Dielectric Fluid Specs Checked Against OCP Immersion Limits

Engineered Fluids lists its ElectroCool EC-120 at 781.9 kg/m³ and 2212 J/kg·K at 40°C. Carrying 100 kW across a 10 K fluid rise therefore takes 100,000 W ÷ (781.9 kg/m³ × 2212 J/kg·K × 10 K) = 0.00578 m³/s, or 5.78 L/s.

The rows on a dielectric fluid sheet that decide whether a non conductive coolant may fill a tank of live servers, such as breakdown voltage, flash point and resistivity, each hang on a test condition the headline number tends to drop. This piece reads three published sheets against the Open Compute Project (OCP) immersion requirements row by row, then sets out the status of the fluorinated option under per- and polyfluoroalkyl substance (PFAS) rules as of 2026-10-09.

PAO, PFPE and EC-120 carry heat per liter within 4.5% of each other

On a mass basis the fluorinated fluid looks weak. Syensqo’s Galden HT135, a perfluoropolyether (PFPE), is listed at a specific heat of 0.23 cal/g·°C. The brochure prints no reference temperature for that row or for any other row of its property table.

Converting with 1 cal = 4.184 J gives 0.23 × 4184 = 962 J/kg·K, so EC-120 carries 2212 ÷ 962 = 2.30 times more heat per kilogram per kelvin.

Source: Syensqo, “Galden HT PFPE Heat Transfer Fluids”

Density reverses most of that gap. The brochure lists HT135 at 1.72 g/cm³. ExxonMobil’s Data Center Immersion Fluid 3152, a polyalphaolefin (PAO), is listed at 0.768 g/ml at 60°C.

The ExxonMobil sheet gives a specific heat of 2.21 at 60°C and prints the unit as J/kg/°C. Read literally, that would be about one thousandth of the EC-120 value, so the arithmetic below takes it as kJ/kg·K.

Source: ExxonMobil, “ExxonMobil Data Center Immersion Fluid 3152” product datasheet

Multiplying density by specific heat gives the heat one cubic meter carries per kelvin:

• EC-120 at 40°C: 781.9 kg/m³ × 2212 J/kg·K = 1.730 MJ/m³·K
• PAO 3152 at 60°C: 768 kg/m³ × 2210 J/kg·K = 1.697 MJ/m³·K
• HT135, no temperature stated: 1720 kg/m³ × 962 J/kg·K = 1.655 MJ/m³·K

The spread is (1.730 − 1.655) ÷ 1.655 = 4.5%. EC-120 and PAO 3152 sit at the temperatures their sheets report and the Galden brochure reports none, so this compares published points, not three fluids in one bath.

For the 100 kW, 10 K case in the opening, HT135 needs 100,000 ÷ (1,655,190 × 10) = 0.00604 m³/s, or 6.04 L/s, against 5.78 L/s for EC-120. A pump and heat exchanger sized in liters per second barely notices the change of family.

Tank weight does notice. That 6.04 L/s of HT135 is 10.39 kg/s of fluid, against 4.52 kg/s of EC-120, and per liter PFPE weighs 1720 ÷ 781.9 = 2.2 times as much. OCP’s immersion requirements list a full-solution static load among the items for public specification sheets, and the fluid choice moves that figure far more than it moves the flow rate.

Viscosity is where the three sheets part ways

OCP’s Rev 2.10 requirements, last updated in August 2023, set out the immersion fluid properties a supplier must publish, including kinematic viscosity at six temperatures, and add a footnote for converting it to dynamic viscosity. The footnote reads: Dynamic viscosity (cP) = Kinematic viscosity (mm²/s) × Density (kg/m³).

Run as printed with EC-120 at 40°C, it returns 16.25 × 781.9 = 12,706 cP. Carried through in consistent units, the result is 16.25 mm²/s × 0.7819 g/cm³ = 12.71 mPa·s, which equals 12.71 cP. The footnote works only with density in g/cm³; a spreadsheet built from it as written overstates every fluid by a factor of 1,000.

Source: Open Compute Project, “Immersion Requirements Rev 2.10”

With the units fixed, the sheets separate. The EC-120 and PAO 3152 sheets both give values at 60°C:

• EC-120: 8.62 cSt × 0.7685 g/cm³ = 6.62 mPa·s
• PAO 3152: 3.2 cSt × 0.768 g/ml = 2.46 mPa·s

At the same temperature, EC-120 is 6.62 ÷ 2.46 = 2.7 times more viscous.

