“Taking resource-intensive infrastructure off Earth has been an idea for years, but it has required launch and satellite costs to come down. We are nearing that point,” said Will Marshall, co‑founder of Planet Labs. That inflection point is now drawing two of the world’s most prominent space entrepreneurs into direct competition over a new frontier: orbital AI computing.
Jeff Bezos and Elon Musk, already locked in a years‑long contest over rockets and satellite broadband, are racing to deploy AI-capable data centers in space. Bezos’ Blue Origin has been working on the technology for over a year, while Musk’s SpaceX is preparing an upgraded Starlink platform designed to carry AI computing payloads. The initiative comes at a time when the AI boom is pushing demand for processing power – and with it, an escalating strain on terrestrial energy grids.
The United States is currently home to more than 4,000 data centers, with heavy concentrations in Virginia, Texas, and California. These facilities are rapidly expanding to support generative AI models, hyperscale cloud services, and other compute‑intensive workloads. In 2024, U.S. data centers consumed 183 terawatt‑hours of electricity over 4% of the nation’s total and by 2030 that figure is projected to more than double. A single AI‑optimized hyperscale facility can draw as much power annually as 100,000 households, and the largest now under construction are expected to use 20 times that amount.
This growth in demand is already transforming the wholesale electricity markets. In areas such as northern Virginia, data centers account for over a quarter of overall grid demand, pulling up rates for homes and small businesses in their wake. The U.S. Department of Energy projects that by 2028, data centers could consume between 6.7% and 12% of all U.S. electricity. In some wholesale markets, they have driven up monthly residential bills by double-digit percentages, creating ripples across the economy.
Against this backdrop, orbital AI computing offers a tantalizing proposition: move the most energy-hungry processing tasks into space, where solar power is abundant and cooling can be managed without tapping terrestrial water or electricity supplies. Already, Google and Planet Labs are pursuing a proof-of-concept with plans to launch two solar-powered satellites in 2027 fitted out with Tensor Processing Unit chips to test a networked AI compute platform in orbit.
The technical challenges are daunting. The satellites carrying AI workloads must be ruggedized to resist radiation, temperature extremes, and latency constraints of space‑to‑Earth data transfer. They require high‑efficiency processors, advanced thermal management systems, and secure, high‑bandwidth communications links. Launch costs, even with the benefits of reusable rockets, still are a major capital barrier. And while solar power in orbit is plentiful, storing and managing that energy to enable continuous AI processing introduces another layer of engineering complexity.
For Bezos and Musk, the potential rewards go far beyond energy savings, however. Space‑based AI data centers could function like sovereign, globally accessible compute resources, unhindered by many of the regulatory and geographic limitations that circumscribe their terrestrial brethren. They would be able to process sensitive workloads above national jurisdictions, provide redundancy for Earth‑based networks, and offer low‑latency services to satellites, spacecraft and remote regions.
The competition plays into the individual strengths of the founders. For one, Blue Origin’s heavy‑lift New Glenn rocket-expected to debut any day now-could loft large, modular AI compute platforms into geostationary or medium Earth orbit. By contrast, SpaceX’s Starlink constellation, already the world’s largest satellite network, provides a ready‑made infrastructure for the distribution of AI workloads and downlinking results to customers worldwide. A revised Starlink generation featuring AI payload capacity could integrate compute directly into the comms layer, paving the way for a vertically integrated space‑cloud service.
Energy policy and grid reliability pressures add urgency to the race. With more than a trillion dollars invested in the grid over the last decade, utilities cannot keep pace as AI demand accelerates. Outages remain stubbornly frequent and electricity rates have risen more than 30% since 2020. By shifting part of the AI load off‑planet, orbital data centers could help relieve the bottlenecks that threaten both economic growth and energy security.
For now, the concept stays at an early developmental stage when prototype missions, like the 2027 launch by Google, will test the viability of solar-powered, AI-capable satellites. With Bezos and Musk joining the competition, acceleration is likely-and perhaps the next decisive battle in their long-running space rivalry will be over who builds the first functioning AI cloud in orbit, rather than who reaches Mars first.