What are the risks and benefits of Elon Musk’s vision of data centres in space?

Down here in Australia, Commbank estimates investment in data centres and AI could reach between $A150 billion and $155 billion by 2030. Driven by artificial intelligence and cloud computing, this reportedly equals roughly 6 gigawatts of potential capacity in development and serves as a primary driver of national business investment. Specific major commitments include Microsoft's $A25 billion push and Amazon Web Services' $A20 billion roadmap over the next few years.

Even more eye-watering is the forecast by Goldman Sachs that global AI-related investment is expected to hit the $US 1 trillion threshold on an annual basis this year, with specific tech-infrastructure and hyperscale capital expenditures projected to reach or exceed $1.1 trillion annually next year. 

Cumulative global AI capital expenditure (it just keeps growing) is projected to total an estimated $US 7.6 trillion over a multi-year build-out period running from 2026 through to 2031. This spending covers terrestrial data centres, specialised AI chips, power infrastructure, and networking hardware built on Earth. 

Such figures could make a dent in even Elon’s massive pay packets, such as his latest long-term Tesla compensation package, which has a theoretical maximum value of up to $US 824 billion (originally estimated up to $1 trillion).

So what could weigh on Elon’s lofty ambitions?

In the shorter term, at least, there are some key bottlenecks to be overcome. (These are widely known – I make no claim to originality or to presenting a technically complete list):

• Launch and Mass Constraints: Lifting millions of tonnes of heavy high-performance computing hardware requires upscaling rocket launch frequencies and regularity by multiple orders of magnitude.
   
• Economic Parity: Current estimates also show orbital deployment costs running multiples higher than ground-based builds until launch economics and in-space manufacturing mature.
   
• Hardware Lifecycles: Repairing or replacing advanced AI accelerators (like NVIDIA platforms) in orbit is exponentially harder than servicing terrestrial servers.
   
• Power Supply. Yes, solar power can theoretically feed large data centres in space, providing continuous 24/7 power without weather disruptions or night cycles. But while early prototype networks and small edge-compute units are being explored, scaling to massive hyperscale facilities faces steep technical and financial hurdles.
   
• Cooling. Bodies such as the World Economic Forum agree cooling data centres in space is technically feasible at a small scale, but it is extremely inefficient and impractical for large-scale or AI facilities. While space itself is cold, it is a near-perfect vacuum, meaning standard convection and conduction cooling do not work. Heat can only be removed via infrared radiation, which they say currently requires massive external radiator surfaces.

So it may be that cheaper and easier-to-use green energy on Earth – plus new cooling liquids and D2C chip cooling – may work better to cool servers down here, rather than cooling in orbit.

And will putting data centres into space ease security and environmental concerns anyway – or worsen them to the point where reputational damage and insurance become stumbling blocks? 

There are two parts to this answer that occupy my thoughts – split into cyber and physical threats. 

While some proponents argue space isolates infrastructure from physical ground attacks, experts note that cyber threats remain location-agnostic. So we may have:

• Persisting Cyber Threats: Any orbital facility connected to Earth remains vulnerable to malware, zero-day exploits, phishing, and API hijacking, which occurs when an attacker steals, predicts, or manipulates authentication tokens or API keys to gain unauthorised control over an application's data and backend systems.
   
• Command and Control Risks: Satellites and orbital clusters face distinct risks such as signal jamming, spoofing, and payload hijacking through vulnerable ground stations (which are covered further ahead in this article, as ongoing physical risks)

Space communication must also consider:

• Communication Latency: Delays in laser or radio link transmission complicate real-time threat detection and incident response from Earth.
   
• And how could you insure a data centre in space? How do you adapt “traditional” aerospace risk models – blowing up on launch pads, for example - to cover ultra-expensive AI hardware, cosmic radiation hazards, and specialised launch phases, through global insurance markets like Marsh and Lloyd's of London. Sensibly, startups and brokers are reported to be already holding preliminary risk briefings to establish how these novel assets can be underwritten. Space startups – including Lonestar Data Holdings, Orbital, Starcloud, and Cowboy Space, alongside major players like Blue Origin and SpaceX – are reported to have held preliminary risk talks with brokers and underwriters to figure out how to insure orbital AI data centres

Physical and Environmental Threats

• Disputes over Orbits: Apart from potential commercial conflicts of interest, space is an increasingly shared and even militarised domain where adversaries can deploy anti-satellite weapons, dazzling lasers, or kinetic strikes. The question then arises: is it easier to provide security on your own territory than to protect a satellite, even one in a low orbit path 500-2000 km above the earth as it traverses above us at about 8km a second. A low Earth orbit (LEO) satellite covers about 40,000 to 43,000 km in one full circuit around the Earth, so there is quite a lot of space to cover.
   
