Skip to content
Eton Technology SolutionsEton Technology Solutions
NVIDIA Starts Qualifying Power and Cooling Kit, and That Changes How AI Sites Get Specified

NVIDIA Starts Qualifying Power and Cooling Kit, and That Changes How AI Sites Get Specified

September 30, 2026

NVIDIA Starts Qualifying Power and Cooling Kit, and That Changes How AI Sites Get Specified

NVIDIA has launched DSX Ready, a qualification programme for facility kit, starting with battery energy storage systems and cooling distribution units. The compute vendor is now shaping the facility bill of materials.

What was announced

  • NVIDIA announced DSX Ready on 21 September 2026, a qualification programme for partner products that meet applicable NVIDIA DSX AI factory reference design requirements.
  • The programme launches with two categories: battery energy storage systems and cooling distribution units, with further infrastructure and software categories to follow.
  • Qualified BESS solutions at launch come from Hitachi Energy, LG Energy Solution and Tesla. Qualified CDU solutions come from LG Electronics, LiquidStack and Vertiv.
  • BESS partners run required qualification tests and submit data for NVIDIA review within a defined qualification boundary. CDU partners use a self-qualification suite.
  • NVIDIA states that passing qualification does not replace site level engineering and does not imply site level stability.

The ETON view

A compute vendor issuing a qualified list for battery storage and cooling distribution units is a structural change, not a product launch. The effective specification for an AI hall now starts with the accelerator roadmap and works outwards into the facility, and the shortlist of suppliers at each layer gets shorter.

For anyone building or leasing UK capacity, that cuts both ways. A qualified list does reduce integration risk on liquid cooling, which is where most first-time GPU deployments lose time. It also narrows negotiating room, and NVIDIA's own wording is careful: qualification is a component level pass, not a guarantee that your hall will behave. Site engineering, and whoever owns it, still decides whether the cluster runs.

Most of our customers are not building a 100 MW factory. They are fitting GPU nodes into 6 to 12 kW colocation racks where the constraint is the rack, not the reference design. Our position there is unchanged: specify the compute against the power and cooling you actually have, keep the vendor list open across HPE, Dell, Supermicro, GIGABYTE, ASUS and Lenovo platforms, and let availability and lead time drive the build rather than a programme badge.

Related infrastructure

Category: AI & GPU · Vendor: NVIDIA, Vertiv, LiquidStack · Technology: DSX reference designs, CDUs, battery energy storage · Last verified: 22 Sep 2026 · ~2 min read

Sourcing this kind of infrastructure? Talk to ETON about availability, lead time and pricing.

Cart 0

Your cart is currently empty.

Start Shopping
Call Us