Post-Novec Immersion Fluid Alternatives in 2026

Three years ago, a single corporate decision changed the trajectory of every active two-phase immersion cooling program in the world. 3M's December 2022 announcement that it would exit all PFAS-related manufacturing by the end of 2025 did not come from a change in market outlook — it came from the same regulatory mathematics now reshaping the entire fluorinated fluid industry. For data center operators who built immersion infrastructure around Novec 7000-series fluids, 2026 is not a transition year; it is a reckoning. Fluid procurement pipelines that assumed continuous supply from a single dominant manufacturer are now being rebuilt against a backdrop of incomplete alternatives, contested regulatory timelines, and two-phase research that is promising but not yet deployable at scale.

The Novec Exit, in Context

3M's Novec line — comprising the 7000 (boiling point 34°C), 7100 (61°C), 7200 (76°C), and 649 (49°C) series — was purpose-engineered for electronics cooling and held an effective monopoly in commercial two-phase immersion deployments from roughly 2018 through 2022. The fluids combined low global warming potential relative to legacy HFCs, non-flammability, electrical inertness, and precisely tunable boiling points that allowed thermal engineers to match condensation curves to chip TDP profiles. Hyperscale operators in Asia-Pacific and Europe, plus domestic U.S. colocation providers pursuing GPU density, qualified their hardware and mechanical infrastructure around these fluids. Some operators invested tens of millions of dollars in Novec-optimized immersion tanks before 3M's announcement.

The announcement itself was framed carefully. 3M did not assert that the fluids were unsafe or non-compliant at the moment of the announcement. Instead, the company cited the "trajectory of PFAS regulations globally" and made a business decision to exit entirely rather than manage fluid-by-fluid regulatory status across dozens of jurisdictions. That framing proved prescient. Through 2024 and 2025, PFAS regulatory pressure accelerated faster than even 3M's internal modeling had predicted.

Two regulatory drivers are most consequential for the data center industry. First, the U.S. Environmental Protection Agency finalized its TSCA Section 8(a)(7) PFAS reporting rule in October 2023, requiring manufacturers, importers, and processors of PFAS-containing substances to report production volumes, use, and exposure data to the EPA. The reporting window opened January 2026 and creates an immediate compliance burden for any operator still holding fluorinated immersion fluid inventory. Second, in February 2023, five European regulatory authorities — Germany, the Netherlands, Denmark, Sweden, and Norway — submitted a universal PFAS restriction proposal under EU REACH covering approximately 10,000 PFAS substances, explicitly including the perfluorocarbon and hydrofluorocarbon chemistries used in Novec-class fluids. A regulatory decision is anticipated in the 2026–2027 timeframe. Norwegian regulators implemented PFAS restrictions on certain non-essential uses as early as mid-2023.

The practical consequence for hyperscale operators who depended on Novec is unambiguous. 3M wound down Novec production in 2025 as committed. Authorized distributor stock exists in the secondary market, but supply chain integrity cannot be guaranteed, pricing has become opportunistic, and any operator restocking Novec today is acquiring a regulatory liability, not building a procurement strategy. The fluid's lifecycle is complete.

The Replacement Landscape

Three distinct categories of replacement options have emerged since 2022, and none of them is a clean answer. Understanding what each category actually delivers — versus what its proponents claim — is the first task for any procurement team re-evaluating a two-phase program.

Category A: Engineered Fluorinated Alternatives

Solvay's Galden PFPE series (principally the HT55 and HT70 grades), AGC's Asahiklin fluorinated solvent line, and residual Chemours Opteon SF products represent the most immediately available engineered alternatives. Galden HT55 offers a 55°C boiling point and Galden HT70 a 70°C boiling point — both within range for high-TDP server workloads. Dielectric strength exceeds 40 kV/mm across the Galden HT line. From a pure thermal performance standpoint, these fluids are competent alternatives to Novec 7100 and 7200.

The problem is that these are also PFAS. Galden PFPE is based on perfluoropolyether chemistry — a distinct molecular structure from 3M's perfluorocarbon compounds, but still within the broad PFAS classification that regulators are targeting. The EU universal PFAS restriction proposal does not carve out PFPEs. AGC's Asahiklin products similarly contain fluorinated compounds within PFAS definitions. Operators who qualify these fluids today are deferring regulatory risk, not eliminating it. Availability is currently better than Novec's end-of-life situation, but both Solvay and AGC face the same regulatory trajectory that drove 3M's exit decision. Treating Galden or Asahiklin as a permanent solution misreads the regulatory environment.

