In collaboration with NVIDIA and Fluence, Siemens has introduced a reference design for AI factories that translates NVIDIA’s AI factory blueprint into a deployable framework. The design is intended for hyperscale and specialised cloud infrastructure, addressing the growing requirements of AI-focused data centres.
The NVIDIA Vera Rubin NVL72 platform introduces new infrastructure requirements, particularly in power and cooling. As a result, AI factory operators face decisions around site selection, grid interconnection, capital efficiency, and the integration of emerging technologies.
Siemens’ reference design provides a total facility capacity of 136 MW, including 100 MW allocated to IT load. Its electrical and control architecture covers the full power delivery path, from 34.5 kV utility connections through modular power blocks to rack interface endpoints. Designed for concurrent maintainability, the system allows individual components to be removed from service without disrupting operations. It also supports phased capacity expansion, from tens to hundreds of megawatts, and is aligned with NVIDIA deployment units.
The design incorporates nVent-aligned parameters to support compatibility with NVIDIA workloads across a range of AI applications. Siemens has also indicated that a future supplement is expected to add further thermal management capabilities.
In addition, the design integrates Fluence battery energy storage systems, enabling AI facilities to operate in environments with limited power availability. This approach is intended to support scalability and reliability as AI infrastructure expands.
The infrastructure also includes a centralised Integrated Data Centre Management Suite, providing operators with visibility across power, cooling, and computing assets. Siemens contributes energy management and automation technologies intended for mission-critical data centre environments.
The collaboration is intended to support AI infrastructure deployment by providing a standardised framework that aims to improve planning predictability, support deployment timelines, and maintain flexibility for future technology developments.