Why Not All Immersion Cooling Systems Are the Same

As datacentres move deeper into the era of AI, high-performance computing, and increasingly dense compute architectures, traditional cooling models are approaching their practical limits. The growing interest in datacentre immersion cooling is a direct response to these pressures. Even relatively simple immersion deployments can outperform air cooling in terms of thermal efficiency and achievable density.Why Modular 12U is a Deliberate Design Choice

However, while immersion cooling is inherently more capable than air-based approaches, not all immersion systems deliver the same level of performance, stability, or long-term operational value. Differences between immersion platforms are often substantial, and many of the challenges encountered in real-world deployments stem not from the cooling fluid itself, but from how the system is engineered.

Immersion cooling is fundamentally an IT cooling technology. When systems are designed primarily from a mechanical or facility perspective, rather than around the behaviour of IT hardware, the benefits of immersion can be diluted or lost entirely.

The Fluid Question

One of the first questions raised in immersion cooling discussions is fluid selection. Operators want to understand whether one dielectric fluid cools better than another, how different chemistries age, and whether fluid choice materially affects operational outcomes.

There are differences between fluids, and these differences matter at a specification and compliance level. Reputable immersion fluids are designed to meet established requirements for dielectric strength, thermal stability, safety, and material compatibility. When these criteria are met and industry guidelines are followed, fluid choice alone rarely determines the success of a deployment.

In practice, the dominant differentiator is the immersion cooling system itself, not the liquid it contains. A compliant fluid placed in a poorly engineered system will underperform, whereas a well-engineered system paired with a compliant fluid will operate predictably and reliably. The fluid enables immersion cooling, but the system determines its outcome.

Immersion Cooling Engineered for IT

A common misconception is that immersion cooling is simply another branch of conventional datacentre cooling infrastructure. In reality, immersion cooling replaces the air-based paradigm entirely and operates at a different physical and engineering level. Traditional datacentre cooling systems like CRAC and CRAH units or airflow containment are designed to condition rooms and manage bulk air temperature. Immersion cooling removes heat directly from IT components within a liquid environment.

Inside a server, heat generation is dynamic and uneven. GPUs ramp rapidly under load, CPUs cycle, and VRMs create localised hotspots. These behaviours cannot be addressed effectively through facility-level engineering alone. Effective immersion cooling therefore requires IT-centric engineering. This includes component and board-level thermal modelling, detailed analysis of heat paths, and fluid flow simulation within the enclosure to eliminate stagnation zones and vortices through controlled liquid distribution.

Systems designed around air cooling cannot simply be adapted to immersion. They must be re-engineered around the physics of semiconductor heat transfer in liquid environments.

Why Thermal Modelling Matters More Than Pump Power

Some immersion cooling solutions emphasise pump size or total mass flow as indicators of performance. While these metrics are easy to communicate, they do not determine chip temperature stability or system efficiency on their own. In immersion cooling, precision consistently outperforms brute force. What matters is not how much liquid is moved, but how it moves: whether the fluid reaches the right components, at the right velocity, and in the right pattern.

Without detailed modelling and simulation, hotspots develop behind tall or densely packed components, cold zones remain underutilised, and fluid stratification becomes unpredictable. Poorly engineered immersion systems may appear functional at low load, but become unstable and difficult to operate as density increases. Well-engineered systems deliver uniform and predictable thermal performance across the entire deployment.

Passive and Active Immersion Engineering Platforms

Different workloads and density targets place fundamentally different demands on immersion cooling systems. For this reason, our immersion portfolio is intentionally built around two distinct engineering platforms: passive (natural convection) immersion and active (forced convection)
immersion. These are not minor variations of the same design. They are foundational categories, each engineered to support a specific performance envelope while sharing common principles around serviceability, predictability, and integration.

Passive (Natural Convection) Immersion Cooling

Our passive immersion platform relies on buoyancy-driven flow, where heated liquid rises and cooler liquid sinks. When the thermal environment is precisely engineered, natural convection delivers stable cooling with extremely low energy overhead and minimal mechanical complexity. This behaviour is not assumed; it is validated through component-level thermal modelling and flow analysis. Tank geometry, server placement, and internal clearances are engineered to ensure predictable convection patterns rather than chaotic circulation. This platform is well suited to moderate-to-high-density deployments where efficiency, simplicity, and long-term operational stability are primary objectives.