HT135 is listed at 1.00 cSt, or 1.72 mPa·s with the brochure’s density, and no temperature is printed beside it. It cannot be set against the 60°C rows above, and since viscosity changes with temperature, any gap at a shared bath temperature remains unknown from these three documents.

OCP attaches a cost to low viscosity: a footnote in Rev 2.10 states that any hydrocarbon at or below 20.5 mm²/s at 40°C should be classified as Category 1 of aspiration hazards.

PAO 3152 is listed at 5.0 cSt at 40°C, well under that line. The low viscosity that helps the pump is the same property that puts this hydrocarbon into Category 1, and OCP’s liquid management rules then require that every operator be trained on the properties of each dielectric fluid in use.

EC-120 also sits under the line at 40°C, but its sheet does not name its base chemistry, so whether the footnote applies cannot be settled from the sheet alone.

Four breakdown figures on three sheets rest on four electrode setups

Rev 2.10 asks for dielectric strength by ASTM D1816, measured at 1 mm and stated in kV/mm. Its amendment table records that the option to estimate that value from a 2.5 mm measurement was removed.

The lifetime minimum is 6 kV/mm. The unused-fluid column for this row is blank, so OCP sets no breakdown floor for fluid as delivered.

The three sheets answer that request three ways, and EC-120 adds a fourth figure from another test.

Table 1. Dielectric breakdown as each document states it, sorted by stated value (kV or kV/mm, as printed)

* Geometry from ASTM’s public summary of D3300 (most common setup); the EC-120 list does not state its own. ** Simple division; Rev 2.10 removed estimating the 1 mm value from a 2.5 mm test. “>” marks a lower bound. Typical values, not guaranteed specifications.

Source: Open Compute Project, “Immersion Requirements” Rev 2.0 (2022) and Rev 2.10 (2023); Engineered Fluids, “Statement of Compliance: OCP Rev2.1, ElectroCool EC-120” (2025 and 2026); ExxonMobil, “Data Center Immersion Fluid 3152” datasheet (2026); Syensqo, “Galden HT PFPE Heat Transfer Fluids” (undated); ASTM International, D3300-20 summary

EC-120’s >25 kV/mm at 1 mm by ASTM D1816 is the only figure in the format Rev 2.10 asks for. Against the 6 kV/mm lifetime floor it leaves a margin of more than four times on a new fluid.

Galden HT135 is listed at 40 kV across a 2.54 mm gap. Dividing gives 15.7 kV/mm, a conversion Rev 2.10 no longer accepts, and the brochure does not name the test method.

PAO 3152 is listed at >10.0 kV at 25°C by ASTM D1816 with no electrode gap stated. Without the gap, there is no way to turn kilovolts into kilovolts per millimeter, so this figure cannot be placed against the OCP floor.

The 205 kV on EC-120’s extended list comes from ASTM D3300, an impulse test. ASTM’s public summary describes it as most commonly run with a negative polarity needle or sharp point against a grounded sphere, a highly divergent field. The figure is valid for its own test and cannot sit on the same axis as a D1816 result.

ASTM’s summary of D1816 adds two limits that apply to every row. A high breakdown voltage does not necessarily indicate the absence of all contaminants, and field results are comparable with laboratory results only when all criteria, including a room temperature of 20 to 30°C, are met.

Source: ASTM International, “D1816-12(2019) Standard Test Method for Dielectric Breakdown Voltage of Insulating Liquids Using VDE Electrodes”

The scope in that summary names petroleum oils, silicone fluids, high fire-point mineral oils and esters. PAO and fluorinated fluids are not on the list.

The 150°C flash point floor moved to a closed cup, and one compliance sheet reads 144°C

OCP’s Rev 2.0 of January 2022 set the flash point floor at 150°C on a “COC” basis, with ASTM D92 and ISO 2592 listed as the methods. Rev 2.10 keeps 150°C for unused and lifetime fluid but changes the reference to “Cleveland Closed Cup”, ASTM D93 and ISO 2719.

The label mixes two tests. UL Solutions’ page on immersion fluid testing lists ASTM D93 as the Pensky-Martens closed cup method and ASTM D92 as the Cleveland open cup method. A lab receiving the Rev 2.10 table will run the method number, so the requirement in practice is a closed cup result of at least 150°C.

Source: UL Solutions, “Immersion Cooling Fluids Testing and Certification”

ExxonMobil’s sheets still report the open cup. PAO 3152 is listed at 157°C on a COC basis by ASTM D92. The sales specification for the related grade 3151, effective October 2023, guarantees a minimum of 150°C, also COC by ASTM D92.