Category B: Hydrocarbon-Based Two-Phase Research Fluids

Several U.S. and Asian research groups are actively developing hydrocarbon-based dielectric fluids capable of two-phase behavior in the 40–70°C range. These would offer zero PFAS exposure, lower raw material cost, and a simpler regulatory compliance path in both U.S. and EU jurisdictions. Early laboratory results published in 2024 and 2025 are technically credible. However, as of mid-2026, no hydrocarbon two-phase fluid has achieved commercial production qualification at data center scale. This category represents a legitimate future direction — not a present procurement option — and no deployment plan should depend on fluids without an existing production supply chain.

Category C: Reformulated Single-Phase and Non-Immersion Systems

The third category abandons the two-phase architecture entirely. Operators in this group are moving immersion capacity to single-phase immersion (mineral oil, synthetic polyalphaolefins, or engineered single-phase dielectrics) or pulling back from immersion altogether in favor of direct liquid cooling to the chip. Direct-to-chip systems bypass the fluid volatility problem by eliminating the phase-change mechanism. The thermal efficiency tradeoff relative to two-phase is real but manageable at current chip TDP levels of 350–1,000 W per package, depending on the hardware platform.

Why "Drop-In Replacement" Is a Myth

The phrase "drop-in replacement" has circulated in data center cooling discussions since 3M's announcement, and it is almost always wrong. No fluid currently available on the commercial market is a true drop-in replacement for Novec 7000-series in an existing two-phase immersion system. The reasons are thermodynamic, not merely chemical.

Two-phase immersion systems are engineered around a specific thermodynamic cycle. The boiling point of the working fluid determines the saturation temperature inside the tank, which sets the temperature gradient between the chip junction and the condensing surface. The condenser — whether a shell-and-tube unit, a plate heat exchanger, or an integrated liquid-to-air assembly — is sized for a specific heat load, a specific saturation temperature, and a specific condensation ΔT relative to the facility's cooling water supply temperature. Change the fluid, and every one of those parameters moves.

Consider the most common substitution scenario: replacing Novec 7100 (boiling point 61°C) with Galden HT70 (boiling point 70°C). That is a 9°C shift in saturation temperature. In a system where the cooling water supply temperature is 25°C, the Novec 7100 system operates with a condensation ΔT of 36°C at the condenser face. The Galden HT70 system operates with a condensation ΔT of 45°C — which means more favorable heat transfer per unit area at the condenser, but the chip junction temperature has risen by approximately 9°C. For high-TDP GPUs operating at or near their thermal design limit, that delta pushes junction temperature beyond the envelope without active derating or flow rate increases that were not specified in the original system design.

The quantified impact of a smaller substitution is still significant. Thermal engineering analysis of two-phase systems shows that a 4°C shift in boiling point propagates to approximately 8% change in required condenser surface area at constant heat rejection load, derived from the heat exchanger duty relationship Q = U × A × ΔT_LM where ΔT_LM changes in proportion to the boiling point offset. An 8% area change in a large data center deployment — representing thousands of square meters of condenser surface — is a capital re-engineering event, not a purchase order.

Tank pressure margins are the second constraint that breaks the drop-in narrative. Novec tanks are designed with pressure relief specifications matched to the vapor pressure curve of the qualified fluid. A fluid with a higher boiling point has a lower vapor pressure at operating temperature, altering the pressure differential across tank seals and vent systems. A fluid with a different enthalpy of vaporization changes how much latent heat the condenser must absorb per kilogram of vapor. Neither adjustment is trivially made without re-qualifying the mechanical containment system under the facility's pressure vessel standards.

Finally, chip-surface thermal interface behavior is fluid-specific. Nucleate boiling — the mechanism that makes two-phase cooling highly effective at the chip level — depends on a fluid's surface tension, contact angle with chip substrate materials, and nucleation characteristics on copper and nickel-coated heat spreaders. Substituting a chemically distinct fluid changes all of these parameters and requires re-characterization of the boiling curve for each hardware configuration. Major OEMs with immersion-qualified server platforms do not automatically extend that qualification to alternative fluids. A separate validation program, typically costing six to eighteen months of engineering time, is required before a new fluid can be used under OEM warranty terms.

