º£½Ç´óÉñ

Artificial intelligence isn't just driving innovation in software — it's demanding a fundamental shift in the very foundations of infrastructure. What were once simple brick-and-mortar buildings are transforming into smart structures capable of handling the extreme digital demands of an increasingly AI-driven society. 

The escalating power densities required by modern AI workloads push data centers beyond the limits of traditional air cooling, necessitating a strategic embrace of liquid cooling technologies. For construction professionals, data center owners and operators, understanding and effectively implementing these advanced cooling solutions is key for future success and operational efficiency.

The Need for More Cooling Power

The efficiency and effectiveness of liquid cooling compared to air are no longer debatable for high-density environments. As server racks become increasingly packed with powerful processors and accelerators, the superior thermal conductivity of liquids proves to offer a more direct and efficient heat dissipation pathway. For owners, this optimal cooling technology is essential for maintaining hardware performance and preventing costly downtime but also for the long-term viability and scalability of their data center operations.

To meet these growing demands for increased power density and energy efficiency, designers and builders are turning to innovative liquid cooling solutions. Among the forefront of these advancements are submersion cooling — which involves fully immersing IT hardware in a dielectric fluid —and direct-to-chip cooling, which channels coolant directly to the hottest components. 

These cutting-edge approaches represent a significant paradigm shift from traditional air-cooling methods, enabling unprecedented thermal management and unlocking the full potential of high-performance computing infrastructure.

Submersion Cooling

Submersion cooling, also known as immersion cooling, involves submerging IT hardware directly into a dielectric fluid that does not conduct electricity. This method offers unparalleled heat removal efficiency by providing direct contact with the heat-generating components. 

Green Mountain, a Norwegian data center provider for clients like TikTok, has successfully implemented in some of its data centers. Servers are submerged in tanks filled with a specialized dielectric fluid (such as Mineral and Silicone Oils). This fluid directly absorbs heat from components, then flows through a heat exchanger to an external cooling system for heat dissipation. This innovative approach allows Green Mountain to achieve significantly higher power densities per rack and realize a much-improved PUE (Power Usage Effectiveness) compared to traditional air-cooled environments.

Then there are companies like (Green Revolution Cooling), a Texas-based immersion cooling startup, who has pioneered . In this method, IT hardware is submerged in a low-boiling-point dielectric fluid, often using .

As components heat up, the fluid boils into a gas, effectively carrying heat away. This gas then rises to a condenser coil, cools and condenses back into liquid, dripping back into the tank to repeat the cycle. This continuous evaporation and condensation provides extremely efficient heat transfer, with projects often boasting impressive PUEs.

Direct-to-Chip Cooling

As immersion cooling has gained traction, so has direct-to-chip liquid cooling, which involves routing a liquid coolant directly to the hottest components on a server, such as CPUs and GPUs, through specialized cold plates. This allows for highly localized and efficient heat removal.

One of the earliest and most significant examples of direct-to-chip cooling was (2010). This groundbreaking system utilized hot water cooling, directly channeling water to cold plates mounted on the processors. Beyond primary cooling, the warm water was cleverly reused to heat the building, showcasing a highly efficient energy reuse model. This project demonstrated the viability of liquid cooling for high-performance computing and its substantial potential for energy savings and improved sustainability.

More recently, Nvidia has been at the forefront of liquid cooling, claiming to address AI’s water consumption issues with direct-to-chip innovations. Nvidia's DGX systems, designed for intensive AI and deep learning, increasingly incorporate direct-to-chip liquid cooling for their powerful GPUs. These systems feature specialized cold plates that sit directly on top of the GPU modules, with coolant circulated through them.

This targeted cooling is essential for maintaining optimal performance of high-wattage components, preventing thermal throttling and enabling denser deployments of critical AI hardware within data centers.

"We have real data with integrity to base our decisions on."

Learn how Vantage Data Centers created a single source of truth to improve data analytics, financial performance and communication across five continents.

Read Their Story

Designing, Building and Operating Next-Gen Data Centers

The adoption of these sophisticated liquid cooling technologies has fundamentally altered the blueprint for data center builds, demanding a new level of collaboration and specialized expertise from all stakeholders across different phases of construction.

Design Phase

Designers are at the forefront, pivoting from traditional HVAC-centric designs to intricate architectures integrating complex piping, fluid containment and advanced leak detection systems. This requires a deep understanding of fluid dynamics and thermal properties. Their role extends to embedding robust safety protocols from the earliest stages, accounting for unique dielectric fluid properties and high-pressure liquid distribution.

The Construction Phase

Builders and construction professionals are tasked with the meticulous execution of complex data center designs. The precision required for installing liquid cooling systems is significantly higher, demanding exact pipe fitting, hermetic seals and flawless integration with power, networking and security infrastructure.

Specialized trades and highly skilled technicians become indispensable, requiring meticulous planning, rigorous quality control and adherence to stringent safety standards. Adapting to new materials, sophisticated sensors and unconventional installation techniques to stay flexible with the project at hand should stay top-of-mind.

Project Delivery & Operation

The successful delivery and ongoing operation of these high-tech data centers hinge on efficient project management and robust information sharing. Given the intricate nature of liquid cooling deployments — involving multiple specialized vendors, novel materials and tight integration schedules — seamless communication and real-time visibility across all project phases are paramount.

Centralized project management platforms that facilitate collaborative workflows, manage documentation, track progress and enable proactive problem-solving become essential. 

Building the liquid-cooled data centers of tomorrow is a collective endeavor, requiring integrated strategies and cutting-edge approaches to enable the continued advancement of AI and pave the way for more sustainable and resilient digital infrastructure.

Written by