Engineered Fluids’ compliance statement for EC-120 does use the closed cup. Its table prints the OCP minimum as ≥150°C by ASTM D93 / ISO 2719 and the EC-120 value as 144°C. The opening paragraph of the same document states that EC-120 fully complies with OCP Rev 2.1 (2023-08-18), the same August 2023 document.

Source: Engineered Fluids, “Statement of Compliance: OCP Rev2.1, ElectroCool EC-120 Dielectric Coolant”

The two figures sit side by side on one page, and the document carries no note on that row. The gap is 6 K. This is the row I would put back to the supplier in writing, with a request for the batch certificate of analysis, before a tank is filled.

The open cup figures raise a second question. A 157°C result on the D92 open cup is not a D93 closed cup result, so the PAO 3152 sheet as published does not show whether that grade clears the Rev 2.10 floor.

Rev 2.10 also raised the auto ignition point minimum from 250°C to 300°C, citing IEC 62368-1. EC-120 lists 340°C. The Galden brochure states that the HT range has no flash point, no fire point and no auto-ignition point, so the flash point row has no value to compare for that family.

ExxonMobil’s guaranteed specification lists six properties, and none is electrical

The PAO 3152 datasheet states that its typical properties are not to be construed as specifications. The 3151 sales specification lists what ExxonMobil does guarantee, each reported on the certificate of analysis: specific gravity, appearance, color, kinematic viscosity at 40°C, pour point and flash point.

Neither dielectric strength nor resistivity is among them. Engineered Fluids describes its EC-120 values as typical figures based on its history of quality control testing and offers a certificate of analysis per batch. For both suppliers, the electrical rows arrive as typical values unless the purchase terms say otherwise.

OCP sets electrical floors for unused and for lifetime fluid separately. For volume resistivity, the unused minimum is 2.00 GΩ·m and the lifetime minimum is 0.20 GΩ·m, a tenfold allowance for decline in service.

PAO 3152 is listed at >2.00E+11 ohm·cm at 25°C. The conversion is 2.00 × 10¹¹ Ω·cm × 0.01 m/cm = 2.00 × 10⁹ Ω·m = 2.00 GΩ·m, which lands exactly on the unused floor.

Galden HT135 is listed at 1.5 × 10¹⁵ Ω·cm, which converts the same way to 15,000 GΩ·m. EC-120 is listed at 1 × 10⁵ GΩ·m.

Source: Open Compute Project, “Immersion Requirements Rev 2.10”

A typical value printed at the floor gives no room for a batch that tests lower. OCP’s quality management table lists particles, moisture, sludge and plasticizers from server components as contaminants, so the lifetime column is where a fluid program is judged. Rev 2.10 places the duty on suppliers to name the tests and intervals that monitor fluid quality, which makes those intervals a contract item rather than a data sheet item, as with the operating assumptions behind other coolant sheets.

3M’s PFAS exit is complete, while law-firm summaries put the EU restriction at committee stage

OCP names three liquid groups for immersion: synthetic hydrocarbons, esters and fluorochemicals. Its acidity limit for unused fluorocarbons is 0.001 mg KOH/g, a tenth of the 0.01 mg KOH/g set for hydrocarbons, and two-phase systems must carry a conditioning system able to remove water, oils, acids and thermal decomposition products. Those rows matter most in two-phase tanks where the fluid boils at the chip.

Whether a fluorochemical counts as PFAS depends on the definition in use. A 2022 iScience paper by Hammel and coauthors quotes the OECD’s 2021 definition, which covers substances with at least one fully fluorinated methyl or methylene carbon, and notes that it still includes carbon-fluorine units joined through oxygen. A PFPE backbone is built that way, so a fluid such as Galden falls inside the OECD definition on its chemistry name alone.

Source: Hammel et al., “Implications of PFAS definitions using fluorinated pharmaceuticals,” iScience (2022)

One part of the supply picture is settled. 3M’s annual report for 2025 states that the company completed its exit of PFAS manufacturing at the end of 2025. Its report for the third quarter of that year had listed engineered fluorinated fluids among the products made from the shorter-chain PFAS compounds it was still producing until then.

Source: 3M Company, Form 10-K for fiscal year 2025

The regulatory part is not settled. As of October 2026, the EU timeline here rests on two law-firm summaries from March 2026, because ECHA’s own pages could not be read for this piece. According to Covington & Burling, ECHA’s Risk Assessment Committee adopted its opinion on the EU-wide restriction on March 2, 2026.