Hydrocarbon Two-Phase Research: Current State

The most intellectually compelling direction in post-Novec fluid development is the pursuit of hydrocarbon-based two-phase dielectrics — fluids with no PFAS content, no fluorine chemistry, and boiling points engineered into the 40–70°C range required for electronics cooling. If this class of fluid succeeds technically, it resolves the regulatory problem permanently while potentially reducing fluid cost by an order of magnitude relative to Novec. The research is real, active, and producing credible results. The gap between research results and deployable product is equally real and is currently measured in years, not months.

Active research programs exist at university laboratories in the United States — primarily at Texas A&M, Georgia Tech, and the Purdue Cooling Technology Research Center — at national laboratory affiliates, and at several Asian institutions, including research groups affiliated with major semiconductor manufacturers in South Korea and Taiwan who have strong incentives to qualify PFAS-free cooling solutions ahead of anticipated regulatory deadlines in their own markets. Published work from 2024 and 2025 has demonstrated proof-of-concept two-phase behavior in modified hydrocarbon formulations at boiling points near 58°C, with dielectric strength measurements in the 25–32 kV/mm range under controlled laboratory conditions.

The technical barriers are structural rather than incidental. Two-phase cooling fluids must satisfy a set of properties that are partially in tension. High dielectric strength requires minimal free charge carriers, which favors low-polarity molecules. Low boiling points in the 40–70°C range are achievable in lighter hydrocarbon fractions, but light hydrocarbons have higher flammability — and most data center facility standards require immersion fluid flash points above 130°C. Adequate vapor density for efficient condensation requires molecular weight distributions that conflict with the boiling point targets. Tuning these parameters simultaneously, without fluorine chemistry to anchor the vapor pressure curve independent of flammability, requires molecular engineering that is substantially more difficult in pure hydrocarbon systems than in fluorinated compounds.

Specifically: achieving a flash point above 130°C while maintaining a boiling point at or below 65°C in a single-component or stable blended hydrocarbon system is not currently achievable with known commercial chemistry. Research groups are exploring synthetic formulations — blended systems, functionalized cycloalkanes, and proprietary molecular structures that modify the vapor-pressure curve relative to the flash point curve. Some results are promising. None has cleared the full qualification gauntlet: extended dielectric stability testing over thousands of hours, materials compatibility verification across copper, nickel, PTFE, and standard PCB laminate substrates, and heat-aged fluid characterization simulating a 3-to-5-year service life.

The realistic commercial production timeline for a deployable hydrocarbon two-phase fluid — one with a published specification sheet, an available supply chain, and at least one major OEM hardware qualification — is 18 to 36 months from mid-2026, assuming current research trajectories hold and a willing chemical manufacturer commits to scaling a candidate formulation from lab to production. That timeline is optimistic. It assumes no fundamental chemistry obstacle emerges in extended stability testing, that a manufacturer commits production capacity before the fluid has an established customer base, and that OEM qualification programs proceed on an accelerated schedule. Infrastructure decisions made in 2026 cannot responsibly depend on this timeline being met.

Single-Phase as the Practical 2026 Immersion Path

Given the binary reality of the 2026 fluid market — fluorinated alternatives with unresolved regulatory exposure, or hydrocarbon two-phase fluids that do not yet exist as commercial products — the case for single-phase immersion as the practical near-term choice has become structurally compelling. This is not a consolation-prize argument. For operators deploying immersion in 2026, single-phase is the choice that holds up across supply, regulatory, and total-cost-of-ownership dimensions simultaneously.

Single-phase immersion fluids — primarily mineral oil, synthetic polyalphaolefins (PAO), and purpose-formulated single-phase dielectrics — have mature global supply chains with no dependency on specialty chemical manufacturers. Mineral oil meeting ASTM D3487 Type I or Type II specifications for transformer insulating oil is available from multiple commodity suppliers in North America, Europe, and Asia at $3–$12 per liter. Synthetic PAO-based dielectrics suitable for electronics immersion run $12–$18 per liter but offer tighter viscosity specifications, better low-temperature pumpability, and extended service life. Neither product category contains PFAS. Neither is subject to EU REACH universal restriction proposals. Neither triggers TSCA Section 8(a)(7) reporting obligations.