• Orbital Debris – Kessler Syndrome: High-density mega-constellations drastically increase collision risks with space junk, which can trigger cascading destruction across an orbital band. This was a warning contained in the Kessler Syndrome, a theoretical scenario proposed by NASA scientist Donald J. Kessler in 1978. It describes a chain reaction where the amount of space junk in low Earth orbit becomes so dense that objects crash into each other. Each crash makes thousands of new tiny pieces, which cause more crashes, making whole parts of space unsafe to use.
   
• Physical Protection of Collateral Facilities on Earth – an area that encompasses the expertise of my company’s parent, the Royal Boon Edam physical security entrances group. We operate in 27 countries, including Australia and New Zealand, with some of the world’s biggest companies including some of the biggest data centre operators and user groups. We can see that, for Earth-based facilities, there is the obvious ongoing and hugely expanding physical need to protect Earth-bound infrastructure and State and private businesses where data centre outputs are employed locally, nationally, and internationally. These range from entire data networks, logistics chains serving them (and, if space technology does lift off) safeguards of launch facilities, logistics and support facilities, plus the ongoing needs of corporate facilities using on Earth the data outputs generated in space.
   
• National and international compliance and security standards and statutory obligations. There are huge commercial and geopolitical issues at stake here, with data centres and AI, not just commercial interests. This is why national and world bodies are endeavouring to get in early to set standards before issues become a problem. Many of these new laws for artificial intelligence change how systems handle information and to help ensure operational resilience. Service disruptions in cloud or edge data centres impact continuous delivery to not only commercial operators, but also vital public services and military capabilities. The Australian Government, for example, is developing a mandatory national data centre framework and supporting legislation following voluntary guidelines published under the Department of Industry, Science and Resources. Legislation to formalise the mandatory standards is expected to be introduced and pass Parliament in early 2027. True, this legislative advance does not explicitly address or regulate data centres in space. Yet. 
   
• Foreign threats to national interests. The Australian Security Intelligence Organisation (ASIO) has warned that foreign powers and state-backed hackers are targeting Australia’s critical infrastructure, telecommunications, and digital data storage networks to conduct espionage and pre-position for high-impact sabotage. ASIO Director-General Mike Burgess has revealed that state-sponsored actors have been actively scanning and probing Australian networks across communications, energy, transport, and data services. ASIO has warned that adversaries are no longer focused solely on stealing data or traditional espionage; they are actively “pre-positioning” inside critical digital networks to create a disruption or destruction capability that can be activated during a future geopolitical crisis.

So the question arising from the points above is: How do data centre and AI interests respond? And how does transferring the issue to new territory – space – affect the equation?

The Bottom Line

Twenty years ago, opinion on Elon Musk was already deeply divided. He was viewed as a bold, ambitious visionary by supporters for backing Tesla and SpaceX – but dismissed as an overconfident eccentric by others who thought his electric car and rocket ventures were doomed to fail.

Electric cars are now a reality. Only time will tell who is right about data centres in space too. Being right with one bet, of course, doesn’t guarantee success with the next venture. And the stakes here are geopolitical. 

And if Elon is right, how long will it take? A lot of money is on the line – and a lot of reputational risk too. As a result, a lot of business interests will be paying very close attention to events as they emerge, some perhaps making plans for maybe five or 10 years or even decades down the line, when the risk issues (some of which I have mentioned in this article) are clearly defined and satisfactorily resolved. Easier said than done.

And, at the end of the day, we may also find that, rather than lessening cyber and physical security risks for the data and AI industry, it may actually increase them by introducing new terrestrial user and supply chain risks, in addition to the space environment risks. Distributed environments increase the risk of unauthorised access or data leaks.

My bet, as a professional in the physical security industry, is that whatever eventuates, the need for protections and risk management will boom, rather than decline. 

Prudent risk management will see products and services evolve to fit whatever new model of data centres and AI environments emerge.

Such evolution is the history of business, and I don’t see it changing as we move forward into the future.

Download Boon Edam’s white paper, “Best practices for data centre security and efficiency” for more insights into data centre security. 
 

About the Author

*Mike Fisher is Managing Director of Boon Edam Australia, which is part of the privately owned international Royal Boon Edam group, which provides architectural revolving door and layered security solutions to some of the world’s largest companies, Fortune 500 companies, and companies in Australia, New Zealand, and Papua New Guinea including financial, data and telecommunications, Federal and State Government, hospitality, health and age care, logistics, retail, and distribution facilities. Boon Edam Australia operates under Master security licence number: 000104487.