Partner POV | Cooling the AI Factory: Boyd Thermal by Eaton powers NVIDIA MGX™
In this article
This article was written and contributed by Eaton.
AI data centers are scaling rapidly to meet the intense compute demands of generative models, large language model inference and agentic AI workloads. These workloads drive greater compute intensity, contributing to higher energy use and thermal complexity. As racks grow denser and hotter, maintaining efficiency and uptime becomes both challenging and critical.
Boyd Thermal by Eaton solves these challenges with innovative liquid cooling solutions built for the architecture that is setting the production standard for AI infrastructure: NVIDIA MGX™.
The thermal challenge for NVIDIA MGX installations
Modern agentic AI systems plan tasks, invoke tools, execute code and coordinate across continuous multi-step workflows. These interactions generate large volumes of reasoning tokens, expand key-value caches and place sustained high-throughput demands on GPU clusters. The result is thermal loads that conventional air-cooling systems cannot address.
Operators managing NVIDIA accelerated GPU clusters are facing higher rack power densities and liquid cooling requirements that are driving a fundamental evolution in thermal management architecture. At hyperscale, even minor inefficiencies compound into major operating expenses. Innovative cooling strategies that maximize performance per watt are no longer optional. They are essential.
What is NVIDIA MGX™?
NVIDIA MGX™ is the modular reference architecture that is defining how AI factories are built, from single-node servers through to full rack-scale deployments. Third-generation MGX racks are designed for around 100 percent liquid cooling, with a modular, cable-free and fan less architecture that places extreme demands on every component in the thermal loop. For Boyd Thermal by Eaton, NVIDIA MGX is the platform we engineer to.
Meeting that standard means delivering precise coolant distribution, minimal pressure drop, and consistent thermal performance at every level of the stack. That is exactly what our NVIDIA MGX-validated solutions are built to do.
Boyd Thermal by Eaton NVIDIA MGX solutions
Boyd Thermal by Eaton liquid cooling solutions designed and validated for NVIDIA MGX rack-scale infrastructure, delivering scalable thermal performance for next-generation AI and high-performance computing environments.
VR200 inner manifold
Engineered for NVIDIA MGX compute trays, the VR200 inner manifold delivers optimized coolant distribution at the component level, minimizing pressure drop and ensuring uniform thermal performance across high-density GPU configurations.
VR200 rack manifold
Operating at rack scale, the VR200 rack manifold manages coolant flow across the full NVIDIA MGX rack architecture, supporting the cable-free and hose-free architecture that defines third-generation NVIDIA MGX design.
ROL2300 in-row CDU
The ROL2300 is a high-capacity in-row cooling system designed to deliver efficient thermal management by leveraging facility water. With a cooling capacity of up to 2.3 MW, it supports multiple server racks simultaneously, making it well suited for the power densities demanded by NVIDIA MGX rack-scale deployments.
Chip-to-Ambient™
Boyd Thermal by Eaton takes a system-level view of thermal management. Our innovative solutions address the full cooling chain, from component-level manifolds and cold plates through rack-scale distribution and facility-level coolant distribution units (CDUs), engineered to work together as one integrated thermal system.
Our liquid cooling solutions minimize pressure drops and maintain precise flow control across the entire loop, efficiently dissipating heat while reducing water and energy consumption. Our in-row CDUs support low-approach-temperature operation, giving data center operators the flexibility to adopt free air cooling even in warmer climates, reducing total cost of ownership without compromising uptime.
We work closely with NVIDIA accelerated system integrators and data center operators to deliver validated, closed-loop cooling systems tailored to their specific infrastructure. From initial design through deployment and ongoing support, Boyd Thermal by Eaton is a long-term thermal partner for the most demanding AI factory environments.
Built for what's next
AI factory infrastructure is evolving fast, and the thermal demands that come with each new generation of accelerated computing will only grow. Boyd Thermal by Eaton is built for that reality.
Our NVIDIA MGX-validated solutions are designed with forward compatibility in mind, ensuring that as NVIDIA platforms evolve, our customers can build on their existing facilities and infrastructure rather than replacing them. We engineer our innovative liquid cooling solutions to integrate with and enhance what is already in place, so data centers can scale capacity and performance without starting from scratch.