The long-running rivalry between Elon Musk and Jeff Bezos is entering a new phase, and this time the conflict involves the future of artificial intelligence infrastructure. Both Musk’s SpaceX and Bezos’s Blue Origin are now developing AI data centers designed to operate in Earth’s orbit, a concept that only recently moved from speculative to feasible.
The catalyst is AI’s accelerating appetite for power. The energy required to train and operate large-scale models already challenges Earth-based electric grids, and forecasts point sharply upward. Goldman Sachs estimates AI-driven electricity demand could grow 165% by 2030, a trajectory that has pushed technology companies to explore alternatives that bypass Earth’s constraints. The idea of hosting computing clusters in orbit, powering them with uninterrupted solar energy while using the cold vacuum of space as a cooling system, is increasingly viewed as a viable answer.
Growing Enthusiasm
In the past year, enthusiasm for off-planet data centers has intensified. Former Google CEO Eric Schmidt has funded his own orbital-computing venture. Google and Planet Labs plan a 2027 mission to test satellite-based AI chips.
SpaceX’s contribution is aided by existing infrastructure. Musk has said that upgraded Starlink satellites, already equipped with laser links for high-speed data transfer, could evolve into compute platforms. The company reportedly floated the concept during discussions with investors as part of a share sale. SpaceX now operates more than 6,000 Starlink satellites, giving it a sizable base for retrofitting AI capabilities.
Blue Origin’s approach is just as ambitious. According to reporting shared with investors and industry observers, the company has quietly spent more than a year developing orbital data center components. Bezos has long described a vision of moving heavy industry into space, and argues that gigawatt-scale orbital clusters are achievable within 10 to 20 years. He cites the advantages of 24/7 solar power and the absence of weather-related intermittency, conditions impossible to replicate on Earth.
Big Obstacles, Big Plans
Yet the obstacles are formidable. Even with reusable rockets lowering launch costs, placing heavy computing infrastructure into orbit remains wildly expensive. Heat management in space is tricky. Without atmospheric convection, servers must dissipate thermal loads through alternative radiative systems, which adds weight. Latency of data transmission from space could limit certain AI applications. Critics suggest that these physical and economic constraints make orbital computing a long-term research project rather than a near-term commercial solution.
Still, momentum is building. Google’s estimate that a one-gigawatt orbital data center may require thousands of satellites underscores the scale of the challenge. Yet if such constellations could be assembled, they could reshape the launch industry by increasing flight cadence and encouraging new forms of modular in-orbit assembly. SpaceX, with its high-frequency Falcon 9 operations, and Blue Origin, preparing its New Glenn heavy-lift rocket, are positioned to benefit from that shift.
And demand on Earth is certainly not easing. That pressure gives the space-based concept its appeal, given that the alternative may be simply running out of Earth-based capacity.
For now, orbital data centers remain a bold experiment, supported by billionaires who are accustomed to pursuing larger-than-life initiatives. Whether these ideas become essential infrastructure or remain big dreams will depend on engineering breakthroughs still ahead. But the competition is underway, and well-funded contenders are eying the prize.