Active (Forced Convection) Immersion Cooling

For sustained high-density AI and HPC workloads, our active immersion platform introduces controlled internal flow using engineered pumping. This enables precise liquid delivery to high-heat-flux components and maintains uniform temperatures under extreme and highly dynamic load conditions. Rather than relying on increased mass flow alone, these systems are designed around targeted, modelled flow paths that eliminate stagnation zones and thermal gradients. This ensures consistent thermal behaviour across all server positions as power density increases. Both platforms reflect the same principle: immersion cooling must be engineered for IT behaviour, not adapted from facility cooling concepts.

Why Modular 12U is a Deliberate Design Choice

Across both passive and active immersion platforms, we standardise on 12U modular immersion compartments. This is a foundational engineering decision rather than a packaging preference. In immersion cooling, the enclosure geometry defines the thermal environment. Smaller, modular compartments allow for tighter control of flow patterns, more uniform heat transfer, and predictable thermal behaviour across the IT estate. They also reduce operational risk by limiting fluid volume per compartment and improving mechanical accessibility.

Compared to large, monolithic tanks, 12U modular compartments promote homogeneous and repeatable cooling, reduce thermal stratification and flow uncertainty, and simplify servicing and hardware access. Additionally, they support clean cabling and integrator-friendly layouts and scale predictably into multi-megawatt deployments. By using the same modular containment approach across both immersion platforms, consistency is maintained in deployment, operation, and expansion, while still addressing very different density and workload requirements. In immersion cooling, modular precision consistently delivers better outcomes than scale without control.

Immersion Cooling Systems vs. Immersion Tanks

In industry discussions, the terms immersion cooling system and immersion tank are often used interchangeably. While related, they describe different levels of abstraction and conflating them can lead to incomplete or misleading comparisons.

An immersion tank refers to the core product functionality: a physical enclosure designed to contain dielectric fluid and cool IT hardware through direct liquid contact. At this level, the focus is on containment, fluid compatibility, mechanical integrity, and basic thermal performance. An immersion cooling system, by contrast, is a broader architectural concept. It encompasses not only the tank itself, but also the surrounding design choices that determine how the solution behaves in a real datacentre environment. This includes how tanks are configured, connected, monitored, serviced, and integrated into operational processes.

A system-level approach focuses on delivering configurations tailored to specific use cases, such as designs optimised to meet a particular service-level agreement, redundancy requirement, safety profile, or operational model. These decisions influence how the system responds to faults, how maintenance is performed, and how predictable performance remains under sustained load. This distinction matters because many of the most important operational KPIs are shaped by system-level design choices, not by the tank alone. Factors such as availability, maintainability, resilience, safety, and operational risk are often only partially visible in a standalone product datasheet.

As a result, two immersion tanks with similar headline specifications may perform very differently once deployed as part of a complete system. Redundancy architecture, compartmentalisation, fluid segmentation, monitoring strategy, and service workflows all affect outcomes that are critical to operators but difficult to infer from product specifications alone. For this reason, immersion solutions should be evaluated in context. Understanding how a vendor defines and engineers the complete system is as important as reviewing the technical characteristics of the tank itself.

Immersion Cooling Should Enable Integration, Not Complicate It

For immersion cooling to succeed at scale, it must support not only datacentre operators but also system integrators building AI clusters, storage platforms, and network fabrics.

An effective immersion cooling system should simplify cabling and physical access, reduce deployment time, minimise node-to-node thermal variability, preserve predictable mechanical interfaces, and support flexible cluster architectures. If an immersion system makes integration harder, it undermines its own value. The objective is not simply to remove heat, but to enable higher performance, improved reliability, and more sustainable infrastructure design.

Conclusion

The fluid matters but the engineering matters more.

Immersion cooling is becoming central to the future of datacentre infrastructure. But its success depends far more on system engineering than on fluid chemistry. Within accepted standards, differences between fluids have limited operational impact.

Poorly engineered immersion systems introduce long-term challenges in performance, stability, and serviceability. Well-engineered systems unlock the full potential of immersion cooling. The future of datacentre cooling belongs to immersion, but only to solutions designed with an IT-first philosophy, rigorous thermal modelling, modular precision, and integrator-friendly architecture. Not all immersion cooling systems are the same. The most effective ones are engineered deliberately, not improvised.