For defence organisations trying to keep command and control working, the challenge is no longer a lack of data. It’s whether that data – and the systems that process it – can be trusted when networks are under pressure from malicious actors.
A growing number of examples have made that axiom clear. In recent conflicts, attackers have tended to go after networks and data centres early, usually through cyber means. Even when those systems stay up, they can be slowed down, interfered with, or quietly tampered with. That reality is turning ground-based compute into a weak point and forcing defence organisations and others handling sensitive data to reckon with a challenging question about how they provision their networks. Can systems that live only on Earth can still be relied on in a fight?
This is why interest in off-planet data centres is growing: defence organisations need ways to keep running operations even when ground systems are disrupted or compromised. The focus, notably, is on resilience under cyber pressure rather than on space technology for its own sake.
That idea fits with how space is already used in competition short of war. Interference does not have to involve missiles or physical damage to be effective. States can disrupt satellite operations, interfere with signals, or target the ground systems that support them without crossing obvious red lines. Much of that activity is digital, aimed at software, control links, and supporting infrastructure rather than hardware in orbit.
As Todd Harrison, a senior fellow at the American Enterprise Institute, has put it, orbit is “ideal” for this kind of pressure precisely because it often stays below the threshold that would trigger a direct response.
From satellites to orbital compute
For a long time, satellites mostly gathered data and sent it back to Earth for processing, but that setup is starting to look fragile. Links get jammed, ground stations are easy to target, and cloud systems rely on power and fibre routes that aren’t hard for adversaries to disrupt. The more steps involved, the more there is to interfere with – which is why some work is now being pushed closer to the sensor, and sometimes off the ground altogether.
Major technology firms are already testing the idea. Microsoft’s Azure Space division has partnered with Loft Orbital to validate cloud software in orbit, aiming to make it “simple … to deploy AI applications in space at scale”; Amazon’s AWS Ground Station offers a commercial model for downlinking and processing data without building dedicated global ground networks; and Google is exploring its own space-based, scalable AI compute architecture through Project Suncatcher, a research initiative to deploy solar-powered satellite constellations equipped with AI accelerators that could one day scale machine learning in orbit.
Defence-leaning firms are also dipping their toes in orbit. Anduril, best known for AI-driven systems, is adapting its autonomy and sensor fusion stack for space applications, and Kayhan Space is building space-traffic management tools to prevent collisions among proliferating defence and commercial satellites. In the UK, innovators such as Satellite Vu are deploying thermal imaging satellites with onboard analytics for infrastructure and security monitoring.
Many of these companies are backed by investors who see space compute as a clear dual-use opportunity, including funds such as Space Capital and Generation Space (formerly Seraphim Capital), as well as strategic backers like Lockheed Martin Ventures.
Governments are moving in the same direction, though at a more measured pace. In the US, DARPA’s Blackjack initiative has been testing ideas around resilient, proliferated LEO networks that could support more autonomous processing in orbit. In the UK, the Ministry of Defence has awarded £65 million to Borealis, a system designed to more reliably integrate data from multiple space sensors. UK Space Command has also shifted its language, describing space as “integral” to multi-domain operations rather than a supporting capability – an acknowledgement that resilience, including cyber resilience, no longer stops at the atmosphere.
Why it matters – and why it’s risky
Those backing off-planet compute usually come back to the same two arguments. Pushing some processing into orbit means less dependence on a small number of high-value ground sites. It also allows data to be handled closer to the sensor, speeding up decision-making and reducing the frequency with which information must travel across links that are routinely interfered with. Taken together, a mixed ground-and-space setup is simply more complicated to knock out with a single cyber hit.
That doesn’t mean it’s an easy fix. Space is a hostile place to run hardware, and once systems are launched, they’re hard to touch. Making them tough enough to survive adds cost and weight, and components don’t always behave the way planners would like over long missions. Launch prices may be falling, but it’s still not clear that putting compute into orbit is better than spending the same money making systems on the ground harder to knock over.
What’s more, cyber risk doesn’t go away just because systems are in orbit – and in some ways, it gets harder to deal with. Once a satellite is launched, fixing a compromised system is slow at best and often impossible, and any weakness baked in before launch can sit there for years. That’s why defence teams are starting to look at things like zero-trust designs, post-quantum encryption, and even the idea of repairing hardware in orbit. It follows the same thinking as cyber defence on the ground – assume something will break and work around it – but it also adds layers of complexity that attackers tend to exploit.
What’s actually at stake
Stripped of the hype, this is not really a story about putting data centres in space. It’s about trust.
Defence organisations are trying to work out where their most critical data can still be processed, verified and acted on when networks are under constant pressure and systems cannot be assumed to be clean. Governments are testing new architectures. Industry is building pieces of the stack. Analysts remain divided on cost, sustainability, and risk.
But it’s getting harder to argue that keeping all critical compute on the ground is still the safe choice. Cyber pressure isn’t something that comes and goes anymore – it’s just there. That’s why space keeps coming up in these discussions. Not necessarily because it’s secure, but because the systems everyone relies on down here are proving easier to interfere with than many expected.
Seen that way, off-planet compute doesn’t look especially bold. It looks like a workaround.
