Data centers and cooling: what are the alternatives to PFAS?

The growth of data centers dedicated to artificial intelligence is rapidly increasing the amount of heat that needs to be removed from chips and servers. With ever-increasing power densities, liquid cooling is taking on a growing role compared to traditional air cooling, but it also raises a new environmental issue: the choice of heat transfer fluids.

Indeed, some two-phase cooling technologies use fluorinated fluids belonging to the large family of PFAS , substances characterized by high environmental persistence. The potential European restriction of PFAS is therefore accelerating the search for alternatives capable of maintaining high thermal performance, electrical safety, and reliability.

For refrigeration designers and operators, this is a particularly interesting topic because not all liquid cooling technologies are equally dependent on PFAS . In some cases, fluorinated fluid-free solutions already exist; in others, especially in two-phase cooling, replacement still requires development and validation.

Single-phase liquid cooling: alternatives to PFAS already available

The first distinction concerns the way in which the fluid transfers heat.

In single-phase systems, both direct-to-chip and immersion, the liquid absorbs heat without changing state. These configurations can use PFAS-free fluids, including treated water, water-based mixtures, or alternative dielectric fluids, depending on the system architecture.

Direct-to-chip delivers coolant directly to the cold plates installed above the CPU and GPU, then transferring the energy to secondary circuits and heatsink systems. This solution is already one of the main directions adopted for managing high-density servers.

The situation becomes more complex in two-phase systems, in which the fluid evaporates, absorbing heat, and is then condensed to repeat the cycle. In these applications, fluorinated fluids have been used for their combination of electrical insulation, chemical stability, and controllable boiling temperature.

Two-phase cooling remains the main challenge

The advantage of two-phase cooling is the ability to remove large amounts of heat by exploiting the phase change of the fluid, a feature that becomes increasingly interesting as the thermal power of processors increases.

However, two-phase technologies still represent a limited share of the global data center market . Most infrastructures continue to use air cooling or single-phase liquid cooling solutions, both of which can be achieved without PFAS fluids.

Alternatives for two-phase systems are also emerging. Dielectric fluids declared PFAS-free are under development, as are new architectures that aim to maintain the benefits of phase transition while avoiding the use of persistent fluorinated substances.

The criticality doesn't just concern thermal performance. The fluid must simultaneously guarantee:

• compatibility with materials and electronic components;
• adequate dielectric strength;
• stability over time;
• correct evaporation temperature;
• limited corrosiveness;
• safety during maintenance and management of the system.

For this reason, the replacement must be evaluated on the basis of the entire system and not just by comparing the heat transfer capacity of the different fluids.

From direct-to-chip to direct-to-silicon: the cooling architecture is also changing

Research isn't just working on fluid replacement. A second direction involves completely rethinking the way heat is extracted from the chip .

Among the most advanced technologies are microfluidic systems in which the coolant is circulated through microchannels integrated directly into the silicon or in structures placed immediately in contact with the processor.

In these cases the distance between the heat source and the fluid is reduced to a minimum, increasing the dissipation capacity and opening up the possibility of also using deionized water or other non-fluorinated fluids .

This shift could also change the distribution of responsibilities in thermal design: some of the cooling could be integrated directly into the electronics, leaving the HVAC/R infrastructure to manage the secondary circuit and final heat rejection.

PFAS and data centers: sustainability enters the fluid choice

The European debate on PFAS therefore comes at a time when data center cooling is already undergoing a profound technological transformation.

The growth of AI loads makes it necessary to increase dissipation capacity, but at the same time there is growing attention towards energy consumption, water use and the environmental impact of the fluids used .

For the refrigeration sector, the challenge will therefore not be simply finding a replacement fluid. It will be necessary to evaluate the entire thermal architecture, choosing between air cooling, single-phase, two-phase systems, and new direct-to-silicon solutions based on power density and the actual needs of the infrastructure.

The direction, however, appears clear: the sustainability of data centers will increasingly depend on the choice of fluids and the ability to develop high-performance cooling systems with a lower environmental impact .