Image from Wikimedia Commons
The artificial intelligence race is about to move beyond the boundaries of Earth. SpaceXAI will use Nvidia’s new Vera Rubin platform to expand the infrastructure powering its AI systems and intends to take the same architecture into space through Starmind, its orbital computing project.
The first generation of the Starmind satellite is expected to use an optimised version of the Nvidia Vera Rubin NVL72 system, technology developed to support intensive artificial intelligence workloads. In practice, the partnership aims to bring into orbit some of the computing capacity that currently relies on enormous terrestrial data centres.
The initiative is still under development, but it signals a new frontier for an industry already worth hundreds of billions of dollars: after the race for chips and data centres, AI infrastructure is beginning to look towards space.
Training and operating advanced artificial intelligence systems requires ever-increasing computing capacity. That means more chips, more data centres, more energy and cooling systems capable of keeping all that infrastructure running.
This is precisely where space begins to enter the conversation.
Orbital computing could allow large volumes of data captured by satellites to be processed in space itself, rather than being transmitted in full back to Earth. This could reduce the amount of data that needs to be transmitted, lower latency in certain applications and open up new possibilities for geospatial intelligence, autonomous operations and real-time data analysis.
But the challenges are enormous. Energy, temperature, radiation, maintenance, connectivity and reliability make an orbital data centre far more complex than simply placing servers inside a satellite.
SpaceX and Nvidia believe this barrier can begin to be overcome.
Over the past few years, the artificial intelligence race has created intense competition for one essential resource: computing capacity.
Nvidia has become one of the world’s most valuable companies by supplying the technology that powers this infrastructure. SpaceX, meanwhile, has built an unprecedented ability to put equipment into orbit, alongside a global satellite network through Starlink.
The collaboration between the two companies demonstrates how different industries are beginning to converge around AI.
Technology, energy, telecommunications and the space industry can no longer be viewed in isolation.
There is also an interesting business logic at play. The greatest opportunities created by artificial intelligence may not lie solely in the models that reach consumers, but in the infrastructure required to keep them running.
First, the competition was for chips. Then came data centres and energy. Now, a new frontier is beginning to emerge above our heads.
It is still too early to know whether large-scale orbital computing infrastructure will be able to compete economically with terrestrial data centres. But Starmind reveals just how far companies are willing to go in search of the next advantage in the artificial intelligence race.
If the next major constraint on artificial intelligence is the infrastructure required to sustain it, could space become the next great frontier in this race?