The thermal performance tradeoff relative to two-phase is real and should be stated directly. Single-phase immersion does not achieve the near-zero chip-to-coolant thermal resistance of optimized two-phase systems. Published data from deployed single-phase installations show chip junction temperatures running 10–18°C higher than equivalent two-phase configurations at the same chip TDP, depending on fluid flow rate, tank geometry, and chip package design. For 350W-class CPUs, this delta is manageable within standard thermal design margins. For 700W–1,000W accelerator packages, the thermal headroom calculation requires engineering analysis specific to the hardware platform and the facility's coolant supply temperature.

The counterargument is direct: a single-phase system that can be deployed in 2026 with a qualified, regulatory-clean fluid from a stable supply chain delivers more operational value than a two-phase system that cannot be deployed because the fluid no longer has a reliable manufacturer. Single-phase also eliminates the pressure management systems, vapor-tight sealing requirements, and condenser sizing constraints that make two-phase mechanical infrastructure substantially more expensive at the capital level. Total cost of ownership analysis for 36-month deployment cycles generally shows single-phase as cost-competitive with two-phase once fluid procurement uncertainty, PFAS compliance costs, and potential re-validation costs for alternative fluorinated fluids are included in the model.

"If you're committing capital to immersion cooling in 2026, single-phase mineral oil or a synthetic PAO is the only bet I can defend to a procurement committee. The two-phase fluorinated fluid market is structurally broken right now, and the hydrocarbon two-phase alternatives are not ready for production deployment. Single-phase is not a compromise — it is the rational choice given actual fluid availability and the regulatory trajectory of every fluorinated option on the market." — Andre Taki, Director of Products & Sales & Cooling Chemistry Practice Leader, Alliance Chemical

Disposal and Legacy Inventory

For operators who already hold Novec inventory — whether in active systems or in stockpiled drums — fluid disposal is a compliance event, not an operational afterthought. The regulatory status of Novec as a PFAS-classified substance creates specific disposal pathway requirements that differ substantially from standard industrial fluid waste handling. Operators who route fluorinated immersion fluid through conventional industrial waste channels face potential EPA enforcement exposure under both TSCA and CERCLA remediation liability frameworks.

3M established a Novec take-back program concurrent with its manufacturing exit announcement, administered through 3M's Environmental Health & Safety and product stewardship groups. The program accepts drums of spent or unused Novec fluid for return and controlled destruction. Operators with existing 3M customer relationships and documented purchase histories should initiate the take-back process through their 3M account representative or the 3M product stewardship team directly. As of mid-2026, the take-back program remains operational for legacy customers with documented purchase records, but capacity is not unlimited. Operators should not assume this option will remain available indefinitely and should initiate conversations now rather than waiting until disposal becomes operationally urgent.

For smaller volumes — typically sub-200-liter quantities held by colocation operators, OEM test labs, or research institutions — specialty PFAS recyclers and licensed hazardous waste processors are the appropriate disposal channel. Companies including Clean Harbors, Veolia North America, and several regional hazardous waste processors with specific PFAS handling permits can accept spent fluorinated dielectric fluid. Costs for specialty PFAS disposal run $8–$25 per liter depending on volume, contamination level, and geographic location. At these rates, disposal cost is a non-trivial line item against fluid values of $45–$90/L, and it should have been — and going forward must be — included in total lifecycle cost accounting for any fluorinated fluid procurement.

Two disposal pathways should be explicitly prohibited in operator site procedures. Standard used-oil recycling programs are not equipped to handle PFAS-classified materials and are prohibited by permit conditions in many jurisdictions from accepting fluorinated fluids. Municipal hazardous waste collection programs similarly lack the thermal destruction or advanced treatment capabilities required to destroy PFAS compounds. Routing fluorinated immersion fluid through either channel creates downstream environmental liability and, under EPA enforcement interpretations, potential operator liability for remediation costs if the fluid reaches a municipal waste stream or groundwater pathway.

Operations teams should treat disposal planning as a mandatory component of fluid procurement decisions going forward — not just for fluorinated fluids but for any immersion fluid with limited or specialized disposal pathways. Any contract for fluorinated alternatives from Solvay or AGC should include explicit documentation of the approved disposal pathway and estimated per-liter disposal cost as line items in the total cost of ownership model. Facilities that qualified Novec without disposal planning are now learning this lesson at above-market disposal rates under time pressure. That experience should not be repeated with the next generation of fluid decisions.

