Data Center Liquid Cooling Gaskets, Seals & TIM by Zone | H-O Products

Home › Blog › Telecom & Data Centers › Liquid cooling gaskets, seals & TIM

Telecom & Data Centers · Materials Guide

Data Center Liquid Cooling Gaskets, Seals and TIM: Material Selection by Zone

EPDM, fluorosilicone, FKM and FFKM on each side of the CDU, and why ASHRAE's liquid-cooling guidance still leaves the material spec to you.

820x360|

Every liquid-cooled rack adds wetted joints that never existed in an air-cooled hall: CDU door and panel gaskets, manifold flange seals, cold-plate thermal interfaces, insulated facility piping. Racks that once pulled single-digit kilowatts are now pulling 60 to 120 kW, and the chip roadmaps behind them are not slowing down. ASHRAE Technical Committee 9.9, the group that has set data center thermal guidance since 2004, has spent the last two years acknowledging that its liquid-cooling guidance was written for hardware that no longer exists.

A new framework is in development. Until it arrives, the gap between what the standard says and what your rack needs is being closed by individual facility teams, and by the sealing, insulation, and thermal interface materials specified into their cooling loops. Here is the short version of what changed, what didn't, and what it means for the parts you buy.

Request a quote →

What ASHRAE TC 9.9 changed, and what it hasn't

ASHRAE's air-cooled classes (A1 through A4) are mature and most facilities still design to them. Liquid cooling has its own facility-water classes, now labeled by supply temperature (W17, W27, W32, W40, W45, and W+; formerly W1 through W5). Those classes are still valid as a starting reference.

What changed is the hardware. Accelerator packages now draw hundreds of watts to more than a kilowatt per socket, and their tolerance for temperature excursions shrinks as power climbs. TC 9.9's recent communications point to two consequences:

  • A brief supply-temperature spike that older servers absorbed without notice can now trigger compute throttling in seconds.
  • The same excursion can damage hardware, not just slow it. TC 9.9 is calling for more thermal mass in the loop and active primary-to-backup cooling redundancy.

What TC 9.9 has not finished is a single prescriptive liquid-cooling template with defined supply and return ranges, coolant chemistry limits, and pressure specs for this generation of hardware. In the meantime, chip vendors are publishing their own coolant supply targets, commonly 25 to 45°C for single-phase direct liquid cooling, with no shared ASHRAE reference to reconcile them.

So the "tightening" is about process (more thermal inertia, active redundancy, monitored coolant quality), not a new number you can look up. The actual temperature and chemistry targets are being set by whoever is closest to the rack.

Diagram showing the coolant distribution unit as the boundary between the facility water system and the technology cooling system, with sealing and insulation materials labeled by zone
Where the material decisions sit. The CDU is the boundary TC 9.9 designs around: upstream is the facility water system (FWS), downstream is the technology cooling system (TCS). Fluid, pressure, and failure consequence all differ by zone, so the materials do too.

Why a guidance gap becomes a material problem

When a standard fully defines the environment, picking a gasket or seal is a lookup: choose a family whose service window covers the spec with margin. Air-cooled data centers have worked that way for a decade.

High-density liquid cooling does not offer that shortcut yet, for three reasons:

  1. Coolant chemistry is not standardized. Glycol/water blends are common on the technology side, but concentration, pH, and corrosion-inhibitor packages vary by facility and vendor. The inhibitor package, not the base glycol, is usually what attacks an elastomer over its service life.
  2. Temperature and pressure targets are still moving. A loop commissioned against this year's accelerator may need to run cooler, or at a different pressure, for the next generation. A gasket specified with no margin gets reopened at the next refresh.
  3. Leak consequence is higher. An air-side seal failure is an efficiency problem. A wet-side failure at the rack, manifold, or CDU is a hardware and uptime problem. That raises the bar on compression-set retention and validated adhesive bonding at every wetted joint.
The one detail that saves a revision. State the coolant concentration and the inhibitor package, not just the fluid family. "Propylene glycol/water" is not enough to run a compatibility check, and the inhibitor package is usually what attacks the elastomer.

Coolant distribution unit gaskets: which side of the CDU is the joint on?

The one thing TC 9.9's guidance is consistent about is the demarcation point: the CDU separates the facility water system from the technology cooling system, and the two sides see different conditions.

Facility water (upstream of the CDU): larger volume, often building chilled water, lower pressure cycling, longer runs. This is where pipe insulation and moderate-duty EPDM sealing live.

Technology cooling (downstream of the CDU): smaller volume, tighter chemistry, higher pressure and cycle rate at the rack and manifold. This is where fluorosilicone, FKM, and validated adhesive-backed gasketing earn their cost.

A single elastomer family rarely covers both zones well. The first question to answer about any joint is which side of the CDU it sits on.

