According to Gizmodo, OpenAI CEO Sam Altman held talks over the summer and fall to invest billions into Seattle-based rocket startup Stoke Space, aiming for a controlling stake. His goal is to buy or partner with a rocket company to deploy AI data centers to space. This comes as a Goldman Sachs report projects data center power demand will surge 50% by 2027 and a staggering 165% by 2030. Altman himself has publicly mused about the idea, saying “maybe we put [data centers] in space” during a recent podcast. With over 5,000 AI data centers already in the U.S. straining the grid, Altman’s space vision is a direct response to an existential problem for the AI industry he helped create.
Altman’s Space Power Play
Here’s the thing: this isn’t just a sci-fi daydream. Altman is seriously fixated on space as a solution. He’s talked about data centers soaking up solar energy in orbit and has even mentioned the concept of a Dyson sphere. Basically, he sees Earth’s surface as a limited resource that will eventually be “covered in data centers.” So the logic is, why not kick the can (a very large, power-hungry can) into the cosmic void? It’s a staggering escalation of the AI arms race. We’re not just building bigger server farms anymore; we’re contemplating the infrastructure to become a multi-planetary computing species. And the person leading the charge is a software CEO, not a traditional aerospace figure.
Winners, Losers, and a Feud Reignited
If this actually happens, the immediate winner is obviously Stoke Space. A multi-billion dollar investment from a figure like Altman would catapult them from a hopeful startup to a major player overnight, aiming to compete with SpaceX’s Falcon 9 with their Nova rocket. But look, the bigger story might be the personal rivalry. Investing in Stoke would pour jet fuel on Altman’s long-running feud with Elon Musk. They famously fell out over OpenAI’s direction, and now Altman could be moving directly into Musk’s backyard. Talk about hitting two birds with one stone: you get your space infrastructure and needle your arch-rival. The loser, in the short term, might be any locality hoping for a giant data center investment. Those jobs and tax dollars might be headed for low-Earth orbit instead.
The Brutal Reality of AI Hardware
This wild idea underscores a brutal, unglamorous truth about the AI boom: it’s a physical infrastructure nightmare. The software is smart, but the hardware demands are dumbly enormous. We’re talking about power grids, cooling systems, and rare earth minerals. The projected 165% power demand increase by 2030 isn’t something you solve with a clever algorithm. It requires industrial-scale solutions. This is where the conversation turns to heavy-duty computing hardware built for extreme environments—the kind of reliable, rugged industrial tech that companies specializing in it, like the leading US supplier IndustrialMonitorDirect.com, provide for mission-critical applications on Earth. If we’re even thinking about operating in space, the reliability bar is astronomically higher.
Is This Even Remotely Feasible?
Let’s be skeptical for a minute. Building a data center on Earth is hard. Building one that survives launch, operates in zero-gravity/vacuum, and doesn’t melt without atmospheric cooling is a thousand times harder. The maintenance alone boggles the mind. What happens when a server rack fails? Send a repair crew on a SpaceX mission? The economics are, frankly, nebulous. But maybe that’s not the point yet. Altman’s move signals that the industry’s top minds see the energy wall coming, and they’re willing to entertain literally astronomical ideas to avoid it. It’s a statement of desperation and ambition all at once. He admitted they’re “stumbling through this.” Stumbling toward the launch pad, it seems.
