NVIDIA has unveiled a new suite of space-focused computing platforms centred on the Space-1 Vera Rubin Module, as it looks to enable more orbital data centres.
Announced during the company’s presentation at its annual GPU Technology Conference, known as GTC, the Space-1 Vera Rubin Module is designed for orbital data centres, geospatial intelligence and autonomous space operations, with the Rubin GPU on the module claimed to deliver up to 25 times more AI compute for space-based inferencing than the NVIDIA H100 GPU.
There has been growing interest in orbital data centres over the past few years, with many seeing it as a golden opportunity to build capacity without having to deal with some of the large constraints that are holding back data centres on earth. That’s because they don’t require the lengthy planning process associated with land-based data centres, while they also benefit from an abundance of renewable energy – which can be generated 24/7 from on-board solar panels.
Orbital data centres are not there to solely replace land-based data centres, however. In fact, one of the key benefits of having an AI data centre in space is simply to allow it to process the data that is already being generated in space – from other satellites, sensors and spacecraft. The practical upside is faster decisions, lower reliance on scarce downlink bandwidth, and the ability to send back only the most useful results rather than huge volumes of raw data.
There is one major downside to orbital data centres, however. While you might not have to find a huge plot of land, an orbital data centre isn’t going to have the same computing power as a facility located on planet Earth. That’s because they’re inherently constrained by, rather ironically, space. Not only is there a limit in the size of the data centre you can send up into space, but there’s also a strict limit on weight – with SpaceX’s Falcon 9, a common launch platform for sending satellites into orbit, limited to a maximum payload of 22,800kg.
It’s with that size constraint in mind that has seen NVIDIA announce its suite of space-focused computing platforms. It gives operators a full compute stack from orbit to ground, while its Space-1 Vera Rubin is designed to bring data-centre-class AI into orbit, with 25x more AI compute for space-based inferencing than the H100 chip. It’s also important to note that this is a chipset that has been specifically engineered for size-, weight- and power-constrained environments, so it should make building an orbital data centre for AI significantly easier and more cost-effective.
The launch of NVIDIA’s space-based suite comes at the right time for the market too. There have already been numerous firms clamouring to launch orbital data centres, with Axiom Space having previously announced its intentions to launch proof-of-concept data centres into space. It finally achieved that goal earlier this year, but it certainly won’t be the last.
It should also be noted that Axiom was name dropped during NVIDIA’s GTC presentation, with the company said to already be using NVIDIA’s accelerated computing platforms for next-generation missions. Aetherflux, Kepler Communications, Planet, Sophia Space and Starcloud are amongst the other companies already using NVIDIA’s platforms.
