Siemens and partners collaborate on NVIDIA AI data center reference architecture

Siemens, in collaboration with NVIDIA, Fluence, and nVent, has joined forces to develop a cutting-edge reference architecture tailored for next-generation AI data centers centered around NVIDIA’s DSX Vera Rubin NVL72 platform. This innovative design targets hyperscalers, colocation operators, and cloud providers seeking to implement high-density AI infrastructure at scale.

The initiative comes at a time when AI factories are revolutionizing data center design requirements, particularly in the realms of power delivery, cooling systems, and operational resilience. Siemens emphasizes that the reference architecture effectively translates NVIDIA’s AI factory concept into a practical electrical, power, and controls blueprint capable of scaling from tens to hundreds of megawatts.

Engineered for Intense AI Workloads

The reference design boasts a total facility capacity of 136 MW with a 100 MW IT load, reflecting the escalating power needs of AI compute clusters. This comprehensive architecture spans the entire electrical chain, starting from a 34.5 kV utility connection, through medium-voltage distribution, down to low-voltage modular power blocks at the rack level.

For readers of eeNews Europe, this project underscores how AI is propelling novel approaches to industrial power systems, modular infrastructure, and energy management. It also showcases how traditional industrial automation vendors are strategically positioning themselves within the AI infrastructure market.

Siemens has indicated that the foundational architecture is geared towards Tier III concurrent maintainability, enabling the removal of any single component without disrupting IT operations. The modular approach aims to streamline phased expansion while minimizing redesign needs as capacity expands.

The design also integrates nVent-aligned electrical parameters and is anticipated to evolve further with enhanced thermal management capabilities, particularly focusing on liquid cooling technologies essential for high-density AI racks.

“nVent has globally deployed over two gigawatts of liquid cooling capacity,” stated Sara Zawoyski, President of nVent Systems Protection. “This operational experience enables us to assist partners like Siemens in translating reference architectures into deployable thermal solutions that deliver reliable performance from day one at this scale. Platforms like NVIDIA Vera Rubin NVL72 are pushing rack densities beyond the capabilities of traditional air-cooled infrastructure.”

Enhanced Power Resilience and Expedited Deployment

A primary focus of the architecture is to mitigate deployment risks and accelerate revenue generation for AI infrastructure operators. Siemens has highlighted that its prefabricated and factory-tested medium- and low-voltage skids are designed to reduce on-site construction complexity and commissioning durations.

“Siemens’ profound expertise in power systems, controls engineering, modular infrastructure, and industrialized delivery is clearly evident in this latest joint reference architecture design,” remarked Ruth Gratzke, President of Siemens Smart Infrastructure USA. “Our pre-engineered, prefabricated, and factory-tested medium- and low-voltage skids help streamline on-site construction complexity, shorten commissioning cycles, and enhance quality, safety, and repeatability across deployments.”

The architecture also integrates Fluence battery energy storage systems to offer grid support and operational flexibility in power-constrained environments. According to Fluence, functionalities such as black start, demand response, and AI load smoothing are becoming increasingly crucial for large AI facilities connected to strained electricity grids.

The entire system is interconnected through a centralized Integrated Data Center Management Suite, providing operators with comprehensive visibility across power, cooling, and compute infrastructure through a unified interface.