Research · October 29, 2026

Vertiv × NVIDIA: Building AI Factories Like Factories

On October 21, 2026, at NVIDIA GTC Berlin, Vertiv Holdings Co. (NYSE: VRT) and NVIDIA introduced an AI factory approach that combines rapid configuration and virtual validation of pre-engineered reference architectures with factory-built deployment. Polar Data Centers is applying the factory-built deployment approach to the first 41 MW phase of its multiphase AI data center at Herøya, Norway, a campus scalable to 100 MW.

The problem it attacks is time, not just hardware. For AI factory operators, securing compute is only part of the challenge: traditional infrastructure development can require power, cooling and facility systems to be engineered and integrated through separate, sequential processes, creating complexity and potential delays. The Vertiv and NVIDIA approach starts instead with pre-engineered AI factory reference architectures that can be rapidly configured to customer and site requirements, virtually validated before deployment, and then translated into factory-built infrastructure.

Configure and validate. The design stage pairs the NVIDIA Vera Rubin DSX AI Factory reference design and the NVIDIA Omniverse DSX Blueprint with Vertiv SimReady power and cooling assets, integrated with Dassault Systèmes' 3DEXPERIENCE agentic platform. In the modeled scenario, that cuts the time from requirements to a configured and validated solution from approximately 30 days to three days, a 10x improvement. Vertiv is also contributing SimReady OpenUSD assets of its power and cooling equipment to the SimReady Central Library on Hugging Face, so engineering teams can discover and reuse validated equipment assets in AI factory digital twins.

Deploy. Once a configuration is validated, Vertiv prefabricated solutions shift more assembly, integration and testing into the factory before equipment reaches the site. The named systems are Vertiv OneCore, a hybrid-built data center solution integrating critical power, thermal management and supporting infrastructure, and Vertiv SmartRun, a prefabricated overhead infrastructure system for high-density power distribution, liquid cooling, networking and containment. In the modeled scenario, the factory-built approach reduces the infrastructure deployment period from approximately 90 days to 45 days, a reduction of up to 50%. Combined with the faster configuration stage, the modeled approach reduces these stages from approximately 120 days to 48 days, making productive AI capacity available about 10 weeks sooner — and, at the same installed capacity, generating up to 30% more cumulative token output within 12 months of project kickoff. The model's arithmetic is 317 versus 245 productive days in the first 12 months.

What this is — and what it is not. Every timeline and output figure in this article is modeled, not measured. The release states its figures are under defined assumptions and that actual results will vary by project; the 30% is a projection, not a deployment result and not a guarantee. No Vertiv–NVIDIA contract value was disclosed, and no Vertiv revenue figure is attached to Polar's phase in this article. What is real is the approach and its first application: pre-engineered designs, virtually validated, then built in a factory — with Polar's Herøya phase as the reference deployment, supplied with prefabricated power and thermal modules, SmartRun overhead infrastructure and SmartMod, supported by Unify software.

"An AI factory earns its return in tokens, and every day spent configuring infrastructure or integrating it on site is output that never comes back," said Giordano Albertazzi, CEO of Vertiv. Polar CEO Andy Hayes put the buyer's version of the same point: "Our customers measure us on how quickly capacity is ready for them." For BlitzLab readers, the through-line is the AI buildout's moving bottleneck. The constraint has shifted from chips to everything around them — sites, power, cooling and now deployment speed. Vertiv sells the picks and shovels of that last leg: if AI factories are going to be built at the pace the chip roadmaps imply, somebody has to industrialize putting the buildings together. This announcement is a claim that the building itself can be manufactured — with the numbers, for now, still on paper.

Sources