What Is PG25 Coolant? The Spec, Freeze Point and Who Uses It

We blend PG25 in Taylor, Texas, so we see the parts of this fluid that a definition skips. The 25 can mean two different numbers. A 275 gallon tote holds 68.75 gallons of glycol and 46.5 pounds of inhibitor, and the gallons do not add up the way you would expect. A refractometer can be a full point of glycol off from a shift too small to notice. This guide covers the spec, the math behind the 25, and what to ask before you fill a loop with anyone's fluid, ours included.

What is PG25 coolant?

PG25, also written PG 25 or PG-25, stands for propylene glycol at 25%. The Open Compute Project (OCP) defines it as "a heat transfer fluid containing approximately 25% propylene glycol by volume, together with water and a formulated corrosion-inhibitor package." In OCP's guideline PG25 means the inhibited, ready-to-use fluid. Plain glycol and water is not PG25 in that sense.

It runs in the technology cooling system (TCS), the closed loop between the CDU and the cold plates on the chips. The CDU hands the heat to the facility water through a heat exchanger. In Dell's words, the two networks "do not mix." OCP expects these loops to run at about 17 to 66 °C (63 to 151 °F).

Glycol helps control biological growth and lowers the freeze point. Water carries the heat. Inhibitors protect the metals and get used up doing it: OCP says they form a protective layer "which consumes the inhibitor gradually."

PG25 spec and datasheet values

OCP's guideline (Revision 2.0, September 2026) is the closest thing the industry has to a shared PG25 datasheet. It lists one set of numbers for new fluid and a looser set for fluid in service. OCP calls these representative values, not specifications, so your equipment maker's limits come first.

Source: OCP Rev 2.0, Table 1 (pp. 14-16), Table 10 (p. 43), Table 14 (pp. 67-70).

PG25 freezing point

New PG25 freezes at 9 to 14 °F (-13 to -10 °C) on OCP's table. Dell lists -10 °C. In service, OCP says 24 to 29% glycol freezes between 9 and 15 °F (-13 and -9 °C).

Burst point is a different number. Below the freeze point the fluid turns to slush, and a system that sits idle can ride that out. Dow gives burst protection to 0 °F (-18 °C) at 25.1% by volume of its fluid, a typical figure for one product. A loop that has to keep running needs freeze protection.

OCP notes glycol also protects "in shipping, storage and use scenarios," so this matters for totes in transit too. Dow advises a freeze point at least 5 °F (3 °C) below your lowest expected ambient. For any other strength, our glycol freeze point calculator gives the freeze and burst point at any percent.

Why 25 percent?

The 25 comes from biology. Freeze protection comes along with it. The sources agree on the floor:

• Dell: 25% glycol "is most common because it is the minimum required to achieve adequate biofouling prevention."
• Dow's glycol guide: "Solutions of glycol less than 25% may be at risk for bacterial contamination." Dow also warns that inhibitor levels below 25 to 30% glycol may not protect against corrosion adequately.
• Motivair by Schneider Electric: at 10%, its thermal engineer writes, "algae and other biomatter tend to feed on PG."

"Bio-growth may also occur if the propylene glycol concentration is not maintained at ~25% or above."OCP Guidelines, Revision 2.0, p. 60

Growth matters because the passages are tiny. Motivair puts cold plate fin spacing at around 50 microns today, and says any growth there can hurt cooling performance.

So why not run more glycol? Dell's answer: viscosity rises and thermal conductivity falls, which means lower cooling capacity and higher pump energy. PG25 is the least glycol that clears the floor, which leaves little room below it. That is why dilution is the risk to watch.

PG25 vs water: how much more flow?

For the same heat load and temperature rise, PG25 needs about 4% more flow than water. Viscosity is the larger penalty. PG25 values below are from Dell's white paper, water from NIST.

Worked example: a 100 kW load with a 10 °C rise across the loop.

That is 5.7 L/min more, or 4%. Dell rounds its specific heat, so read it as 3 to 5%. Recochem's property table for its OAT PG-25 agrees: 4.3% at 20 °C and 3.2% at 40 °C against NIST water.

Viscosity is the real cost. Recochem's table shows 2.55 cP at 20 °C and 1.42 cP at 40 °C, against 1.00 and 0.65 cP for water, so the fluid stays about twice as thick when warm. In a laminar passage, pressure drop scales with viscosity times flow, about 2.5 times water's at 20 °C. Treat that as a rough upper bound, since turbulent sections feel viscosity less, and rerun it with your own fluid's values before you size a pump. Motivair's engineer sums up the trade: PG25 "performs nearly as well as DI water but with far less fluid maintenance."

By volume or by weight?

PG25 is a by-volume name: OCP and Dell both state glycol by volume. Not everyone does. Supermicro's datasheet lists "Propylene glycol 25 wt%," and Recochem's sheet lists 25.0% by weight minimum. Dow's table shows how the bases line up:

Source: Dow DOWFROST Technical Data Sheet, pp. 2-3. Concentrate is Dow's fluid at 96% glycol by weight. Typical values.

