.
Currently, delivering cargo to high orbit costs approximately $2,000 per kilogram, with satellites generating electricity at rates ranging from $18 to $26 per watt. This is nearly double the costs seen in terrestrial data centers, which average around $12 per watt. However, if launch costs were to be halved, the price of energy produced in space would be on par with that of Earth-based alternatives. At a launch cost of $500 per kilogram, space energy could be about 30 percent cheaper, and at $100 per kilogram, it could even reach 50 percent savings.
The feasibility of such reductions hinges on the development of reusable vehicles like SpaceX’s Starship, which can facilitate orbital refueling. Analysts suggest that these advancements could make the cost-effective deployment of space-based computing a reality within the decade.
In their study, 33FG modeled four types of orbital architectures: Starlink-class low Earth orbit (LEO) systems, Starlink-class high Earth orbit (HEO) systems with consistent sunlight exposure, Compute-Optimized Starlink (HEO) systems tailored for computing with lightweight structures, and a promising ultra-light architecture termed Thin-PV Frontier (HEO).
Deploying equipment to high Earth orbit necessitates refueling in low Earth orbit, making this approach about 1.5 times more costly than LEO. Initial costs for Starlink-class systems reach roughly $2,000 per kilogram, resulting in an electricity price of $18 to $26 per watt. However, if launch costs decrease to about $500 per kilogram, Starlink-class systems in high orbit could become economically viable. Meanwhile, the Compute-Optimized Starlink architecture could reach parity with terrestrial systems at about $1,000 per kilogram, surpassing Earth-based infrastructure if launch costs fall below $500 per kilogram.
With launch costs at $100 per kilogram in high Earth orbit, orbital architectures could generate electricity at rates between $6 and $9 per watt, making them 25 to 50 percent cheaper than conventional data centers on the ground. Further reductions in launch prices would have diminishing returns on the economics, as system efficiency would increasingly depend on the cost of the hardware itself rather than rocket launches.
Comparative analyses yielded significant insights. The Compute-Optimized Starlink (HEO) system becomes competitive with terrestrial data centers at a launch cost of around $500 to $1,000 per kilogram. As launch prices decrease further, the advantages for space systems grow markedly. The Starlink-class (HEO) architecture aligns with terrestrial infrastructure when launch costs hit $500 to $600 per kilogram. The Thin-PV Frontier, while promising, faces challenges due to high equipment costs and is expected to remain less competitive as prices evolve into the 2030s.
Analysts assert that orbital energy represents a critical future resource for humanity. Space offers limitless solar energy and ample room for hardware placement, in stark contrast to the growing scarcity of both energy and physical space on Earth. However, achieving a well-designed architecture that balances equipment and launch costs against terrestrial options remains a significant challenge. Currently, the most mass-efficient designs do not coincide with cost-efficiency, necessitating innovative solutions in the field.
Notably, in November, Google announced plans to establish a satellite network in low Earth orbit aimed at harnessing solar energy to power data centers. In May, China successfully launched 12 satellites as part of a broader initiative to create a network of orbital supercomputers, highlighting the global race toward utilizing space for computational power.
Foxconn Invests $569 Million in Wisconsin to Expand AI Hardware Operations, Adding 1,300 Jobs
AI Hardware Market Grows 30% in 2025, Driven by Generative AI and Edge Computing Demand
Conflow Power Group Launches iLamp: Solar-Powered AI Streetlights with Near-Zero Carbon Footprint
Neoclouds Surge as Anthropic Partners with Fluidstack for $50B AI Data Centers
Chinese Startup Zhonghao Xinying Launches GPTPU, Claims 1.5x Speed Boost Over NVIDIA A100
