"Peak insanity": Analysis dissects plans for data centers in space

The rush to develop data centers in Earth orbit has reached a "peak insanity" and carries the risk that the participating actors will neglect the urgently needed expansion of infrastructure on Earth. This is the conclusion of a Gartner analysis of plans for "orbital data centers," which have recently been promoted not only by Elon Musk. Such structures "Orbital data centers are not economically viable, cannot match the performance of terrestrial data centers and fall short compared to grid-connected, liquid-cooled, nuclear-powered campuses," it states in unambiguous terms. Those who work on it are not only wasting money, but also time.

Several high obstacles

The idea of sending data from Earth to space to be processed there and sent back to Earth is "not feasible" because the transmission rates represent a bottleneck, according to the analysis, which is available to heise online. The high costs of launching payloads into space are cited as another obstacle. Finally, author Bill Ray feels compelled to remind us that there is a vacuum in space. There is no air to dissipate heat. Instead, one relies on inefficient and slow alternatives, for example, using ammonia. For the maintenance of such complex infrastructure, engineers would have to be sent into space instead of astronauts: "A capability that does not currently exist."

While orbital data centers are unrealistic for these reasons, there is a risk that better alternatives and essential investments will be neglected in the work on them, it continues. The author refers, for example, to data centers that are submerged in water. Microsoft, for instance, has developed such systems. Data centers can also be built in remote locations such as the Arctic, Iceland, or the deserts of Saudi Arabia without overly cumbersome approval processes. However, if the "bubble" around space data centers lasts too long, there is a real danger that terrestrial infrastructure will not be expanded sufficiently.

That the plans for orbital data centers could develop into a new space race has been apparent for some time. Projects exist at Google, among others, and Amazon, but Elon Musk is once again the loudest. He claims that Earth orbit will be the cheapest location for AI data centers in just a few years. Gartner sees it differently. Nevertheless, the analyst does not want to rule out being wrong. Those who are worried about missing out should keep an eye on the costs of transporting payloads into space. But only a significant price drop could indicate that he was not entirely right. The remaining problems would then still exist. Furthermore, It is possible that politics will drive such projects forward, "regardless of viability."

Innovations possible

Even if orbital data centers cannot be used efficiently for processing terrestrial data, one should keep an eye on whether the work on them promotes innovations, the Gartner analyst writes. He refers to technology for communication between satellites. Work on cooling systems could also lead to inventions. Finally, he expects that innovative methods will be developed in the course of the work to distribute computing tasks across multiple satellites. In space itself, however, for the foreseeable future, only computing tasks on data that is also collected there will be processed – for example, by Earth observation satellites.

(mho)