The two numbers are close because propylene glycol's density is near water's. The difference still matters more at 25% than at 50%:

• The gap is bigger. On Dow's table it is 0.6 points at 25% and 0.1 point at 50%.
• The room is smaller. Near 50% nothing depends on a few tenths of a point. Near 25% you are on the biofouling floor. A fluid at exactly 25.0% by weight is 24.4% by volume, just under the 24.5% where OCP's new-fluid range starts.

The concentrate column is a third trap: 25% of a concentrate is not 25% glycol, because the concentrate is part additive. And never borrow an ethylene glycol chart. There, 25% by weight is 22.9% by volume.

How a 275 gallon tote of PG25 is made

Here is our actual batch rule. Glycol gallons equal the target percent of the batch. The inhibitor is 2.0% of the batch weight. Water is the balance. We use a specific gravity of 1.038 for propylene glycol, 1.165 for the inhibitor, and 8.345 lb/gal for water.

1. Our 25% by volume is 25.6% by weight. Glycol is heavier than water.
2. The water is 73%, not 75%. OCP counts inhibitors and water together as about 75% of the fluid.
3. Gallons do not stack. The sheet adds volumes straight across, which implies 8.45 lb/gal. Our calculated density for the finished fluid is 1.0221 g/cm³ at 20 °C, or 8.53 lb/gal. Glycol and water shrink slightly when mixed, by about 1%, close to 3 gallons on a tote. Our spec records the formula by batch weights, which avoids the problem.

Water is 72% of the weight, so water quality sets the ion numbers. OCP says high chloride, sulfate or hardness may indicate poor-quality water, and that softened water "will not eliminate chloride and sulfate."

How to check PG25 with a refractometer

A refractometer is the fast field check for glycol, and it has a trap. OCP warns that "inhibited formulations, such as those containing corrosion inhibitors and additives, can alter the refractive index, meaning the correlation between refractive index and freeze point may vary from one manufacturer to another."

The math shows how little it takes. On Dow's table, refractive index at 22 °C goes from 1.3615 at 24.4% by volume to 1.3675 at 29.4%. That is 0.0012 per point of glycol. If an inhibitor package shifts the reading by 0.0012, your meter shows one extra point. One point is the distance from OCP's 24% in-service floor to 25%.

The wrong scale is worse. A reading of 1.3615 is 25% propylene glycol by weight, freezing at 14.7 °F. On Dow's ethylene glycol table, about the same reading is 28% glycol, freezing at 9.2 °F. Read PG25 on an EG scale and you credit the loop with 5.5 °F of protection it does not have.

1. Get the refractive index table for that exact fluid from its maker. OCP's example of 1.362 to 1.368 at 20 °C for 24 to 29% depends on the formulation.
2. Use a digital refractometer with temperature compensation.
3. Keep a sample of the fresh fluid, as OCP advises, and read it on your own meter. Trend against that number.
4. Use it for concentration only. OCP says it is "not a direct measure of inhibitor health," and it cannot give an accurate concentration if the loop is badly contaminated.

What a PG25 COA should show

OCP tells operators to keep batch numbers, delivery dates and Certificates of Analysis for every system. A useful PG25 COA has these lines:

• Lot number and date, matching the container label
• Glycol percent, with the basis (volume or weight) and method stated
• Freeze point and method (ASTM D1177 or D3321)
• pH (ASTM D1287) and reserve alkalinity (ASTM D1121)
• Appearance
• Chloride, sulfate and total hardness, the water quality lines
• Inhibitor technology

Ask how each line was produced. A COA line can be a lab measurement or a calculation from the batch sheet. Our sample COA labels which is which: inhibitor and nitrite come from batch weights, freeze point cites ASTM D1177.

PG25 vs DI water, PG55 and ethylene glycol

DI water transfers heat better, and Lenovo recommends it for its GB300 NVL72 rack. It needs more care. Vertiv specifies it "treated with suitable corrosion inhibitors and biocides," and Motivair says the upkeep can significantly increase costs.

Ethylene glycol is less viscous and generally transfers heat better, Dow says, but PG is chosen where toxicity matters for its "low acute oral toxicity vs. the moderate toxicity of ethylene glycols." Motivair says EG is seldom used in data centers because quick disconnects can leave a slightly wet break during servicing. More in our PG vs EG toxicity guide.

Who uses PG25?

Castrol's release shows what a validation covers: corrosion resistance, thermal stability, dilution-water quality, glycol purity and wetted-material compatibility. A fluid can be PG25 by strength and still not be approved for your system.

