Traditional ground-based data centers are gradually becoming a thing of the past—leading tech companies are increasingly exploring the possibility of moving computing power into orbit. While this idea may sound futuristic, it is already attracting significant investment and market attention. Let’s take a closer look at what’s driving this trend and how realistic it actually is.

How Modern Data Centers Work

A data center is essentially a massive warehouse filled with thousands of servers running around the clock. Major companies working in artificial intelligence, such as Anthropic, OpenAI, and Google, rely on these centers for two main purposes:

• Training AI Models: Training advanced language models like ChatGPT or Claude requires running computations on thousands of specialized chips (GPUs) simultaneously, sometimes for weeks or even months at a time.
• Running AI Services: When millions of users interact with chatbots, their requests are processed in data centers, where servers generate responses in real time.

Data centers ensure the seamless operation of thousands of machines and provide the necessary infrastructure for uninterrupted 24/7 service. To achieve this, they are connected to ultra-fast internet and equipped with powerful cooling systems, which demand enormous amounts of energy and put a strain on local power grids and resources.

Why Consider Data Centers in Space?

The concept of moving computing power into orbit is being discussed as a more efficient alternative to ground-based centers. In low Earth orbit, it’s possible to harness virtually unlimited solar energy and use radiative cooling, eliminating constraints related to land, energy, and fresh water availability.

The plan is to deploy powerful GPU servers inside satellites, which will travel in groups along sun-synchronous orbits, transmitting data to one another as they circle the Earth. This approach would allow operators to scale computing resources more easily.

Who Is Already Working on Space Data Centers?

Both established and emerging companies are actively pursuing this direction:

• Google, in partnership with Planet, is developing the Suncatcher project and plans to launch prototype satellites as early as next year.
• Aetherflux, originally focused on transmitting solar energy from orbit to Earth, is now working on a commercial orbital data center hub.
• Starcloud, with support from Nvidia, has already sent a GPU payload into space and trained a large language model there.
• SpaceX and Elon Musk’s company xAI are planning to join forces to launch orbital data centers, aiming to make AI computations in space cheaper than on Earth within the next three years.

Challenges and Obstacles

However, there are many hurdles to overcome before this vision can become reality. As The Verge notes, orbital data centers will need to:

• Safely navigate more than 6,600 tons of space debris and over 14,000 active satellites.
• Address issues of maneuvering and refueling in orbit.
• Efficiently dissipate heat and periodically service equipment, potentially with the help of astronauts.
• Consider the impact of new satellites on astronomical research and light pollution.

Outlook and Reflections

The first attempts to create computing power in space will provide valuable experience and knowledge before such projects can be scaled to national or corporate levels. Despite significant challenges, this could eventually become a reality.

Still, it’s worth considering: will artificial intelligence truly deliver the benefits its creators promise to humanity, and do we really need to keep expanding AI infrastructure—whether on Earth or far beyond its boundaries?