Power-hungry AI systems already strain Earth’s electrical grids – and Sam Altman wants to solve that problem by taking them off the planet entirely. The chief executive officer of OpenAI has been exploring acquiring a reusable rocket company, such as Stoke Space, in a move that would open a new front in his long-running rivalry with Elon Musk and SpaceX.
According to people familiar with the matter, Altman approached Stoke Space this summer and proposed a multistage equity investment in the tens of billions of dollars that would give him control. Stoke-which was founded in 2020 by several former Blue Origin engineers-is working on the fully reusable Nova rocket, a competitor to SpaceX’s Falcon 9 and Starship in terms of launch cost efficiency. The reusability of Nova would significantly lower the cost of deploying extensive orbital infrastructure, an important factor if Altman does plan to establish AI data centers in orbit.
The rationale behind such a venture is self-evident: space offers near-constant solar exposure, unhindered by the interference of the atmosphere and a constant day-night cycle that makes terrestrial solar arrays so spotty. Panels in orbit can collect as much as eight times more power than their earthly counterparts in certain orbits. That would let AI clusters operate without the giant land footprint and environmental impact of ground-based facilities, while sidestepping the grid constraints that have forced companies like Microsoft to idle GPU inventories.
But the construction and operation of such data centers in orbit pose very serious engineering challenges. Concepts such as Google’s Project Suncatcher illustrate the difficulty: satellites need to be part of tightly clustered constellations, with separations of hundreds of meters in some cases, to achieve multiterabit-per-second optical links from node to node. Such links demand that received power levels be many thousands of times higher than in more traditional long-range systems, placing exacting demands for station-keeping and advanced DWDM transceivers. Thermal management, radiation shielding for semiconductors, and autonomous operation in the hostile orbital environment add further layers of complexity.
Another frontier is that of power transmission back to Earth: space-based solar power systems could beam energy via microwaves or lasers back to ground stations, but significant hurdles remain in the form of efficiency losses, atmospheric distortion, and safety regulations. According to the Office of Technology, Policy, and Strategy at NASA, capability gaps must be closed in autonomous assembly, in-space servicing, and efficient power-beaming for such systems to become commercially viable. Current timelines indicate that operational systems might not emerge until mid-century, although smaller-scale constellations dedicated to in-orbit computing could arrive sooner.
Reusable rocket technology is the linchpin for making these ambitions economically feasible. To date, SpaceX has demonstrated that boosters can fly up to 20 times and land both on land pads and drone ships. Stoke’s Nova will also be capable of at least that, with the first launch scheduled for 2026. By cutting launch costs and turnaround times, reusable rockets make the sustained deployment of orbital infrastructure-from solar arrays to AI compute clusters-a realistic proposition. Altman’s interest in aerospace isn’t new: as former head of Y Combinator, he oversaw early investment in Stoke Space, and has publicly mused on building a rocket company or even energy project at the Dyson sphere scale.
As it is, his comments to Theo Von’s podcast – Maybe we put them in space… it actually makes no sense to put these on Earth – underscores his belief that humanity’s future energy consumption will far exceed what can be generated on-planet. The rivalry with Musk adds a competitive edge to these ambitions. Musk has floated the idea of Starlink V3 satellites serving as AI data centers, while also pursuing brain–computer interfaces through Neuralink a field Altman has entered with Merge Labs. Both leaders see the fusion of AI and aerospace as a defining battleground for the coming decades, where mastery of reusable launch systems, orbital power generation, and distributed computing could determine technological dominance.
To investors and industry watchers, Altman’s possible jump to space is more than a publicity-garnering tiff. It represents a convergence of two high-capital, high-risk sectors-AI and aerospace-where breakthroughs in propulsion, energy harvesting, and orbital systems engineering could reshape both markets. Whether or not those Stoke Space talks resume, the trajectory is set: the next phase of the AI arms race may well be fought not in data halls on Earth but in the vacuum of low-Earth orbit.
