GPU Rack Leasing (GB300 B300 and Vera Rubin)

Bare-metal GPU servers vs cloud GPU instances at rack scale

bare metal gpu vs cloud gpu instances
How we researched this article
Written by Anton Liskevich
Reviewed by Dmitry Ogievich, COO
Why we wrote it Buyers comparing bare-metal GPU servers with cloud GPU instances need to know what changes at rack scale before they choose a procurement model.
How we researched it We read the top 3 Google results, the Google AI Overview and ChatGPT answers to the buyer questions (November 2026), and NVIDIA's own specification pages; the details come from CambridgeNexus operating documents.
How AI was used We drafted this article with AI, using CambridgeNexus documents and the sources below. People checked every fact, and the reviewer approved the final text.
Sources NVIDIA GB300 NVL72, NVIDIA NVLink and NVLink Switch, NVIDIA GPU Operator: time-slicing GPUs. Full list in References.
Facts checked against Official CambridgeNexus internal documentation
Research date 17 Nov 2026
Last updated 18 Nov 2026

GPU Rack Leasing (GB300 B300 and Vera Rubin)

Whole NVIDIA racks, one operator

GB300 NVL72 racks, B300 systems and Vera Rubin allocation. Installed in 75 to 90 days from contract.

Check rack availability

Tags GPU Rack Leasing (GB300 B300 and Vera Rubin)