Schneider Electric has announced a new set of tools and reference designs developed with NVIDIA and AVEVA to support the design and operation of large-scale AI data centres.

Unveiled at NVIDIA GTC 2026 in San Jose, the announcement centres on three developments aimed at the growing infrastructure demands of AI factories.

The first is a new validated reference design for NVIDIA Vera Rubin NVL72 racks. Schneider Electric said the design covers power and cooling requirements for NVIDIA’s latest rack-scale systems, including support for 480 VAC power distribution, a 45°C technology cooling system loop supply temperature, and an IT room layout designed around clustered AI racks.

The company said the design also supports different GPU operating points, including MaxP and MaxQ, with the latter intended to improve tokens per watt where power efficiency is a priority. The reference design has been validated using ETAP models for electrical system design and ITD CFD models for layout and airflow.

The second part of the announcement focuses on digital twins. AVEVA and NVIDIA have developed a new lifecycle digital twin architecture using NVIDIA Omniverse, with Schneider Electric creating SimReady assets to help model AI facilities before deployment.

That will allow operators to simulate areas such as power distribution, thermal dynamics, airflow and controls before construction begins. As AI facilities continue to scale, that kind of pre-deployment testing is likely to become increasingly important in reducing design risk and speeding up delivery.

Schneider Electric also said it has successfully tested NVIDIA Nemotron models as part of a new agentic AI alarm management capability. The system is designed to interpret alarms across multiple systems, identify likely root causes and recommend corrective actions using real-time IoT data. 

While all the updates will be welcome for customers using Schneider’s Electric’s ecosystem, much of it will be familiar. That’s because its work with NVIDIA is an evolution on what it has already developed, with Schneider having previously debuted reference designs for NVIDIA’s computing clusters back in 2024. It later updated that work in 2025 to support Blackwell-era designs and Omni-verse based digital twins, and right on schedule, it’s announcing yet another update specifically for NVIDIA’s Rubin platform. 

What is notable, however, is that Schneider is now trying to pull all of the strands of its system together into a more complete offer around AI factories: power, cooling, simulation and operational tooling in one package. That suggests the company is pushing beyond helping operators deploy high-density racks and is instead trying to position itself much earlier in the design process, and much deeper into the day-to-day running of large AI sites. That will be crucial given the speed at which AI data centres are needed.