PG25 inhibitors: OAT, nitrite and silicate

PG25 names a strength. The inhibitor package is the real difference between one PG25 and another: OAT (organic acid technology), IAT (inorganic acid technology) or a hybrid. Our OAT vs NOAT vs HOAT guide goes deeper. OCP's wording on specific chemistries sits in its section on flush cleaning solutions, and it is worth reading exactly:

• Silicates: "Silicate-based additives should be avoided as these additives can form insoluble deposits on metal surfaces which are difficult to remove."
• Nitrites and molybdates: "Certain inhibitors and additives, such as chromates, nitrites, nitrates and molybdates, may present environmental, regulatory or disposal considerations depending on local requirements and site-specific waste handling practices."
• Amines: "Amines may also warrant consideration, as some can react with nitrites to form nitrosamines under certain conditions."

That is a caution, not a ban. OCP says use "should be evaluated in the context of applicable regulations and facility capabilities." Some fluids avoid the question: Recochem's sheet says its OAT PG-25 contains no nitrite, amine or silicate. Others, including ours, contain nitrite. Your equipment spec decides.

Flushing, filling and filtering a PG25 loop

Check wetted materials first. OCP's starter lists include copper, brass (under 15% zinc preferred), 304L and 316L stainless, titanium Grade 2, peroxide-cured EPDM, FKM, PTFE, polypropylene, HDPE and PEEK. Sulfur-cured EPDM can add sulfate over time. Quick connects should be stainless. Aluminum is not on the metals list, and carbon steel is "not recommended."

Flush clean. OCP's flush water limits are under 25 µS/cm, chloride, sulfate and hardness each under 25 ppm, and bacteria under 100 CFU/mL. Keep biocides and surfactants out of the final flush, and fill within 24 hours if you can.

Watch leftover water. Motivair says concentration can change significantly on the first pass through the CDU as the fluid meets residual flush water. The math is short:

Four percent leftover water takes a 25.0% fill to OCP's 24% in-service floor. A 500 gallon loop at 23.75% needs about 20 gallons drained and replaced with PG55 to get back to 25%. Only correct with a concentrate your fluid supplier approves.

Filter. OCP says even new premix can carry particulates and highly recommends 5 to 10 micron absolute filtration during fill. In service, sidestream filtration diverts about 5 to 15% of flow, and OCP strongly recommends it above 250 gallons. Overly fine filtration can remove antifoam.

Start up and sample. Verify concentration after fill and circulate for at least 30 minutes while you check for leaks. Take a sample in the first 1 to 3 days and keep it for the life of the fluid, along with a sample of fresh fluid. Our twelve failure modes guide shows how to use that baseline.

How often to test PG25

OCP's floor: every TCS loop gets a qualified lab test at least once a year. Beyond that, it scales with risk.

Source: OCP Rev 2.0, Table 13 (pp. 63-64). After 1.5 years of stable results, medium may drop to yearly and high to every 6 months. Loops under 100 gallons can be sampled as a subset of racks.

The annual lab panel covers glycol, pH, reserve alkalinity, appearance, inhibitor concentration, chloride, sulfate, total hardness, degradation acids and dissolved metals. Pull 4 to 6 oz from well-mixed circulating fluid, never from an expansion tank or dead leg. If an on-site test comes back out of range, OCP says to get a sample to a lab within a week.

Advice varies, so check your maker's terms. Dell's paper says OCP recommends quarterly on-site testing, citing OCP's 2022 edition. Motivair suggests daily on a new system, then weekly, then about monthly. Some makers tie it to warranty terms. Our ASTM D1121 guide covers the alkalinity side.

Store unopened containers dry and out of direct sun, and skip galvanized tanks. Never pour fluid down a drain. OCP says most used PG fluids can go to glycol recyclers.

Questions to ask your coolant supplier

1. Is your glycol figure by volume or by weight?
2. What is the inhibitor technology? Does it contain nitrite, amine, silicate, molybdate or phosphate?
3. Has my equipment maker approved this fluid, in writing?
4. Do you have ASTM D8040 corrosion data for copper, brass and stainless? OCP's limits are 0.2 mpy for copper and brass, 0.1 mpy for 304L and 316L.
5. Can I have your refractive index table for this exact fluid?
6. What water do you blend with? OCP's dilution water limit is under 25 µS/cm, with chloride and sulfate under 10 ppm.
7. Do you filter at fill? OCP says 50 microns or finer, with 25 or finer preferred.
8. What does your COA report, and which lines are measured?
9. Which concentrate or additive do you approve for corrections, and which lab reads the samples?

Our PG25: what it is and what it is not

We blend Inhibited Propylene Glycol 25/75 with OAT Inhibitor in Taylor, Texas: 25.0 to 26.0% propylene glycol by volume in deionized water, freeze point 14.0 °F (-10.0 °C) max, typical pH 7.0 to 9.0, inhibitor at 2.0% by weight, not dyed. It comes in sizes from 1 quart to a 330 gallon tote.

Need the Certificate of Analysis for your lot? Just ask, no charge. Here are the technical data sheet and the safety data sheet. For other strengths, try the coolant finder or our data center cooling chemicals page.