In Episode 5 of the Splashcast podcast, Standard Fluids’ technical experts Steve Pignato and Kelvin Cabrera break down one of the most important distinctions in modern data center cooling technology. Single-phase and two-phase immersion cooling both involve submerging electronics in dielectric fluid. However, the fundamental physics of how they remove heat creates dramatically different performance characteristics and operational considerations.
Single-Phase Immersion Cooling: Conductive Heat Transfer
Single-phase immersion cooling keeps the dielectric fluid below its boiling point throughout the entire cooling cycle. The system circulates fluid past hot components, where the liquid absorbs heat through conduction. Warmed fluid then travels to a heat exchanger, releases its thermal energy, and returns to the tank to repeat the cycle.
This approach relies on the sensible heat capacity of the fluid. The amount of heat removed depends on the fluid’s specific heat capacity, the volume of fluid circulated, and the temperature differential between the fluid and the components. Increasing cooling capacity requires either more fluid flow or lower fluid temperature.
Many single-phase systems use synthetic hydrocarbons because they’re relatively inexpensive and provide adequate dielectric properties. However, these fluids typically have higher viscosity than engineered fluorochemicals. Higher viscosity means more pumping power is required to achieve necessary flow rates. The mechanical circulation system becomes a critical component that must run continuously for the cooling system to function.
Single-phase immersion cooling works well for moderate heat densities. It’s proven technology with a straightforward implementation path. Major deployments exist in China and other markets where cost considerations favor this approach. The basic principles of single-phase cooling remain sound for appropriate applications.
Two-Phase Immersion Cooling: Phase Change Physics
Two-phase immersion cooling operates on entirely different physics. The dielectric fluid is selected specifically for its boiling point, which is engineered to match the thermal design point of the hardware. When fluid contacts a hot chip surface, it boils and transitions from liquid to vapor.
This phase change is where the dramatic performance advantage appears. The latent heat of vaporization allows the fluid to absorb enormous amounts of thermal energy without increasing temperature. For comparison, raising water temperature by one degree Celsius requires 4.18 kJ/kg of energy. Converting water from liquid to vapor at the same temperature requires 2,260 kJ/kg. The phase change absorbs roughly 540 times more energy than a one-degree temperature increase.
Engineered fluids like Standard Fluids™ SF 649™ Engineered Fluid provide similar phase-change performance tailored specifically for electronics cooling. The vapor rises naturally due to buoyancy, condenses on cooling coils at the top of the tank, and returns as liquid through gravity. The system can operate passively with minimal or zero pumping within the tank itself.
The heat transfer coefficient during nucleate boiling reaches 10,000-100,000 W/m²·K. Single-phase convection typically delivers 500-5,000 W/m²·K. This order of magnitude difference in heat transfer efficiency explains why two-phase cooling can handle chip densities that overwhelm single-phase systems. Research from IEEE on boiling heat transfer validates these performance characteristics.
Practical Implications for Data Centers
The choice between single-phase and two-phase involves multiple considerations beyond pure thermal performance.
Single-phase systems offer simplicity in fluid selection and potentially lower upfront fluid costs. They work with a wider range of dielectric liquids. However, they require continuous pumping and may struggle with the highest-density AI accelerators coming to market.
Two-phase systems demand more sophisticated tank design and precise fluid selection. The fluids typically cost more per liter. However, they provide superior cooling capacity, passive operation capability, and better scalability for future chip generations. The reduced reliance on mechanical circulation also means fewer potential failure points.
As Steve Pignato explains in the Splashcast episode, two-phase immersion cooling leverages physics that engineers have understood for decades. The technology simply waited for hardware heat densities to reach levels where the advantages become compelling. We’ve reached that point with modern AI workloads.
The Path Forward
The data center industry is witnessing parallel development of both approaches. Single-phase deployments continue in applications where moderate heat densities and cost sensitivity drive decision-making. Two-phase systems are increasingly deployed where maximum thermal performance and future scalability matter most.
Standard Fluids provides engineered fluids for both approaches. Our SF 649 fluid for two-phase cooling and SF 188 fluid for single-phase applications reflect our understanding that different use cases demand different solutions. The technical consultation and fluid purity standards we provide ensure optimal performance regardless of which approach customers choose.
The conversation in Splashcast Episode 5 makes clear that understanding these technical differences helps operators make informed decisions. Both technologies have legitimate applications. The key is matching the cooling approach to the specific thermal requirements, operational constraints, and future growth plans of each data center deployment.
As Kelvin Cabrera notes, choosing correctly for the correct time matters. Some solutions work better now. Some will work better as chip densities continue increasing. The physics of phase change suggests that two-phase immersion cooling will become increasingly advantageous as we move deeper into the AI era.