NVIDIA unveils Vera Rubin NVL72 compute tray with one-minute automated assembly
NVIDIA introduced the Vera Rubin NVL72 Compute Tray, a single-wide rack-scale unit that the company says assembles in one minute with full automation and no external cables, hoses, or fans. The tray integrates liquid cooling and connectivity via third-generation MGX racks and backplane blind-mating, aiming to speed cluster ramp and time to revenue. Reporting ties the design to SpaceX plans to quintuple compute by 2027 on Vera Rubin, potentially requiring over two million GPUs, and to an optimized NVL72 payload on the Starmind AI1 satellite that Elon Musk said could launch next year. Operators should note marketed assembly times are vendor claims, not independent benchmarks, while the broader Vera Rubin supply and deployment timeline remains unsettled.
NVIDIA unveils Vera Rubin NVL72 compute tray with one-minute automated assembly
NVIDIA introduced the Vera Rubin NVL72 Compute Tray as a single-wide unit designed to scale, assembled in one minute with 100% automation. The design has no external cables, hoses, or fans, aimed at faster cluster ramp and time to revenue.
Key takeaway
NVIDIA is packaging Vera Rubin NVL72 as an integrated, cable-free compute tray built for automated rack-scale deployment rather than discrete components.
What happened
NVIDIA announced the Vera Rubin NVL72 Compute Tray as a single-wide unit designed to scale, with the company claiming one-minute assembly, full automation, and no external cables, hoses, or fans.
Related reporting cites third-generation MGX racks with 100% liquid cooling at 45°C via backplane blind-mating, and links the platform to SpaceX plans that could require over two million Rubin GPUs by 2027 and an optimized NVL72 on Starmind AI1 next year.
Evidence
NVIDIA says the Vera Rubin NVL72 Compute Tray assembles in one minute with full automation and no external cables, hoses, or fans.
NVIDIA (X) · attributed
NVIDIA introduced the Vera Rubin NVL72 Compute Tray as a single-wide unit designed to scale, assembled in one minute with 100% automation.
The tray design uses third-generation MGX racks with liquid cooling at 45°C via backplane blind-mating.
NVIDIA (X) · attributed
This design leverages third-generation MGX racks to achieve 100% liquid cooling at 45°C via backplane blind-mating
SpaceX plans to more than quintuple compute capacity by 2027, betting exclusively on Nvidia Vera Rubin and potentially requiring over two million GPUs.
The Decoder · attributed
SpaceX plans to more than 5x its compute capacity by the end of 2027, betting exclusively on Nvidia's Vera Rubin platform.
Elon Musk said SpaceX will exclusively use Nvidia GPUs and that an optimized Vera Rubin NVL72 could launch into space next year.
Tom's Hardware AI · attributed
Elon Musk says SpaceX will exclusively use Nvidia GPUs 'because they are the best' — says optimized Vera Rubin NVL72 will be launched into space next year
NVIDIA says SpaceX Starmind AI1 satellite compute payload will be powered by Vera Rubin NVL72.
NVIDIA (X) · attributed
SpaceX's Starmind AI1 satellite compute payload is powered by NVIDIA Vera Rubin NVL72
Why it matters
SpaceX's exclusive Vera Rubin bet and orbital Starmind AI1 plans signal that hyperscale buyers are anchoring procurement on NVIDIA's rack-scale roadmap.
Limits and uncertainties
One-minute tray assembly is a vendor claim; the packet provides no independent deployment benchmarks.
Orbital NVL72 timing is attributed to Elon Musk, not a confirmed launch schedule.
Projections of over two million Rubin GPUs for SpaceX are reporting estimates, not firm orders.
Practical implications
Data center operators evaluating Vera Rubin must plan around MGX rack integration and liquid-cooled blind-mate trays.
Procurement teams should factor large exclusive Vera Rubin commitments into supply and lead-time planning.
What to watch
Independent benchmarks for NVL72 tray assembly, cooling performance, and cluster ramp timelines.
SpaceX Vera Rubin GPU procurement scale and Starmind AI1 orbital deployment milestones.
Original reporting: Vera Rubin NVL72 Compute Tray. Designed to scale. Assembled in 1 minute. 100% automated. No cables, no hoses, no fans. A revolutionary new design that means faster ramp and faster…
