Siemens and nVent announced a collaboration to develop a reference architecture for hyperscale AI data centers, targeting facilities of up to 100 MW and naming NVIDIA DGX GB200 systems as an intended compute environment. It is a design blueprint, not a publicly documented customer deployment: the announcements do not establish a completed facility or independently measured performance.
What did Siemens and nVent announce?
The companies said they would develop a liquid-cooling and power reference architecture for large-scale AI infrastructure. Siemens describes the design as Tier III-capable and says it brings together Siemens electrical and automation systems, NVIDIA DGX SuperPOD reference designs, and nVent liquid cooling. NVIDIA DGX GB200 systems are identified as an example of the intended compute environment. nVent’s December 10, 2025 announcement and Siemens’ announcement the same day describe the collaboration.
The 100 MW figure is the target facility scale for the design, not the measured output of a built site or confirmation that a particular 100 MW project is underway.
What does the architecture include?
A reference architecture is a design framework showing how components and systems can fit together. It can guide implementation, but it is not itself a finished data center or a guarantee that every deployment will use identical equipment.
#1 Best Overall
Siemens: electrical and automation systems
Siemens’ described scope covers industrial-grade electrical and automation systems. Its broader data-center portfolio includes medium- and low-voltage power distribution, automation, and energy-management software. The announcement does not provide a complete project bill of materials or specify every component for the joint design.
nVent: liquid cooling
nVent contributes the liquid-cooling scope. In a separate November 17, 2025 portfolio announcement, nVent listed row- and rack-based coolant distribution units (CDUs), technology cooling system manifolds, updated racks, intelligent power distribution units (PDUs), and installation and maintenance services. That broader portfolio context does not confirm that every listed product or service is part of the Siemens–nVent architecture.
NVIDIA: compute-system context
The design names NVIDIA DGX GB200 systems and DGX SuperPOD reference designs as its compute context. NVIDIA is not one of the two companies named in the collaboration headline.
Is the design available, or is it a blueprint?
The cited announcements describe a reference architecture to be developed; they do not document a named customer installation, a completed facility, or a final implementation available as a deployed system. They also do not publish independent operating results for this architecture.
Recommended Free Tools
For an organization evaluating an implementation, the announcements leave key configuration questions open: the target compute platform and rack density, facility power capacity and distribution, cooling and heat-rejection requirements, redundancy and availability design, serviceability, and deployment scope. The available details are not enough to rank particular configurations on those criteria.
What does “Tier III-capable” mean here?
“Tier III-capable” is the companies’ characterization of the design. The announcement does not certify a particular facility or establish that every implementation will meet a defined uptime outcome. A capability claim about a blueprint should not be treated as a certification or a guarantee for a future site.
How do power and liquid cooling work together?
AI data centers need electrical infrastructure to deliver power to computing equipment, while cooling systems remove heat from that equipment. In a liquid-cooled design, cooling components such as CDUs and manifolds can form part of the path that circulates coolant through the facility’s cooling system. The Siemens–nVent announcement presents the two scopes as parts of one architecture, but it does not publish enough engineering detail to specify a particular site’s electrical layout, cooling loop, heat-rejection system, or redundancy configuration.
What performance claims are established?
The companies describe intended benefits, including faster deployment, sustainability, energy efficiency, and improved tokens-per-watt. Those are statements about the design’s aims, not independently verified results. Siemens executive Ciaran Flanagan said the architecture “accelerates time-to-compute and maximizes tokens-per-watt, which is the measure of AI output per unit of energy.” nVent executive Sara Zawoyski said the architecture would help data center managers deploy cooling infrastructure for AI. Neither statement supplies measured outcomes for an operating joint-architecture deployment.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →How does the 2026 Siemens design relate?
In a separate announcement on June 1, 2026, Siemens described a DSX Vera Rubin-aligned reference design developed with NVIDIA and Fluence, with nVent-aligned design considerations. Siemens said a forthcoming supplement would expand the design to include advanced thermal-management capabilities. This is a separate development; it does not establish that the 2025 Siemens–nVent architecture was deployed or that its final design was released. Siemens’ June 2026 announcement provides that later context.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
