Will orbital data centers soon be a reality?

Jeff Bezos, the founder, and former CEO of Amazon, suggested a few years ago that to limit the impact of global warming on Earth, all heavy industry should be transferred to space. This is not a new idea: way back in the late 1960s, NASA was exploring the possibility of building huge structures in space, such as solar power plants. The concept was first introduced in 1968 by the American engineer Peter Glaser. With climate change now taking hold, installing large orbital platforms, including these famed solar power plants, could be an option.

Within the next decade, however, we could also be considering putting data centers in orbit. At the end of 2022, there were 264 data centers in France. They consume a significant amount of power because cooling them requires a lot of energy. According to a European Commission report published in 2020, the electricity consumed by European data centers rose from 53.9 to 76.8 TWh/year between 2010 and 2018. The report goes on to say that this consumption is likely to increase by a further 21% by 2025, to around 92.6 TWh/year.

Over the next ten years, terrestrial data centers could consume a further 15-20% of energy. The most energy-hungry aspect is not data storage, but the computing performed by microprocessors. So, to limit their environmental footprint, why not install them straight into space? At the end of 2022, the European Commission asked the Franco-Italian company Thales Alenia Space (TAS), a joint venture between Thales and Leonardo, to explore this option.

Similar size to the ISS

In January 2023, TAS began work on the European Commission’s initiative known as ASCEND (Advanced Space Cloud for European Net zero emission and Data sovereignty). Several companies have also joined forces with TAS, bringing complementary expertise in the environmental field (Carbone 4 and Vito), cloud computing (Orange, Cloudferro and Hewlett Packard Enterprise Belgium), launch vehicles (ArianeGroup) and orbital systems (Airbus DS and the German Aerospace Centre, DLR).

This project, which could help achieve the European Green Deal target of carbon neutrality by 2050, has two objectives. The first is to demonstrate that deploying future data centers in sun-synchronous orbit (SSO) at 1,400 km above the Earth’s surface (as suggested by TAS) emits less CO2 than those installed on the ground. Initial information provided by the Franco-Italian company suggests that installing data centers in SSO orbit is required to overcome data transmission latency constraints and to maximise solar energy yield.

The second objective is to determine the most appropriate – and greenest – launch solution for putting them into orbit around the Earth. The study is still only at the preliminary stage, but on paper “there’s nothing to stop it being done”, says Yves Durand, Director of Technology at TAS and leading the ASCEND project. In technical terms, we’re talking about structures that could be as big as the International Space Station (ISS), or roughly the size of Wembley Stadium in orbit. A single data centre building block would need to be the size of a large truck container, weighing around 30 tonnes, to provide the functionality required for Earth-based needs.

“To supply it with electricity, we need solar panels about the size of a football pitch,” explains Durand. They could also be powered by future space-based solar power plants, which the European Space Agency (ESA) is also looking into as part of its SOLARIS project. TAS is also studying whether using orbiting power plants is feasible. The idea is to use the energy generated outside the Earth’s atmosphere to power the data centre, and to link up with the ground only for high-speed data transfer over optical communications.

Need for a super launcher

To keep it financially viable, the system being considered is based on the use of artificial intelligence and future automated orbital tugs, which could become commonplace within the next decade. In the short to medium term, there is no plan for any human intervention. But the idea is to provide a structure in space that is highly modular.

This will ensure that components that need replacing can be returned to Earth by a robotic vehicle. From 2026, these techniques will be tested in orbit with the planned EROSS IOD (European Robotic Orbital Support Services In Orbit Demonstrator) vehicle. If this experiment is successful, a larger-scale project could be given the green light by 2035. But whatever form the first orbital data centre takes, if it does materialise, it will inevitably require the use of new launch vehicles. This is the second point. And it will very probably be one of the successors to Ariane 6, whose maiden flight is scheduled for 2024 at the time of writing.

The installation of heavy orbital structures requires heavy-lift launch vehicles. Right now, the highest-capacity vehicles for launching payloads into space are NASA’s Space Launch System (SLS) for the Artemis lunar programme, with a capacity to take 95 tonnes into low Earth orbit (LEO), and Space X’s Falcon Heavy, with 63.8 tonnes for its non-recoverable version. ESA’s planned launch vehicle, which is still only at the drawing board stage, is called PROTEIN (euroPean Reusable and cOsT hEavy lift transport InvestigatioN).

ESA has signed two contracts with ArianeGroup and Germany’s Rocket Factory Augsburg (RFA), each worth €500,000, to work on this issue. Based on the information currently available, the launcher should be recoverable and capable of delivering up to 100 tonnes into low orbit, so that it can be used to launch up to 10,000 tonnes a year. This project is part of ESA’s Future Launchers Preparatory Programme (FLPP), which has been allocated equally between the French National Centre for Space Studies (CNES), the Italian Space Agency (ASI) and its German counterpart, DLR.

There is a precedent for this. In the 1990s, CNES asked EADS/Astrium to work on a launcher concept with a capacity on a par with Saturn V during the Apollo years. This was the “Super Ariane 5” designed to send a manned vehicle to the Moon. It would have been able to deliver a payload of around 90 tonnes into low orbit. With PROTEIN, no human transport is planned, although this option is technically possible.

Preparing for intruders

There is also the issue of the security of future orbital data centers. By definition, an orbital system is isolated, and an attempt to board it cannot be ruled out. “Human intruders at an altitude of 1,500 km would not exactly be plain sailing, even though some countries might try it just to make trouble,” says Durand. He also points out that radio frequency hacking remains a possibility.

It is technically possible to take control of a satellite remotely, as Thales demonstrated at the third annual CYSAT conference on 26 and 27 April 2023, using ESA’s small technological demonstrator, the 7-kilogram (3U) OP-SAT nanosatellite. Thales’s offensive cybersecurity team rose to the challenge, identifying vulnerabilities that could disrupt satellite operations. To take control of the satellite, the four-strong red team managed to break into the satellite by exploiting several weaknesses.

The team compromised the data transmitted back to Earth by modifying the images captured by the satellite. However, ESA still had access to the satellite during the demonstration. Future data centers installed in orbit around the Earth will therefore have to be highly secure, not only from a maintenance perspective, but also for future customers, who have yet to be identified. If this major project becomes a reality, (all?) our data could one day find a home between the Earth and the Moon.