Key Takeaways
- Single-phase uses a fluid that stays liquid throughout; two-phase uses a fluid that boils and condenses for heat transfer.
- Single-phase is more mature, lower fluid cost, and simpler to maintain — the right choice for most enterprise deployments.
- Two-phase offers higher heat removal per unit volume and passive circulation — ideal for extreme-density GPU clusters.
- Two-phase fluids cost significantly more and require tighter containment to manage vapour losses.
- The right choice depends on workload density, budget, operational capability, and risk tolerance.
How Single-Phase Immersion Cooling Works
In single-phase immersion, IT hardware is fully submerged in a dielectric fluid — typically mineral oil, synthetic ester, or a hydrocarbon-based fluid. The fluid absorbs heat from components through direct contact and remains in its liquid state throughout the process. Pumps circulate the warm fluid through an external heat exchanger (plate or shell-and-tube), where heat is transferred to a secondary cooling loop connected to dry coolers, adiabatic units, or chilled water.
The system is conceptually straightforward: submerge, pump, exchange, return. This simplicity is its greatest strength. Single-phase systems have been deployed in production data centres for over a decade and are well-understood operationally. Common fluid options include:
- Mineral oil: Lowest cost, widely available, proven track record. Typical fluid cost: $2–$5 per litre.
- Synthetic esters: Higher thermal performance, biodegradable, slightly higher cost. Good balance of performance and environmental profile.
- Engineered hydrocarbons: Purpose-designed for immersion cooling with optimised thermal properties and material compatibility.
How Two-Phase Immersion Cooling Works
Two-phase immersion uses engineered fluorocarbon-based fluids with very low boiling points (typically 34–60°C). Hardware is submerged in the fluid, and as components generate heat, the fluid boils at the chip surface. The resulting vapour rises to a condenser coil at the top of the tank, where it condenses back to liquid and drips down — creating a passive, gravity-driven circulation loop.
The phase change (liquid to gas) absorbs significantly more heat per unit volume than single-phase convection — the latent heat of vaporisation provides an enormous thermal buffer. This makes two-phase particularly effective for extreme heat flux scenarios. However, the engineering requirements are more demanding:
- Vapour management: The tank must be sealed or well-contained to minimise fluid loss through evaporation.
- Fluid cost: Engineered two-phase fluids typically cost $50–$200+ per litre — an order of magnitude more than single-phase options.
- Environmental considerations: Some two-phase fluids have high global warming potential (GWP), though newer formulations are addressing this.
Performance Comparison
| Factor | Single-Phase | Two-Phase |
|---|---|---|
| Fluid cost | $2–$15 per litre | $50–$200+ per litre |
| Heat removal capacity | Good (convective) | Excellent (latent heat of vaporisation) |
| Pumping requirement | External pumps required | Passive (gravity-driven) or minimal pumping |
| Containment complexity | Standard (open-top tanks common) | Higher (vapour containment, sealed or semi-sealed) |
| Maintenance | Simpler — fluid testing, filter/pump service | More complex — vapour loss monitoring, fluid top-up |
| Maturity | 10+ years in production | Growing but fewer large-scale deployments |
| Typical rack density | 50–100 kW | 100–200+ kW |
| Environmental impact | Generally low GWP | Some fluids have high GWP; newer options improving |
When to Choose Single-Phase
Single-phase is the right starting point for most organisations. Choose it when:
- Your rack density is in the 20–100 kW range — well within single-phase capability
- Budget sensitivity is a factor — lower fluid cost and simpler infrastructure
- Your operations team is new to immersion cooling — simpler training and maintenance
- You want a proven, low-risk path — single-phase has the longest production track record
- You're retrofitting existing facilities — open-top tanks are easier to integrate
For most enterprise data centres, HPC environments, and even moderate AI deployments, single-phase delivers the thermal performance needed at a fraction of the complexity.
When to Choose Two-Phase
Two-phase makes sense in specific high-demand scenarios:
- Extreme density requirements (100+ kW per rack) where single-phase approaches its limits
- Space-constrained deployments where passive circulation eliminates the need for external pump infrastructure
- Environments where noise reduction is critical — no pumps, no fans, near-silent operation
- Organisations with the technical capability and budget to manage higher fluid costs and tighter containment
Two-phase is increasingly relevant for next-generation AI training clusters where per-GPU power consumption continues to climb. However, the higher fluid costs and operational complexity mean it's not the default choice for most deployments today.
Making the Decision
The choice between single-phase and two-phase isn't binary — it depends on your specific workload, facility, budget, and operational maturity. Key questions to ask:
- What is your current and projected rack density? If under 80 kW, single-phase is almost certainly sufficient.
- What is your fluid budget? Two-phase fluid costs can be 10–50x higher. Factor this into TCO.
- How mature is your operations team with liquid cooling? Start with single-phase if this is your first immersion deployment.
- What are your noise and space constraints? Two-phase's passive circulation has advantages in tight or noise-sensitive environments.
We help organisations evaluate both approaches during our audit phase. A structured comparison based on your actual workload data and facility constraints removes guesswork from the decision. Contact us to discuss your specific requirements.
Related: Immersion Cooling for AI Data Centres · The Business Case for Immersion Cooling
Frequently Asked Questions
Can I start with single-phase and upgrade to two-phase later?
In principle, yes — but the tanks and containment systems differ significantly. A more practical approach is to pilot single-phase, validate your operational processes, and deploy two-phase in new capacity if your density requirements grow. The two systems can coexist in the same facility.
Are two-phase fluids safe for the environment?
Some legacy two-phase fluids (particularly perfluorocarbons) have high global warming potential and are under regulatory scrutiny. Newer formulations from manufacturers like 3M (Novec series) and others are designed to have lower GWP. Always verify the environmental profile and regulatory status of your chosen fluid, especially for Australian compliance.
Which approach has better long-term economics?
For most deployments under 80 kW per rack, single-phase offers better long-term economics due to lower fluid costs and simpler maintenance. For extreme-density deployments, two-phase can be more economical per kW cooled despite higher fluid costs — because its superior heat transfer allows smaller, denser systems. The answer is workload-dependent.