What Hyperscalers Are Actually Doing

Market intelligence from operators in hyperscale, colocation, and enterprise immersion segments through the first half of 2026 reveals a picture more pragmatic, and more conservative, than public narratives from tank manufacturers and fluid suppliers would suggest. Three anonymized operator postures illustrate the range of strategic responses currently in execution.

A major North American hyperscaler with an active two-phase immersion pilot — approximately 3 MW deployed across three data center campuses — announced internally in Q1 2026 a suspension of all greenfield two-phase expansion pending resolution of the fluid supply chain. The operator has not dismantled existing two-phase installations, which continue to operate on remaining Novec inventory supplemented by limited Galden HT70 procurement. New capacity additions planned for 2026 and 2027 are being redirected to single-phase immersion and direct-to-chip configurations. The cooling engineering team concluded internally that the total cost of two-phase — including fluid cost, supply risk premium, re-validation costs for alternative fluids, and potential PFAS compliance costs — has exceeded the thermal efficiency premium that two-phase provides at current accelerator TDP profiles. The decision was framed as temporary, pending maturation of the hydrocarbon two-phase market, but with no committed re-evaluation date.

A European colocation operator with GPU-dense AI inference capacity has moved in the opposite architectural direction: it has formally qualified a PAO-based single-phase fluid as its new immersion standard across all future AI pod deployments. The operator negotiated a multi-year volume supply agreement with a European synthetic lubricant producer — not a specialty cooling fluid supplier — pricing the PAO at commodity-adjacent rates and eliminating the specialty chemical supply risk that characterized Novec procurement. The operator's chief engineer characterized the decision internally as "the last boring choice we're making about immersion fluid for several years while the two-phase market figures itself out." That framing captures a broader sentiment observable across mid-market operators: the goal is to make a defensible procurement decision and move on, not to optimize at the margins of a fluid market that is structurally broken.

A third operator — a large Asia-Pacific cloud provider with significant in-house hardware development capability — has shelved immersion entirely for its next-generation accelerator rack program and is deploying an internally designed direct-to-chip liquid cooling system at scale. The decision was driven primarily by hardware integration requirements for its proprietary AI accelerator architecture, not by fluid market conditions, but the PFAS regulatory environment was cited as a confirming factor in the decision not to invest in immersion infrastructure qualification for a new platform.

For procurement teams making 2026 decisions, these narratives converge on a set of concrete recommendations. Do not acquire Novec from secondary distributors unless you have a specific operational continuity need and a documented, permitted disposal plan — the regulatory and supply chain risks outweigh the short-term availability benefit in almost all scenarios. If two-phase capability is a hard architectural requirement for your thermal design, engage Solvay or AGC directly for Galden PFPE supply and require written commitments on supply continuity through your system's planned service life, plus an explicit statement of the fluid's regulatory status under EU REACH and EPA TSCA as of the contract date. Treat those written commitments as material representations in the procurement contract. Evaluate rigorously whether your specific GPU or CPU workload thermal profiles actually require two-phase efficiency margins, or whether a well-engineered single-phase system at adequate flow rates stays within thermal design envelopes — for most H100-class AI inference and training deployments, single-phase immersion is thermally viable. And build disposal cost into every fluid procurement model from this point forward; it is no longer an afterthought that can be deferred to end of life.

The post-Novec landscape is not a temporary market disruption. It is a permanent restructuring of the dielectric immersion fluid industry, driven by regulatory forces with no reversal mechanism in the medium term. Operators who plan 2026 and 2027 immersion deployments around that structural reality — rather than the assumption that a like-for-like fluorinated replacement will emerge — will have procurement strategies that hold up over a three-to-five-year infrastructure commitment horizon. Hydrocarbon two-phase chemistry is a legitimate future direction and deserves continued research investment; the operators who are tracking that research closely today will have an advantage when deployable products eventually reach the market. But infrastructure decisions made in 2026 must be grounded in what is available, qualified, and regulatory-clean now. The fluid market will evolve. The capital committed to this year's immersion deployments will not.

For deeper technical context on specific fluid chemistries, thermal modeling approaches for single-phase and two-phase architectures, and Alliance Chemical's supply capabilities for immersion-grade dielectrics, explore the related sub-pillars in this series — including our guides to mineral oil specifications for electronics immersion, PAO selection criteria for high-TDP GPU environments, and the PFAS compliance primer for data center operators — or return to the AI & Data Center Cooling Chemicals pillar for the full scope of resources available from Alliance Chemical's cooling chemistry practice.