Liquid cooling gasket and TIM materials by zone

Zone / dutyWhat the material has to surviveTypical family
FWS piping insulationCondensation control at dew point, long unattended serviceClosed-cell elastomeric foam; aerogel blanket on chilled runs
FWS static seals, CDU door and panel gasketsModerate pressure cycling, dry to occasional weepEPDM, silicone sponge
TCS manifold flange and rack-level gasketsGlycol/water blend, high cycle rate, thermal cyclingFluorosilicone, FKM sheet
TCS cold plate and chip interfaceDirect thermal path, compression set under clamp loadGap-filler pad, TIM film, graphite spreader
Aggressive-chemistry or immersion zonesDielectric fluid, aggressive inhibitor packagesFFKM sheet; PTFE and FEP fluoropolymer films
Leak-adjacent EMI seams (rack, CDU enclosure)Continuity under vibration, humidity, occasional moistureNickel-graphite or conductive-fabric-over-foam gasket

This is a starting point, not a spec. The right call in each row still depends on the actual glycol concentration, pressure, and duty cycle at that joint. The zone-by-zone material reference lives on our Data Center Cooling & Liquid Infrastructure page.

EPDM vs fluorosilicone vs FKM vs FFKM for glycol and immersion loops

Most liquid cooling seal questions come down to four elastomer families. The ranges below are typical published values for sheet and sponge grades H-O converts; the grade you specify is confirmed against the maker's current technical data sheet and your fluid's actual concentration and inhibitor package.

FamilyTypical service rangeWater / glycol coolantHydrocarbon and dielectric fluids (immersion)Where it fits in the loop
EPDM (sponge or solid)About -50 to 150 °CGood. The default for water and glycol on the facility side; confirm the inhibitor package.Poor. Swells in oils and hydrocarbon dielectrics.FWS flanges, CDU doors and panels, dry-side enclosure gaskets
Silicone spongeAbout -55 to 200 °CFair. Fine for panel and door gaskets that see occasional weep, not for long immersion.Poor.CDU cabinet and rack panel gaskets where flame rating or temperature range matters
FluorosiliconeAbout -60 to 175 °CGood.Good. Resists fuels, oils and many dielectric fluids.TCS manifold flange gaskets; wide-temperature wetted joints
FKM / Viton® sheetAbout -20 to 200 °C (low-temperature grades lower)Grade-dependent. Standard grades are fair in hot water and glycol; specialty grades are good. Check the compound.Excellent.Aggressive inhibitor packages, immersion tanks, high-temperature technology-side joints
FFKM sheetAbout -15 to 300 °CExcellent.Excellent. Broadest chemical resistance of the four.Where nothing else survives the chemistry; cost limits it to the joints that need it

Two things the table cannot tell you: which side of the CDU the joint is on, and what the inhibitor package in your specific coolant does to the compound over years, not weeks. Those are the two inputs we ask for before recommending a family.

Two claims to be careful about

"This material is ASHRAE-compliant." There is no ASHRAE certification for a gasket, seal, or TIM. TC 9.9 publishes design guidance for facilities and equipment; it does not test or list sealing materials the way UL does for flammability. A supplier can say a material's service range covers a given W-class. A supplier cannot say the material is "ASHRAE compliant."
"This elastomer is glycol-compatible." Compatible with what concentration, what inhibitor package, at what temperature, for how long? A generic resistance-chart entry is the start of a conversation, not a substitute for checking your fluid's actual formulation against the compound's swell and hardness-shift data.

How H-O Products helps

H-O does not write thermal guidance or test coolant chemistry. We are a die-cutter and converter of the gasket, seal, insulation, and thermal interface materials that get specified into a cooling loop once those engineering decisions are made. Concretely:

  • Cross-referencing your actual coolant chemistry and concentration against elastomer compatibility data before you commit to a family
  • Matching compression-set and closure-force requirements at CDU and manifold joints to real gasket classes (ASTM D1056) instead of a nominal durometer
  • Converting FWS and TCS parts to the tolerances liquid loops require, die-cut, waterjet, or CNC knife-cut, with the lot traceability data center customers typically ask for
  • Supplying TIM options (gap-filler pad, film, graphite) thickness-matched to your actual clamp load and gap, not a catalog default

Three questions to ask now

  1. Which side of the CDU is this joint on? If that takes more than one sentence, the spec underneath it needs a second look.
  2. What is the exact coolant chemistry and concentration? "Propylene glycol/water" is not a complete answer for a compatibility check.
  3. Does the spec have margin for the next hardware refresh? Given how fast supply-temperature targets have moved, a spec with no margin will be revisited within 18 months.

Have a liquid-cooling material stack reviewed

Send your coolant chemistry, concentration, pressure, and joint drawings. We will tell you plainly whether your current spec has margin, recommend a family per zone, and convert samples on your geometry.

