The idea of putting data centres in orbit sounds like science fiction, but serious money is now flowing into the concept.
Cowboy Space has raised $275 million to build space-based data centres. Bajobot, founded by one of the co-founders of Robinhood, is pursuing the same goal and has already secured a launch partnership with SpaceX. The thesis is straightforward: as AI workloads grow exponentially and terrestrial power and cooling constraints become binding, moving compute infrastructure into space, where solar energy is abundant and cooling is effectively free, starts to look rational.
But Bajobot's journey reveals the core problem facing every company in this space. The startup originally planned to beam solar energy from orbit back to Earth. It pivoted to data centres after recognising that the economics of space-based solar were not yet viable. Now it faces a different constraint: there are not enough rockets to go around.
SpaceX dominates the commercial launch market, and it has been raising prices over the past two years. The company's own Starlink constellation is consuming a significant share of its launch capacity, and it remains unclear how much room will be available on Starship for third-party payloads. The result is a crunch that is squeezing startups across the space sector.
Bajobot's response has been bold to the point of recklessness: it is now planning to build its own rocket. That is an extraordinary escalation for a company whose core business is supposed to be data centres, not launch vehicles. Building rockets is phenomenally expensive, technically demanding, and littered with the wreckage of companies that tried and failed. But Bajobot's logic is that relying on a single launch provider, one that is also a potential competitor for orbital infrastructure, is an untenable long-term position.
The broader market dynamics are real. AI training runs are consuming ever-larger amounts of energy. Terrestrial data centre construction is running into planning objections, grid capacity limits, and water supply concerns. Space offers a theoretical escape from all three.
The practical barriers remain enormous: the cost of launching hardware into orbit, the difficulty of servicing equipment that cannot be physically accessed, and the latency issues inherent in routing compute through space.
Related reading
- The $10 trillion race: How Nvidia and Alphabet are powering, and profiting from, the AI industrial revolution
- AI can write code and prove theorems. It still can't book me a restaurant
- YouTube is using AI to let brands hijack the moments viewers care about most, and creators should be paying a…
But the capital is flowing, the engineering talent is available, and the terrestrial constraints are tightening.
Somebody is going to make this work. The question is who survives long enough to get there.