Digital twin technology is not just for data center design. It can also be used to optimize management and operations of AI infrastructure.
Digital twin (DT) technology is not an entirely new concept in the data center industry, but its use has traditionally been limited to basic design validation and equipment mapping. For example, Building Information Modeling (BIM) can be used to create simple DTs to visualize planned infrastructure deployments. Also, computational fluid dynamics (CFD) has been used to model thermal behaviors and optimize cooling accordingly. These use cases, while helpful, have generally been tactical rather than strategic.
From design validation to operational intelligence
That’s changing. As highlighted in the recent Vertiv Frontiers report, the evolution of DT technology is one of the trends expected to reshape data center strategy in the near term. New and emerging DT platforms are enabling more informed, dynamic facility design and increasingly have the potential to support ongoing operational management.
Operational modeling is becoming especially critical for modern AI facilities supporting high-density racks and variable AI workloads. AI is associated with power loads that can pulse from a 10% idle to a 150% overload in a flash. A DT platform capable of ingesting real-time equipment data and modeling this level of AI load fluctuation becomes an important tool for managing underlying infrastructure.
NVIDIA Omniverse and the real-time digital twin
Specific tools have already been developed to support this approach. For example, Vertiv is collaborating with NVIDIA on a digital simulation platform designed specifically for AI factories. NVIDIA Omniverse is a suite of libraries and microservices for building physically accurate, real-time simulations of complex systems including data center DTs. Leveraging NVIDIA’s expertise in accelerated computing and AI, Omniverse enables high‑fidelity visualization, shorter development cycles, and greater precision. Users can take advantage of Omniverse’s interoperability, physically based rendering, and real‑time collaboration features to build sophisticated virtual facility models.
With NVIDIA Omniverse, operators can use a DT not merely as a static representation of design intent but as a dynamic virtual proving ground. The model can be continually updated with real operational data, simulate complex failure or stress scenarios, and predict how systems will behave under changing loads. Operators can stress-test power and cooling systems, validate real-world performance against nameplate ratings, and observe how isolated component failures ripple across interconnected systems.
Several pilot projects are already underway, including work at data center operator Digital Realty’s research center in Virginia and the Stargate Project in Texas. The expectation is that mature DTs will enable operators to more seamlessly shift workloads, direct power where it is needed most, adjust thermal settings dynamically, and ultimately model and manage interactions with the electrical grid.
The case for operational digital twins
The potential benefits, even from seemingly modest efficiency gains, are significant. A typical 100MW data center may spend up to $100 million annually on electricity, depending on location. Every 1% improvement in energy efficiency translates into roughly $1 million in savings1. A fully leveraged DT can act as a powerful optimization engine. And that is only part of the value: predictive maintenance, reduced downtime risk, and improved capacity planning offer additional return on investment.
To be clear, the industry is still in the early stages of this type of DT-driven operational modeling. The current pilots are expected to generate meaningful data in 2026, and they may drive more substantial investment later in the year and beyond. Realizing the full potential of operational DTs will take time but momentum is rapidly building.
1Based on 100 MW × 8,760 hrs/yr × $0.1275/kWh (U.S. average commercial electricity rate, 2024). Source: U.S. Energy Information Administration (EIA), Electric Power Annual 2024, Table 2.10 — Average Price of Electricity to Ultimate Customers by End-Use Sector (eia.gov, released October 2025). Costs vary significantly by location; rates in low-cost states may reduce this figure.
Digital twin-driven design and operations is one of the technology trends reshaping data center infrastructure identified in Vertiv Frontiers.