Samples typically ship in 3 to 5 business days for common die-cut configurations; standard production runs about two weeks; special orders are quoted to the job.

Family-owned in Winsted, Connecticut since 1971 · ISO 9001:2015 certified · 1,200+ customers worldwide send us a drawing and get back their part.

Related reading

Frequently asked questions

Does ASHRAE certify gasket and seal materials for liquid cooling?

No. TC 9.9 publishes thermal design guidance for facilities and equipment; it does not test or list individual sealing, insulation, or thermal interface materials. Flammability and dielectric certifications come from separate bodies such as UL. A supplier can state that a material's service range covers a given W-class, but "ASHRAE compliant" is not a rating ASHRAE issues.

Is there a published ASHRAE liquid-cooling class we should design to?

The facility-water classes (W17 through W45 and W+) exist and remain a reasonable reference for facility-water-side design. Guidance specific to the power densities and tight excursion tolerances of current AI accelerators is still in development, which is why chip vendors are publishing their own supply-temperature targets in the meantime.

Our coolant vendor says the fluid is compatible with EPDM. Is that enough to spec EPDM everywhere?

Not on its own. Confirm the concentration and inhibitor package that compatibility claim was tested against, and confirm it matches what is actually running in your loop today. Formulations get adjusted, and a claim tested at one concentration does not automatically extend to another or to the other side of the CDU.

Is EPDM compatible with propylene glycol coolant?

Generally yes. EPDM is the default elastomer for water and propylene-glycol or ethylene-glycol coolant on the facility-water side, and most generic compatibility charts rate it well. What the charts do not cover is the corrosion-inhibitor package blended into a data center coolant, which is what actually attacks the compound over a service life. Confirm the specific inhibitor package and concentration against the EPDM grade's swell and hardness-shift data before specifying it downstream of the CDU.

EPDM or FKM for a coolant distribution unit (CDU) gasket?

It depends on the joint. CDU door, panel and facility-side flange gaskets that see water or glycol at moderate temperature are usually EPDM or silicone sponge. Technology-side manifold flanges with tighter chemistry and higher cycle rates move to fluorosilicone or FKM sheet, and dielectric-fluid or immersion joints move to FKM or FFKM. Both are die-cut from sheet; the fluid and the temperature decide which one.

How often should we revisit cooling-loop material specs?

Given the pace of the last two accelerator hardware generations, a review at every major hardware refresh is a reasonable default until TC 9.9's liquid-cooling thermal template is finalized and widely adopted.

Request a quote

Request a quote for your data center cooling materials

Send H-O your coolant chemistry, pressure, and joint drawings. Engineering reviews every RFQ, recommends a family per zone, and converts samples on your geometry.

Contact
Company address
Your application
Part & quantity
Typical response in one business day. Samples typically 3–5 business days; production in about 2 weeks.

Last updated · Reviewed by H-O Products engineering (ISO 9001:2015 certified converter, Winsted, CT).

References

  • ASHRAE TC 9.9. Thermal Guidelines for Data Processing Environments, 5th ed. (2021). ASHRAE Datacom Series. Defines the A1 through A4 air-cooled classes and the W17 through W45 and W+ liquid-cooling classes.
  • ASHRAE TC 9.9. Liquid Cooling Guidelines for Datacom Equipment Centers, 2nd ed. (2014). ASHRAE Datacom Series.
  • ASHRAE TC 9.9. Liquid Cooling: Resiliency Guidance for Cold Plate Deployments, technical bulletin (September 2024).
  • ASTM D1056, Standard Specification for Flexible Cellular Materials: Sponge or Expanded Rubber. ASTM International.
  • Manufacturer technical data sheets for the named material families; grade-specific values, fluid-compatibility data, and compression-set figures are confirmed on the maker's current TDS.

Standard summaries in this article are informational and reflect publicly available ASHRAE TC 9.9 communications as of September 2026. ASHRAE's liquid-cooling thermal template for high-density accelerated computing was still in development at time of writing; confirm current published guidance and your own mechanical engineer's design basis before finalizing a material spec.

H-O Products is a precision die-cut converter of sealing, insulation, and thermal interface materials, not a raw-material producer; we do not write thermal design standards or test coolant chemistry. Elastomer compatibility with a specific coolant formulation is confirmed against the material maker's current technical data sheet and your fluid's actual concentration and inhibitor package.

H-O Products is an ISO 9001:2015 certified organization; that certification covers our quality management system, not individual materials or parts. Grade-specific values are confirmed against the manufacturer's current technical data sheet and an engineering drawing review. All parts are custom manufactured to order and minimum production run quantities apply.

H-O Products Corp · Precision die-cut converting · Winsted, CT · ISO 9001:2015 certified organization