SpaceX claims it will put a Vera Rubin NVL72 rack-scale system into orbit next year
Elon Musk has said SpaceX will launch a "space-optimised" Nvidia Vera Rubin NVL72 rack-scale system into orbit in the fourth quarter of next year, with plans to scale up significantly by 2028. The announcement, made on X and echoed in a slightly differently worded Nvidia press release, marks an ambitious step towards putting datacentre hardware in space, though neither company has explained what "space-optimised" actually involves, and SpaceX has yet to build such a satellite.
A full NVL72 system contains 72 Rubin GPUs, 36 Vera CPUs, ConnectX-9 SuperNICs and BlueField-4 DPUs across 18 compute trays and nine NVLink switch trays, weighing around 1,800kg per rack and drawing 120kW, which Nvidia says will use novel liquid cooling. SpaceX has outlined a satellite design featuring a 210kW solar array and liquid radiators to dissipate heat in orbit, and its Falcon 9 rocket—capable of lifting 22,800kg to low Earth orbit and flying over 100 missions a year—could plausibly accommodate several such racks per launch. However, the piece notes Musk's patchy record on public predictions, citing a New York Times analysis finding he fulfilled only 19% of 602 public promises, with 35% delivered late or not at all.
- SpaceX plans to launch an Nvidia Vera Rubin NVL72 datacentre rack into orbit in 2027.
- Larger-scale deployment is promised for 2028, per Musk and Nvidia statements.
- Musk has a mixed track record on delivering public promises on time.
New here? Start with this
Nvidia designs computer chips and systems, including the Vera Rubin NVL72, a large rack of processors used for demanding computing tasks. SpaceX, founded by Elon Musk, builds rockets and spacecraft; its Falcon 9 rocket regularly carries satellites into low Earth orbit, the region relatively close to Earth where many satellites operate.
Putting powerful computers in orbit could let companies run some computing tasks using equipment powered by sunlight, but keeping it cool and maintaining it remotely would pose challenges. The proposal brings together Nvidia’s computing hardware and SpaceX’s launch and satellite plans, though the companies have not explained what changes would make the system suitable for space.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
SpaceX has demonstrated genuine capability executing complex aerospace projects including Falcon 9 reusability, and Nvidia has already developed the required hardware. Space-based computing addresses real market demand, providing strong commercial incentive for execution. Even accounting for Musk's characteristic timeline optimism, a 2026 deployment from a well-resourced aerospace company remains a credible objective.
The case against
Musk's track record shows only 19 per cent of public promises fulfilled on time, with 35 per cent delivered late or abandoned. Space systems require extensive testing and certification that characteristically extend timelines substantially beyond announcements. Without demonstrated proof of concept for space-optimising this hardware, the announcement appears driven by promotional optimism rather than grounded engineering assessment.