Lori Data Center Server Liquid Cooling Components: Precise and High Quality Cooling Assurance for Your NVIDIA GB200 AI Servers

With the proliferation of artificial intelligence and high-performance computing (HPC) workloads, the need for efficient thermal management in the data center has never been greater. nvidia's groundbreaking GB200 NVl72 AI server racks, powered by Blackwell GPU's, are redefining computing power, but their immense power also generates unprecedented amounts of heat. At Lori, we specialize in delivering high-performance liquid-cooled components that ensure your GB200 infrastructure runs at peak efficiency, reliability, and sustainability. 

NVIDIA GB200 NVL72 Makes AI and HPC Stronger

NVIDIA GB200 is a super AI chip based on the Blackwell architecture, combining Blackwell GPUs and Grace CPUs designed for AI and High Performance Computing (HPC), with ultra-high bandwidth interconnections via NVLink 5.0 to dramatically increase arithmetic power and energy efficiency. 

That arithmetic power comes at a price: a single GB200 NVL72 rack consumes up to 140 kW, far exceeding the limits of traditional air cooling. As GPU temperatures soar and energy costs rise, liquid cooling is not just an option, but a necessity.

Lori delivers strong support for your AI and HPC servers

At Lori, we design every component to meet the stringent requirements of NVIDIA GB200 servers, reducing PUE to 1.04 and dramatically lowering operating costs. Our engineers' time-tested expertise ranges from single racks to hyperscale data centers, and our solutions grow with your needs.

Direct-to-chip liquid cooling plates

Blackwell GPU's like the B200 have a Thermal Design Power (TDP) of 1,000W, while the GB200 “superchip” (CPU+dual GPU) reaches 2,700 W. Lori's direct liquid cooling solution utilizes advanced cold plates mounted directly to the GPU and CPU. These plates circulate coolant and absorb heat at the source, achieving 30x higher thermal conductivity than air.

Server Manifold & Quick Disconnect

Our server manifolds distribute coolant evenly across the rack while quick-disconnecting for seamless maintenance without downtime. These components are OCP compliant and compatible with NVIDIA's MGX platform, ensuring easy integration into existing infrastructure.

               

Server Mainfold

The copper rack manifold can be manually connected and flexibly arranged vertically or horizontally within the rack to provide an interface between the cold plate loop in the server and the CDU.

               

Quick Disconnect

Lori quick connectors prevent coolant leaks and provide flexible, easy assembly and servicing for your direct liquid cooling loops.

The Future of Artificial Intelligence is Cooler with Lori

The development of AI Artificial Intelligence makes liquid cooling methods will become more and more widespread in data centers. Using Lori's direct liquid cooling solutions can provide your liquid to air (L2A) and liquid to liquid (L2L) cooling systems with the flexibility to be assembled quickly and in a space-saving manner. Allowing you to fully utilize the potential of your NVIDIA GB200 AI servers while minimizing operating costs and environmental impact. Ready to upgrade your data center? Explore our direct-to-chip liquid cooling portfolio and join the ranks of leading AI innovators, trust Lori to provide you with precise thermal management.