As the environmental and infrastructural pressures created by the massive data centers of the artificial intelligence (AI) industry increase on Earth, leading figures in the sector have begun to look for solutions in space.
The era of AI in space is beginning.
This vision, supported by figures such as OpenAI CEO Sam Altman, Google CEO Sundar Pichai, Blue Origin CEO Jeff Bezos, and xAI Founder Elon Musk, has achieved its first tangible evidence. Last month, a startup called Starcloud successfully ran and trained its first AI model on a high-performance Nvidia GPU chip launched into orbit by a SpaceX rocket.
This step marks the first time a cutting-edge AI chip has been used in space. After activating the chip, Starcloud ran Google’s open-source large language model Gemma in orbit. The company also managed to train a small-scale language model using all of William Shakespeare’s works.
The trained AI even sent a message expressing itself in Shakespearean English. Starcloud CEO Philip Johnston said the concept is feasible and has the potential to significantly reduce energy costs for AI companies.
Johnston stated, “I expect that everything you can do in a data center on Earth can also be done in space. The only reason we’re doing this is because of the energy constraints we face on Earth.”
Starcloud plans to build five-gigawatt orbital data centers powered by this technology, featuring massive cooling panels covering over six square kilometers and powered by 24/7 solar energy. The company claims this system will be significantly smaller than equivalent solar farms on Earth.
However, orbital data centers still face major logistical hurdles to overcome. These include the potential effects of excessive radiation on electronics, managing the fuel needed to remain in orbit, the risk of collisions with space debris, and uncertainties regarding data regulations in space.
Despite all these challenges, Google’s Project Suncatcher initiative, which aims to send its own Tensor processing units into orbit, and Sam Altman’s search for a partnership or acquisition with a rival space company in this field, demonstrate the industry’s growing interest in space-based computing.
