Why Belimo is the Control Solution of Choice for AI Data Center Cooling
Published by Jason Tables
on
AI data centers are scaling rapidly, and liquid cooling is already central to how the industry is managing thermal loads. Belimo valves are a natural fit for this evolution. With field-proven performance in high-density environments, precise control across variable conditions, and seamless integration into advanced control architectures, Belimo is already supporting the infrastructure behind today’s most demanding AI and HPC applications.
Where Are Belimo Valves Showing Up in Data Centers?
Belimo valves have become the go-to choice across the AI and HPC data center space. In fact, they are being used by the biggest players in the industry. If you watched the Microsoft Build 2025 keynote, you might’ve noticed something interesting behind Satya Nadella during his conversation with NVIDIA CEO Jensen Huang. On screen was a high-density server rack running a closed-loop liquid cooling system, and mounted front and center were six Belimo valves. These are part of Microsoft’s AI infrastructure, helping manage thermal loads in systems built to support massive compute workloads.
Who else is using Belimo valves for AI and HPC cooling?
Belimo hardware is being used in some of the most advanced computing environments in the world. One example is Japan’s new quantum-AI supercomputer, named ABCI-Q, developed by the National Institute of Advanced Industrial Science and Technology (AIST). The system is designed for quantum simulation and AI research, leveraging 2,020 NVIDIA H100 GPUs along with direct-to-chip liquid cooling to manage its massive thermal output. As shown in the photo below, Belimo valves are integrated into the liquid cooling infrastructure, helping ensure thermal stability in one of the world’s most advanced research systems. The use of Belimo in this application underscores the trust placed in our technology to deliver precise flow control under extreme computational conditions.
Why are Belimo products a good fit for liquid-cooled AI and HPC systems?
High-density computing environments require fluid control solutions that can respond in real time to load variability, hardware changes, and dynamic thermal conditions. Belimo valves, particularly the Energy Valve and EPIV, are designed to maintain precise flow across cold plates, accommodate changing delta T conditions, and support power-based control strategies directly from the BMS. When server configurations shift, Belimo valves maintain design flow or fixed pressure drops automatically, protecting both thermal performance and equipment lifespan. Built-in diagnostics, remote commissioning, and seamless integration with existing control architectures make them a solid choice for liquid-cooled data center applications.
What kinds of control does Belimo support for these systems?
For direct-to-chip cold plate cooling, Belimo supports flow control, power control, and differential pressure control.
Flow control uses pressure-independent valves like the EPIV to hold flow steady, regardless of pressure fluctuations. Power control, enabled by the Energy Valve’s integrated sensing and logic, goes a step further by using delta T and flow rate to modulate thermal energy output precisely. This helps match cooling output to actual compute demand without overcooling or wasting pump energy.
Differential pressure control is a game changer for systems where server configurations change frequently. Taking a server offline alters the hydraulic resistance in the loop. The Belimo Energy Valve, when paired with a differential pressure sensor, detects these changes and adjusts automatically to maintain the correct pressure drop, preserving safe, efficient flow across the remaining cold plates.
Can Belimo valves handle the fast-changing demands of AI cooling?
They already are. Our Energy Valves and EPIVs are providing dynamic control, real-time performance data, and flexible commissioning in the most demanding environments where nothing is static. Systems with high-density GPUs, experimental hardware, or fluctuating workloads need valves that adjust on the fly, and Belimo is built for that. With tools like integrated delta T monitoring and BACnet/Modbus compatibility, you’re not just getting a valve. You’re getting system-level intelligence designed for the speed and scale of modern data center workloads.
What makes Belimo different?
It’s the combination of being Reliable, Scalable, and Efficient. Our valves are trusted in the kinds of data centers where failure isn’t an option. The hardware is rugged and flexible. The control is accurate and responsive. And the built-in diagnostics and performance feedback enable consistent, responsive cooling, even as conditions shift. When AI workflows push infrastructure to the edge, Belimo delivers proven cooling control solutions that keep thermal performance stable, efficient, and ready to scale.
Image Credits and Sources
Figure 1: Screenshot from Microsoft Build 2025 | Satya Nadella Opening Keynote by Microsoft, published on YouTube, May 19, 2025. https://www.youtube.com/watch?v=ceV3RsG946s&t=6060s
Figure 2: Image sourced from NVIDIA Powers World’s Largest Quantum Research Supercomputer, NVIDIA Newsroom, published May 18, 2025. https://nvidianews.nvidia.com/news/nvidia-powers-worlds-largest-quantum-research-supercomputer