Verified, then logged.
Sun 22 Nov 2026 · 00:16 KST / 15:16 UTC한국어 | ENSearch /

Breaking02 / IT

NVIDIA pitches DSX POD as faster path to AI factory capacity

The pitch is aimed at operators that must run compute, cooling and other systems as one unit.

NVIDIA introduced DSX POD in a developer blog post on 21 Oct 2026, saying it lets AI factory operators secure capacity faster and more predictably.

According to the blog, several systems must be running before an AI workload produces its first token, the unit of model output. They are compute, networking, storage, power, cooling and software. Operators have to bring them online together as one productive unit.

NVIDIA opened the post with the line "AI factories turn energy into intelligence". The sentence defines that conversion as the role of an AI factory.

A conventional data centre build starts with site selection, NVIDIA said. Power and cooling are then designed for the deployment. Facility layout follows as the next step.