Solar data centers in orbit? Musk wants to move AI into space

Elon Musk has come up with one of the boldest technological visions of recent years: building a network of solar-powered data centers in Earth’s orbit. He says the project would address the growing energy demands of artificial intelligence and bypass the limits of terrestrial power grids. Experts, however, warn of enormous technical and economic obstacles.

The rapid growth of artificial intelligence is bringing not only technological progress but also a major problem in the form of high energy consumption. Modern data centers that run large language models and other demanding algorithms are among the biggest electricity consumers. In some regions, their operation is already putting a strain on distribution grids and raising questions about where the energy for further expansion will come from.

Billionaire and SpaceX and xAI founder Elon Musk wants to tackle this problem with his vision of orbital data centers. He outlined his plan in greater detail in a recent interview on the Dwarkesh podcast. The plan involves deploying a vast network of satellites equipped with computing hardware and large solar panels to process artificial intelligence workloads directly in space.

Musk argues that “it’s always sunny in space” — meaning that solar panels in orbit are not limited by the alternation of day and night or by cloud cover. Orbital infrastructure could therefore draw on an almost continuous supply of energy. 

Solar power almost nonstop and unaffected by the weather

The project’s energy rationale is simple: move computing power to where energy is most readily available. In low Earth orbit, satellites receive a much more stable supply of solar radiation than ground-based solar power plants.

In low Earth orbit, satellites are exposed to solar radiation for approximately 60–70% of each orbit, with one orbit around Earth taking about 90 minutes. In higher orbits, or with a suitably designed constellation, the proportion of time in sunlight can be even greater. By contrast, ground-based solar power plants typically have an annual capacity factor of just 10–20% in Central Europe and 20–30% in the best desert areas. The problems of nighttime outages and seasonal fluctuations are eliminated.

Another difference is the absence of atmospheric attenuation. The atmosphere absorbs and scatters approximately 25–30% of incoming solar energy, meaning that photovoltaic panels in space have access to a higher and more stable flow of energy, unaffected by cloud cover, dust or the seasons.

A radical paradigm shift, but significant obstacles too

Musk says the satellites would function as a distributed “cloud” network. Individual nodes would be connected by optical (laser) links, and the results of computations would be transmitted back to Earth. Musk has even suggested that, in the long term, space could become the cheapest place to run powerful AI systems.

From an energy perspective, this would represent a radical paradigm shift: instead of building ever-larger terrestrial data centers connected to overloaded grids, an extraterrestrial computing infrastructure powered entirely by solar energy would be created. But enthusiasm is tempered by the realities of the space industry. Publications such as Forbes, Euronews and TechCrunch point out that the project faces major obstacles.

The first is economics. Launching thousands or even hundreds of thousands of satellites carrying computing hardware would entail enormous costs, even with SpaceX’s reusable rockets. Every kilogram of payload in orbit is expensive, and specialized AI hardware is not lightweight.

The second problem is the space environment itself. Electronics in orbit are exposed to cosmic radiation, extreme temperature fluctuations and the risk of micrometeoroid impacts. While a faulty server can be quickly replaced in a terrestrial data center, servicing a satellite is many times more complex and costly.

Cooling is another specific challenge. Although people often point to the “cold of space,” heat in a vacuum cannot be carried away by airflow, but only through radiation. This requires large radiators and precise thermal management, increasing the system’s weight and complexity.

Data centers in orbit as a geopolitical issue

From an energy perspective, the orbital approach may be appealing, but from a telecommunications perspective, latency is a concern. Data transmission between Earth and a satellite, even in low Earth orbit, is slower than communication over a terrestrial fiber-optic network. For some applications — such as real-time control of industrial processes or financial trading — delays could be critical.

Analysts quoted by international media therefore warn that even if orbital data centers are built, they would probably complement, rather than replace, terrestrial infrastructure.

Geopolitics and regulation add another dimension. Euronews warns that if the project were carried out primarily under the SpaceX banner, Musk could gain unprecedented control over key infrastructure for AI and cloud services. This raises questions about competition, security and individual countries’ dependence on a private player.

At the same time, it is clear that other major powers are also beginning to take an interest in space-based computing capacity. If the concept proves viable, it could usher in a new phase of technological competition — this time over solar energy and computing power in orbit.