Structured cabling for AI: lessons from the NVIDIA and Corning guide
Archive focus: January 2025 · Prepared 17 October 2026 · XConnect FZE
Retrospective: This article was prepared on 17 October 2026 for our January 2025 archive. It reviews the announcement and its infrastructure planning implications; product plans described at the time may have changed.
The development
On 9 January 2025, Corning described its work with NVIDIA on an 800G InfiniBand and Ethernet connectivity guide. The guidance addresses structured connections from servers to leaf switches and onward through spine and core layers. It presents organised cabling as part of scaling an AI cluster, alongside the equipment and network design.
Practical implications
The following is XConnect’s planning interpretation.
The useful deliverable is a repeatable connection plan. In a large cluster, a spreadsheet of component quantities is insufficient unless each assembly can be traced to the correct ports. Infrastructure teams should agree how the first rack or pod will be checked, then repeat the validated arrangement with controlled exceptions.
What to review
- Translate the network topology into a port-to-port cabling schedule.
- Validate one representative rack or pod before ordering the full rollout.
- Document spare assemblies, replacement access and the process for updating records.
Explore our network solutions and data centre solutions, or contact XConnect to discuss your site requirements.
Source: Original announcement — corning.com. Reviewed 17 October 2026.