The Socio-Economic Analysis Committee’s draft opinion followed on March 26, with consultation open until May 25. Covington and Arnold & Porter both expect that committee’s final opinion by the end of 2026.

Arnold & Porter describes the proposal as covering a class estimated at more than 10,000 substances, with derogations that are mostly time-limited. Covington expects no Commission adoption before the third quarter of 2027 and no application before 2029, a law firm estimate rather than a published date.

Neither summary names immersion cooling among the derogated uses, and no later committee outcome appeared in the sources opened for this piece.

A fluorinated fluid selected in 2026 is therefore a purchase from a supply base that 3M, a maker of engineered fluorinated fluids, has left, under an EU proposal that, according to those law-firm summaries, had no final committee opinion or Commission text. For a tank expected to run past 2029, regulatory status needs the same scrutiny in design review as breakdown voltage, and qualified engineering review is required before any of these figures is used for a safety or design decision.

The sheets agree on heat per liter, within 4.5% across three fluid families, and diverge on nearly every row that depends on a test setup: the viscosity conversion, the electrode gap, the flash point cup and the line between typical and guaranteed values. One compliance statement lists 144°C beside a 150°C OCP minimum in its own table. For fluorinated fluids, the manufacturing exit of one major producer is complete, while the EU restriction, as of October 2026 and per law-firm summaries, is a set of committee opinions without a Commission text.

References

• Immersion Requirements Rev 2.10 / Open Compute Project / August 18, 2023 / https://www.opencompute.org/documents/ocp-acs-immersion-requirements-rev-2-1-pdf
• Immersion Requirements Revision 2.0 / Open Compute Project / January 21, 2022 / https://opencompute.org/documents/ocp-acs-immersion-requirements-rev-2-v1-00-pdf
• Statement of Compliance: OCP Rev2.1, ElectroCool EC-120 Dielectric Coolant (with Extended Characteristics List) / Engineered Fluids / April 15, 2026 / https://www.engineeredfluids.com/wp-content/uploads/2026/06/EC-120-OCP-Compliance.pdf
• ExxonMobil Data Center Immersion Fluid 3152, Product Datasheet / ExxonMobil / revision date February 12, 2026 / https://www.exxonmobilchemical.com/en/chemicals/webapi/dps/v1/datasheets/150000104833/0/en
• ExxonMobil Data Center Immersion Fluid 3151, Sales Specification / ExxonMobil / effective October 10, 2023 / https://www.exxonmobilchemical.com/en/chemicals/webapi/dps/v1/datasheets/130000108780/1/en
• Galden HT PFPE Heat Transfer Fluids brochure / Syensqo / undated / https://content.syensqo.com/galden-HT-PFPE-heat-transfer-fluids-v2.2
• D1816-12(2019) Standard Test Method for Dielectric Breakdown Voltage of Insulating Liquids Using VDE Electrodes, public summary / ASTM International / 2019 / https://www.astm.org/d1816-12r19.html
• D3300-20 Standard Test Method for Dielectric Breakdown Voltage of Insulating Liquids Under Impulse Conditions, public summary / ASTM International / 2020 / https://www.astm.org/d3300-20.html
• Immersion Cooling Fluids Testing and Certification / UL Solutions / undated / https://www.ul.com/services/immersion-cooling-fluids-testing
• Form 10-K for fiscal year ended December 31, 2025 / 3M Company / 2026 / https://www.sec.gov/Archives/edgar/data/66740/000006674026000014/mmm-20251231.htm
• Form 10-Q for quarter ended September 30, 2025 / 3M Company / 2025 / https://www.sec.gov/Archives/edgar/data/66740/000006674025000089/mmm-20250930.htm
• Implications of PFAS definitions using fluorinated pharmaceuticals / Hammel, Webster, Gurney, Heiger-Bernays, iScience 25(4) / March 2022 / https://pmc.ncbi.nlm.nih.gov/articles/PMC8933701/
• ECHA Launches a New Public Consultation on a Proposed Universal Ban on PFAS in the EU / Covington & Burling / March 2026 / https://www.cov.com/en/news-and-insights/insights/2026/03/echa-launches-a-new-public-consultation-on-a-proposed-universal-ban-on-pfas-in-the-eu
• ECHA Committees Advance Broad PFAS Restriction Under REACH / Arnold & Porter / March 2026 / https://www.arnoldporter.com/en/perspectives/advisories/2026/03/echa-committees-advance-broad-pfas-restriction-under-reach