Schneider Electric unveiled its infrastructure design and operational direction for large-scale AI data centers at NVIDIA's annual event, GTC 2026. This announcement is significant as it presented an integrated infrastructure concept to handle gigawatt-class facilities, focusing on the fact that power, cooling, control, and digital technologies must be considered together during the construction of AI data centers.

The company announced on the 18th that it introduced key technological advancements related to next-generation AI data centers during its participation in GTC 2026. At the event, Schneider Electric officials are scheduled to participate in panel sessions and keynote presentations to explain ways to improve the design and operational efficiency of large-scale AI infrastructure.

The key features of this announcement are a reference design and digital twin technology tailored to NVIDIA's latest rack-scale systems. The new design is based on the NVIDIA Vera Rubin NVL72 rack and includes power and cooling infrastructure, configured to work in conjunction with Schneider Electric's control system design. By incorporating Omniverse-based digital twin technology, it is possible to review performance and operating conditions in a virtual environment prior to actual implementation.

Along with this, initial tests of agentic AI applicable to data center operations were also unveiled. A representative example is an alarm management service utilizing the NVIDIA Nemotron open model. In large-scale AI data centers, rapidly analyzing alarms and anomalies occurring during operation becomes critical as the scale of facilities increases, and this is an attempt to assist with this process using AI.

The target applications are data centers and AI factories requiring ultra-large-scale AI computing infrastructure. Its potential for use is being discussed not only by cloud providers and companies operating high-performance computing environments, but also across the entire industry that provides services based on large-scale GPU resources. It differs from existing data center construction methods in that it requires simultaneous consideration of power supply, cooling, building design, and computing operations, rather than simply expanding servers.

Collaboration with industrial software company AVEVA has also expanded. Together with NVIDIA, AVEVA announced a lifecycle digital twin architecture for gigawatt-scale AI factories, presenting a structure that connects the design phase to operations and maintenance. As the importance of pre-simulation and operational optimization grows with the scale of AI infrastructure, this announcement is interpreted as an example demonstrating that the data center industry is shifting from a hardware-centric approach to an integrated operating system-centric one.