NVIDIA Launches DSX Ready Program to Certify Power and Cooling Gear for AI Factories

NVIDIA introduced a program called DSX Ready on Tuesday, a qualification effort meant to certify power and cooling equipment that pairs with its designs for AI data centers. The company framed the move as a response to a practical bottleneck: as AI clusters grow, the electricity, heat, water and grid connections available at a given site increasingly determine how much computing a builder can actually turn on. Products that clear the program will be listed as compatible with what NVIDIA calls the complete factory design.

The program arrives as AI infrastructure spending keeps climbing and utilities struggle to keep pace. Operators building large GPU clusters now routinely negotiate multi-hundred-megawatt power contracts years ahead of construction, and cooling choices range from direct-to-chip liquid loops to immersion tanks, each with different water and retrofit demands. NVIDIA's pitch is that a validated catalog shortens the engineering work of matching those systems to its rack-scale architectures, rather than leaving buyers to reconcile vendor specs on their own.

The strategic logic cuts two ways. For NVIDIA, tighter control over the surrounding infrastructure makes its platforms easier to deploy and harder to swap out, since a certified power and cooling stack becomes part of the purchase decision. For equipment makers, inclusion in the DSX Ready list functions as a sales credential in a market where buyers are moving fast and have little appetite for unproven hardware. Neither NVIDIA nor the blog post disclosed which vendors have signed on or when the first qualified products will appear.

The announcement also nods to constraints that go beyond the data hall. Siting, water rights and grid interconnection queues have delayed projects in regions such as Northern Virginia, Arizona and parts of Europe, and some operators have turned to on-site generation to bridge the gap. By tying qualification to the full factory design, NVIDIA is signaling that the limiting factor for AI capacity is no longer only chips, but the physical plant around them.

What to watch next is the list itself. The program's value depends on how many power distribution, cooling and water-handling vendors actually earn the mark, how testing is verified, and whether hyperscalers and neoclouds treat the certification as a requirement or a suggestion. NVIDIA has not said when it will publish qualified products or how often the list will be updated, leaving the program's real weight undetermined for now.