AI is beginning to push up against the physical limits of today’s data centre infrastructure. Power, cooling and available land are all under increasing strain, as demand for AI workloads accelerates. In some regions, energy supply is already becoming a barrier to building new capacity. All of this pressure explains why more radical ideas, including data centres in space, are starting to be explored. Not because they are ready to replace infrastructure on Earth – but because they reflect how quickly the conversation is shifting.
But despite the speed at which orbital computing is moving from concept towards reality, there are still major questions to answer. The sustainability and scalability case is not yet proven, and significant engineering and connectivity challenges remain. As latency stretches beyond Earth and reliability becomes harder to guarantee, it raises a more immediate question: why look to space when many organisations are still focused on making today’s infrastructure resilient and fit for purpose?
Why space changes the infrastructure equation
The real shift is conceptual, because moving above the atmosphere removes some of the limits that have always shaped data-centre infrastructure. On Earth, compute is constrained by land, power grids, cooling requirements and the pace of construction, but in orbit those assumptions start to change, with abundant solar energy, highly efficient vacuum-based cooling and expansion no longer tied to physical space. This is what makes the idea so compelling, as physics begins to work in favour of infrastructure rather than against it, opening up the possibility for enterprises to scale in line with launch capacity instead of being held back by the timelines and constraints of construction on the ground.
Recent experiments have made that idea feel less abstract. Starcloud recently launched a satellite containing an Nvidia H100 GPU and trained a small AI model in orbit, showing that the industry is testing whether AI workloads could run beyond Earth.
But orbital compute is not simply about putting servers in space. It is about asking whether the current model can keep scaling, or whether compute, energy and networks will need to be designed together at global scale.
The sustainability and scalability case
The sustainability case is the most obvious place to start. If solar generation in orbit can produce more consistent power, and if vacuum cooling can reduce one of the largest burdens of terrestrial data centres, then space starts to look less like a novelty and more like a serious infrastructure question. It would arguably also be more cost-effective over time, because building a data centre on Earth means securing land, grid capacity, cooling systems and connectivity, all of which can take years before capacity is available. In orbit, the constraints change, and that changes the economics of scale.
But this does not make orbital data centres an automatic answer. The engineering challenges remain enormous. Hardware must survive radiation, infrastructure cannot easily be repaired, and launching compute into orbit remains extremely expensive.
The ambition is real and the experiments are happening, but for now this matters more as a signal than a solution.
What this means for enterprise behaviour
The other major shift is latency, because most enterprise infrastructure decisions today are still built around geography. Organisations ask where their users are, where their cloud regions sit, where their data is generated, and where workloads need to run.
Orbital compute stretches that model. As latency stretches from continental to planetary, enterprise behaviour will change dramatically, not because every workload will move into space, but because some workloads could become less tied to place. In that environment, real-time global inference becomes achievable, automation can operate across borders without the same constraints, and applications like continuous climate modelling move out of theory and into practical, scalable use.
This is where the “tech for good” argument becomes strongest, because if space-based compute can support climate intelligence, global monitoring and sustainable infrastructure growth, then the case becomes meaningful. But the value will depend on how the technology is used, who can access it, and whether it supports a resilient digital ecosystem.
Why connectivity becomes more important, not less
The data still needs to move efficiently between Earth and space. Even if compute moves to orbit, the importance of global connectivity will only increase. The next generation of digital infrastructure will not just be defined by compute capacity; it will be defined by how compute, energy and networks are designed to work together at global scale.
For most organisations, the priority remains on the ground, where AI is already increasing demand on networks, cloud environments and data infrastructure. Power availability, scale and sustainability are becoming harder to ignore, which is why enterprises need flexible, resilient networks that can evolve as demand grows.
Data centres in space are no longer science fiction, but they are not the answer yet. They show where infrastructure may eventually go, but real progress right now still comes from getting the fundamentals right on the ground.
Michel Muizer
Michel Muizer is Product Director at Expereo where he leads strategy and portfolio across DIA, Broadband, Enhanced Internet, 5G/FWA, LEO, P2P and emerging NaaS capabilities.
