Two Phase Immersion Cooling: How Does It Work?

What Is Two-Phase Immersion Cooling? Advantages and Limitations

 

As AI and HPC servers generate more heat within increasingly dense deployments, data centers are exploring cooling methods that do not depend on moving large volumes of air through server rooms.

 

Two-phase immersion cooling has attracted attention because it can collect heat directly from submerged server components and reduce reliance on server fans and room-level airflow. However, it also introduces different requirements for infrastructure, cooling fluids, maintenance, and hardware compatibility.

 

This article explains how two-phase immersion cooling works, its main advantages and limitations, and the factors to consider when comparing it with other liquid cooling architectures.

 

What Is Two-Phase Immersion Cooling?

 

The term two-phase immersion cooling describes both the coolant’s behavior and the system’s physical arrangement. Two-phase means that the coolant changes between liquid and vapor. Immersion means that the server hardware is submerged directly in the coolant.

 

The system uses a dielectric fluid that does not conduct electricity, allowing it to contact server components without causing a short circuit. As the fluid absorbs heat from the hardware, part of it changes into vapor. The vapor then releases the collected heat, condenses, and returns to the tank as liquid. This creates a continuous cooling cycle.

 

How Does Two-Phase Immersion Cooling Work?

 

Two-phase immersion cooling uses the boiling and condensation of a dielectric fluid to move heat away from server components.

 

Server components generate heat → dielectric fluid boils → vapor rises → condenser removes heat → liquid returns to the tank

 

When the servers operate, components such as CPUs, GPUs, memory, and power supplies release heat into the surrounding fluid. Once the fluid near these surfaces reaches its boiling point, part of it changes into vapor. This phase change absorbs latent heat from the hardware.

 

The vapor rises to a condenser located inside or above the immersion tank. A cooler liquid flows through the condenser and absorbs heat from the vapor without mixing with it. As the vapor releases heat, it condenses into liquid and falls back into the tank.

 

The dielectric fluid can then absorb heat from the server components again, creating a continuous cooling cycle.

 

What Are the Advantages of Two-Phase Immersion Cooling?

 

Two-phase immersion cooling collects heat directly from submerged server components and uses phase change to carry it away. This provides several potential advantages for high-density computing environments.

 

  • Direct Heat Capture

The dielectric fluid surrounds the server components and absorbs heat directly from their surfaces. This allows the system to collect heat from CPUs, GPUs, memory, power supplies, and other components within the same tank.

 

  • Reduce Reliance on Fans and Airflow

Because the cooling fluid contacts the hardware directly, the servers require fewer or no internal cooling fans. The data center also relies less on high-volume airflow, hot and cold aisles, and room-level air conditioning to remove server heat.

 

  • Higher Deployment Density

Reducing the need for server fans and airflow space can allow computing hardware to be installed more densely. This can support AI, HPC, and other applications that require substantial computing capacity within a limited area.

 

  • Potential for a Lower PUE

Server fans and room-level cooling equipment contribute to a data center’s non-computing energy use. Reducing their workload can help lower this energy demand and improve power usage effectiveness, or PUE.

The actual PUE still depends on the complete data center design, including the immersion tanks, condensers, pumps, heat-rejection equipment, and operating conditions.

 

What Are the Limitations of Two-Phase Immersion Cooling?

 

Two-phase immersion cooling places server hardware directly in a tank of dielectric fluid. This design allows the fluid to collect heat from many components, but it also introduces specific facility, maintenance, and hardware requirements. Before adopting the system, operators should consider the following limitations.

 

  • Floor Loading and Space Requirements

An immersion system combines the weight of the tank, servers, cooling fluid, piping, and related equipment. The data center floor must be able to support this load.

Operators also need enough space to install the tanks, move hardware, and lift servers out for maintenance. These requirements must be considered in both the equipment layout and maintenance area.

 

  • Integration with Existing Infrastructure

Immersion tanks use a different layout from conventional server racks and cold-plate liquid-cooling systems. An existing data center may need changes to its floor plan, piping, equipment transportation routes, power connections, and maintenance procedures.

For this reason, immersion cooling may be easier to include when designing a new facility than when converting a data center built around rack-based equipment.

 

  • Chip-Level Cooling Capacity

Immersion cooling can collect heat from many components across a server. However, this does not automatically determine how much heat it can remove from one high-power chip.

Chip-level cooling capacity depends on the coolant, heat-source area, boiling surface, fluid movement, condenser, and operating conditions. As chip TDP and heat flux increase, operators should verify whether a commercial system can support the intended hardware under continuous operating conditions. 

 

  • Cooling-Fluid Cost and Environmental Impact

Two-phase immersion cooling requires a dielectric fluid with suitable boiling and material properties. The initial fluid volume, possible replenishment, recovery process, and end-of-life handling can add to the system’s long-term cost.

Fluid composition varies. Some two-phase immersion fluids are fluorinated and may raise concerns related to PFAS or global warming potential. Buyers should review the safety data sheet and environmental information for the specific coolant instead of assuming that all dielectric fluids have the same impact.

 

  • Maintenance, Material Compatibility and Warranty

Servers must be removed from the tank before many maintenance tasks can be performed. Components may also need to drain or be cleaned before technicians can handle, repair, or replace them.

The selected fluid must be compatible with circuit boards, plastics, elastomers, seals, cables, and other server materials. Prolonged exposure can affect some materials if compatibility has not been validated.

Warranty coverage also varies among server manufacturers and immersion solutions. Operators should confirm whether the hardware, coolant, and tank configuration are approved and which party is responsible for warranty and service support.

 

Two-phase immersion cooling may therefore be best suited to facilities that can accommodate its weight, layout, coolant-management, and maintenance requirements from the planning stage.

 

Is Two-Phase Immersion Cooling Right for Your Data Center?

 

Two-phase immersion cooling offers direct heat capture, reduced reliance on airflow, and strong PUE potential. However, its suitability depends on whether the facility can accommodate immersion tanks, liquid weight, specialized maintenance, material-compatibility requirements, and the intended chip heat load.

 

Projects that cannot accommodate these requirements may consider a two-phase direct-to-chip system instead. This architecture applies phase-change cooling at selected CPUs and GPUs without submerging the server hardware.

 

Explore CreAlien W2PC for a DI-water two-phase direct-to-chip approach to high-power AI server cooling.