AI Infra In Space? Your Next Data Centre Might Be In Orbit

Did you know that while we are all talking about underwater data centres and nuclear-powered facilities, some companies are aiming even higher, quite literally? There is a whole new frontier emerging where data centres are not just going underground or hitching a ride with small modular nuclear reactors.

Some companies are seriously planning to put data centres up in orbit. At first glance, it sounds like a plot twist from a science fiction series. But scratch beneath the surface, and the logic starts to reveal itself. In space, you have near-continuous sunlight to power solar arrays.

You have no land acquisition battles, no water scarcity debates, and no neighbourhoods protesting against another hyperscale facility coming up next door. In theory, orbital data centres promise an escape from many of the constraints now choking infrastructure expansion on Earth.

At Greyhound Research, we are seeing growing investor and enterprise curiosity around this idea, driven less by novelty and more by the hard limits now facing terrestrial infrastructure globally.

This is not happening in isolation. It is emerging at a moment when AI workloads are exploding, when power grids in key markets are under strain, and when even the largest technology companies are being forced to rethink where the next megawatt will come from.

Against that backdrop, space has re-entered the infrastructure conversation not as fantasy, but as a strategic possibility. Let’s be clear. This is not just a wild idea circulating in whitepapers. Real companies are already at work. Names you recognise, like Google, and names you may not yet, like Aetherflux and Starcloud, are exploring orbital data centres in earnest.

Google is working on Project Suncatcher, which examines running AI accelerators in orbit and networking them together using laser links. There is already a satellite in space today carrying a high-end NVIDIA GPU running inference workloads. This is not hypothetical. It is happening now.

But this is where investors and technology leaders need to pause. Just like data centres on Earth are not simply about plugging into more megawatts, orbital data centres are not simply about having more sunlight. Power generation is only half the equation. The harder problem is heat.

In space, there is no air or water to carry heat away. The only way to cool high-performance systems is through radiation, which is slow and demands enormous surface area. Every additional unit of compute, therefore, requires disproportionately larger radiator structures, more mass, and higher launch costs.

In other words, the real constraint in orbit is not energy. It is physics. Then there is radiation. Modern AI chips are not designed for space. They face constant exposure to cosmic rays that can flip bits, degrade components, and shorten hardware lifespans. Unlike Earth-based data centres, there is no technician on standby to replace a failed board.

Reliability must be engineered through redundancy and error correction, which inevitably reduces usable performance and raises costs. Scale is another hard reality. Much of the excitement around orbital data centres quietly assumes they can one day rival hyperscale facilities on Earth.

That is unlikely in the near term. Even with reusable rockets, launch economics remain a major bottleneck. These visions depend heavily on future launch systems operating at unprecedented cadence and cost efficiency. That is a dependency no serious infrastructure strategy should ignore.

This does not mean orbital data centres are a dead end. It means they are a specialised layer. The most credible near-term use cases are narrow but valuable. Processing Earth observation data in orbit to reduce latency and bandwidth. Defence and intelligence workloads where resilience and proximity to sensors matter more than cost.

Disaster response analytics that cannot wait for downlink windows. These are scenarios where orbital compute offers a genuine advantage. Mainstream enterprise workloads do not fall into that category.

Despite the talk of cloud regions in space, most enterprise applications gain little from orbital execution and inherit significant complexity around governance, integration, and compliance. Which brings us to governance.

Who regulates data processed in orbit? Which country’s laws apply when computation happens beyond national borders but on satellites registered to specific states? How do data sovereignty and export controls adapt? These questions will directly shape adoption.

Today, the answers are fragmented. Space law was not designed for cloud computing. Data protection regimes were not written with orbital execution in mind. Enterprises and insurers will remain cautious until these grey zones are clarified.

There is also a sustainability conversation that cannot be ignored. Rocket launches have environmental costs. Satellite manufacturing and re-entry affect the atmosphere. Large constellations raise concerns around debris and astronomical interference. Shifting computing to orbit does not erase environmental accountability. It relocates it.

So where does this leave us? Orbital data centres are neither science fiction nor a silver bullet. They are an early-stage infrastructure experiment driven by real pressures but constrained by hard limits. They will not rescue the world from AI’s energy appetite in the next five years. They may, however, become a critical adjunct layer for specific, high-value workloads.

For investors, the discipline is straightforward. Strip away the excitement and look at the basics. Test the unit economics, the launch assumptions, and the hardware lifespan. If the numbers do not hold up like any other infrastructure investment, space will not magically make them work.

For technology leaders, the right response is not dismissal but restraint. Track the milestones. Watch who can demonstrate sustained compute under load, reliable thermal management, secure networking, and paying customers.

For policymakers, the message is urgency without illusion. Space-based infrastructure will intersect with national security, economic competitiveness, and digital sovereignty sooner than many expect. Regulatory frameworks need to evolve now, not after the first failure.

Orbital data centres are not the next cloud. They are something else entirely. If we approach them with realism and patience, they may earn a place in the global digital stack. If we oversell them as an escape from terrestrial limits, they will disappoint just as quickly as they have captivated.

In a world where infrastructure is strategy, orbit is not an escape from complexity. It is complexity, amplified. The opportunity is real. The risks are real, too. The next five years will decide which one defines the outcome.

Disclaimer: The views expressed in this article are solely those of the author and do not necessarily reflect the opinion of NDTV Profit or its affiliates. Readers are advised to conduct their own research or consult a qualified professional before making any investment or business decisions. NDTV Profit does not guarantee the accuracy, completeness, or reliability of the information presented in this article.