Why Dielectric Fluids Are Critical for Data Center Cooling

When a GPU rack crosses 100 kilowatts of power density, air cooling stops being an engineering challenge and starts being a losing battle. Fans can only move so much heat. Traditional cooling infrastructure — raised floors, precision air handlers, computer room air conditioning units — was designed for a world where servers drew 5 to 10 kilowatts per rack. That world is gone.

What's replaced it is a race toward liquid, and specifically toward a category of fluid that most data center operators hadn't thought much about until recently: dielectrics.

The core insight is deceptively simple — if you can submerge live electronics in a liquid that conducts heat but not electricity, you've solved your cooling problem at the source. That's exactly what dielectric fluids do. As immersion cooling moves from curiosity to infrastructure standard, understanding these fluids — what they're made of, how they perform, and what they cost you long-term — has become an operational necessity.

What Dielectric Fluids Actually Are

Strip away the chemistry, and the definition is straightforward. "A dielectric fluid is a liquid that conducts heat but does not conduct electricity," says Lucas Beran, director of product marketing at liquid cooling firm Accelsius. "Because it is electrically non-conductive, it can come into direct contact with live electronics without risking short circuits or equipment damage."

That non-conductivity is the entire ballgame. It separates immersion cooling from a very expensive disaster. You're not just keeping servers cool — you're bathing them in fluid, and that fluid has to be chemically inert enough that it won't destroy the boards, corrode the traces, or swell the gaskets it's touching 24 hours a day for years at a time.

Getting that chemistry right is harder than it sounds. As ZutaCore EVP Shahar Belkin explains, manufacturing involves tight chemical synthesis, rigorous removal of ionic contaminants, elimination of moisture, and strict quality control throughout filling and packaging. "Performance and safety depend heavily on purity and consistency, so supplier quality matters," he notes — which is a polite way of saying that cheap, poorly controlled fluid can quietly destroy expensive infrastructure.

The Two Camps: Petrochemical vs. Plant-Based

Walk into any serious conversation about dielectric fluids for data centers, and you'll find the market split along a fundamental axis: where the feedstock comes from.

On one side sit the petrochemical majors. HF Sinclair, Shell, Castrol, ExxonMobil, ENEOS, Gulf Oil, Petronas, and SK Enmove have all brought dielectric cooling fluids to market. Their path here wasn't accidental — many of these companies spent years formulating specialized lubricants and coolants for electric vehicles, developing deep expertise in managing heat in electrically sensitive environments. Pivoting that R&D toward data center immersion cooling was a logical extension. Castrol, for example, describes its immersion cooling products as synthetic hydrocarbons "specially formulated by our chemists and engineers to provide excellent performance."

Even more unconventional entrants are arriving. Infinium, a US developer of synthetic low-carbon efuels, announced in January 2025 that it was entering the data center sector with a cooling fluid produced by combining captured CO₂ with hydrogen — essentially redirecting its sustainable fuel production process toward a new market.

On the other side, plant-based alternatives are gaining serious traction. Cargill's NatureCool fluid is at least 90 percent soy oil-derived. Oleon — a subsidiary of French green chemicals firm Avril — launched its Qloe fluid in 2025, positioning it as fully biodegradable and non-toxic without sacrificing thermal performance.

The petrochemical camp wins on familiarity and supply chain scale; the plant-based camp wins on sustainability credentials and an increasingly favorable regulatory trajectory. Neither has definitively won the market, and smart operators are watching both closely.

What the Performance Numbers Actually Mean

Selecting a dielectric fluid isn't like buying a commodity. The metrics that separate good choices from expensive mistakes are specific and require some translation.

Thermal performance starts with two numbers: boiling point and latent heat of vaporization. These matter most for two-phase immersion systems, where the fluid boils off server components and then condenses back into liquid form in a heat exchanger above the tank. The efficiency of that phase-change cycle — how much heat the fluid absorbs per unit of mass as it vaporizes — determines how effectively you're actually moving thermal load out of your facility.

Then there's the sustainability dimension, which is no longer just a marketing consideration. Global Warming Potential (GWP) and Ozone Depletion Potential (ODP) are now genuine infrastructure decisions. Beran puts it bluntly: operators building infrastructure for the next 20 years need to choose fluids that won't face regulatory phase-outs. His benchmark: a GWP at or below 300 is considered low and effectively future-proofs the selection. For context, some older fluorinated fluids that dominated early immersion cooling deployments carry GWPs in the thousands — the kind of numbers that are already drawing regulatory scrutiny in Europe and increasingly in the US.

Equipment compatibility is the sleeper issue. A fluid that performs beautifully in a lab can quietly degrade the seals, gaskets, and polymer components inside a production tank over 18 months. Operators who skip long-term materials compatibility testing during procurement often discover the problem only after they've signed a multi-year contract with their fluid supplier.

Safety classification matters too, particularly for two-phase systems where vapor is present. Some dielectric fluids are certified non-flammable; others are classified as mildly flammable depending on operating conditions. That distinction has direct implications for facility permitting, insurance, and fire suppression requirements.

Choosing a Fluid: The Questions Operators Should Be Asking

Most procurement conversations about immersion cooling focus on the tanks, the heat exchangers, and the IT equipment. The fluid often gets treated as an afterthought — something to be sorted out later.

That's backwards. The fluid choice shapes everything downstream.

A few questions worth asking any supplier before signing: What is the fluid's viscosity profile across your operating temperature range, and how does that affect pump sizing and energy consumption? What materials compatibility data exists for the specific server hardware you're deploying? What is the fluid's track record in production deployments — not just lab testing? And critically, what does end-of-life fluid management look like, including disposal costs and environmental obligations?

The plant-based fluids have a genuine advantage on that last question. Biodegradability isn't just a sustainability checkbox — it translates to lower disposal costs and reduced liability when a fluid eventually needs to be replaced.

Beau Van Vaerenbergh of Oleon frames the feedstock choice as foundational: "The feedstocks generally fall into two categories: petrochemical-based versus naturally sourced." That binary shapes not just the environmental profile of the fluid, but its regulatory future, its supply chain resilience, and ultimately its total cost of ownership over a facility's operational lifetime.

Where This Market Is Going

The dielectric fluid market is young, fragmented, and moving fast. What's notable is who's entering it — not startups alone, but oil majors, food conglomerates, and synthetic fuel companies. That breadth signals that the opportunity is real and that no single feedstock or chemistry has locked up the market.

The underlying driver isn't going away. As AI infrastructure continues pushing rack densities higher — 200 kilowatt racks are already in production deployments, with 500 kW discussed as a near-term target — the physics of air cooling become more hostile by the quarter. Immersion cooling, and by extension dielectric fluids, graduates from niche solution to baseline requirement.

The operators who treat fluid selection as a strategic infrastructure decision — not a commodity purchase — will build facilities that are more efficient, more compliant, and better positioned for whatever regulatory and density requirements arrive next. The chemistry is settled enough to deploy at scale. The vendor landscape is competitive enough to demand scrutiny. The window to make good choices, before the industry consolidates around defaults, is open right now.

[INTERNAL LINK: dielectric fluids]

[INTERNAL LINK: immersion cooling]

[INTERNAL LINK: data center infrastructure]

EDITOR NOTES

• Consider cutting the paragraph discussing the petrochemical camp's familiarity and supply chain scale for brevity.
• The blog post would benefit from a more compelling CTA at the end, encouraging readers to explore the InfraSale Marketplace for dielectric fluid options.