Vertiv just dropped its first digital twin onto Nvidia’s Omniverse DSX platform, and it’s aimed squarely at one of the most expensive problems in tech right now: building AI factories without spending years and billions on guesswork.
The product, called the Vertiv OneCore Rubin DSX, is a simulation-ready model that converges power, cooling, controls, and lifecycle services into a single system-level digital twin. It lets engineers design, test, and optimize an entire AI data center’s physical infrastructure inside a virtual environment before anything gets bolted to a concrete floor.
What Vertiv actually built
The OneCore Rubin DSX is Vertiv’s contribution to Nvidia’s Vera Rubin DSX AI factory reference design and the broader Omniverse DSX Blueprint. It functions as a set of modular, repeatable building blocks that can be stress-tested in simulation rather than discovered to be wrong after construction.
Vertiv isn’t the only company contributing to the Nvidia DSX ecosystem. Partners in the broader effort include Cadence, Dassault Systèmes, and Siemens, each bringing their own simulation and design capabilities to the platform.
The target metric that keeps coming up is a reduction in time-to-first-token of up to 50%. That’s Nvidia’s way of measuring how quickly an AI factory goes from design to actually producing useful compute output.
The backstory and why digital twins matter here
This isn’t Vertiv’s first move on the Omniverse DSX platform. Back on October 28, 2025, the company announced gigawatt-scale reference architectures for the Omniverse DSX Blueprint. That earlier work focused on flexible deployment methods, offering stick-built, hybrid, and fully prefabricated options for AI data center construction.
The March 2026 update builds directly on that foundation. Where the earlier announcement was about establishing reference architectures at massive scale, this one is about making those architectures simulatable before committing real capital.
What this means for investors
Vertiv, which trades on the NYSE under the ticker VRT, has been positioning itself as the picks-and-shovels play for the AI infrastructure buildout. The company isn’t selling AI models or compute. It’s selling the physical infrastructure that makes AI compute possible: power distribution, thermal management, facility controls.
By contributing a converged infrastructure model directly to Nvidia’s reference architecture, Vertiv is essentially getting its products specified into the default template that companies will use when building Nvidia-based AI factories.
The partnership roster tells you where this is headed. Nvidia, Cadence, Dassault Systèmes, Siemens, and Vertiv are collectively building a design-to-deployment pipeline that treats entire AI factories as simulatable systems.