Water Quality Requirements for Data Center Liquid Cooling Systems: pH, Conductivity, and Beyond

As liquid cooling moves from specialized high-performance computing into mainstream AI, cloud, colocation, and enterprise environments, water quality is becoming a reliability requirement—not an afterthought. Cold plates, coolant distribution units, rear-door heat exchangers, manifolds, quick connects, pumps, valves, hoses, seals, and heat exchangers all depend on a stable fluid chemistry profile. If that profile drifts, the consequences can include corrosion, biofouling, scale, plugged microchannels, reduced heat transfer, higher pump energy, nuisance alarms, premature component replacement, and avoidable downtime.

For facilities teams, the challenge is that “good data center water” is not a single universal specification. Requirements differ by loop type, wetted materials, coolant formulation, OEM warranty language, operating temperature, makeup water source, and risk tolerance. From our perspective as a manufacturer of liquid analysis sensors, the most successful programs treat water quality for data centers as a controlled operating parameter: define the target, measure the right variables, trend the data, and respond before chemistry becomes a cooling problem.

Why water quality matters more in liquid-cooled data centers

Liquid cooling brings water or a water-based heat transfer fluid closer to IT equipment. That proximity improves heat removal, especially where air cooling struggles with high-density racks and increasing chip power. ASHRAE has noted that liquid-only processor chips are already available and that future data center designs should include the capability to add liquid cooling, while the Open Compute Project describes warm-water cooling as an effective option as power density and cooling challenges increase. (ashrae.org)

The same efficiency advantage also reduces the margin for poor water chemistry. A chilled-water plant heat exchanger may tolerate water quality that a server cold plate cannot. Modern cold plates can include very fine internal flow passages, and ASHRAE warns that technology cooling system filtration is especially important because these tighter channels can foul if particles reach the IT cooling hardware. (resourcecenter.ashrae.org)

That is why water treatment solutions for liquid-cooled data centers must be designed around the actual loop being protected. A facility water system, a technology cooling system, and a water-based coolant inside a rack may all contain “water,” but they do not share the same chemistry expectations.

Start with the loop: facility water vs. technology cooling system

Before setting pH, conductivity, hardness, or microbial limits, identify which water loop you are discussing. ASHRAE’s water-cooled server guidance distinguishes between multiple loop types and emphasizes that each has different chemistry, filtration, and pressure requirements. Equipment designed for one loop type may not be suitable for another because of materials of construction and operating conditions. (resourcecenter.ashrae.org)

In practical terms, most liquid-cooled data center projects include some combination of these systems:

• Facility water system: The building-side loop that may serve chillers, dry coolers, cooling towers, plate-and-frame heat exchangers, or CDUs. It is often owned and maintained by facilities.
• Technology cooling system: The IT-side loop between the CDU and the racks, cold plates, or rear-door heat exchangers. It is typically more sensitive to contamination.
• Coolant inside packaged equipment: Some solutions use factory-filled or vendor-specified fluids with additive packages, glycol, inhibitors, or biocides.
• Makeup or fill water: The water used to charge, top up, flush, dilute, or blend the system. Its quality can be very different from the final treated loop chemistry.

A common mistake is applying the wrong specification to the wrong loop. ASHRAE specifically notes that facility water system and technology cooling system recommendations have been misapplied, and that the technology cooling system is typically stricter than the facility water system because the IT-side loop is associated with specific hardware and smaller passages. (resourcecenter.ashrae.org)

Practical target ranges: useful benchmarks, not universal limits

The safest answer for any project is: follow the server OEM, CDU supplier, heat transfer fluid supplier, and water treatment provider. Their requirements should become the governing specification for warranty, commissioning, and ongoing maintenance. However, industry guidance provides useful starting points.

ASHRAE’s water-cooled server white paper summarizes different water quality benchmarks for facility water systems and technology cooling systems. In that summary, facility water guidance includes pH 7 to 9, corrosion inhibitors, limits for sulfate, chloride, bacteria, hardness, residue after evaporation, and turbidity. The technology cooling system guidance shown there is tighter, including pH 8.0 to 9.5, required corrosion inhibitors and biocide, lower limits for sulfide, sulfate, chloride, bacteria, hardness, residue, suspended solids, and a conductivity range of 0.2 to 20 micromho/cm. (resourcecenter.ashrae.org)

Other water-based heat transfer fluid guidance can look very different because it assumes treated water with an inhibitor package. OCP’s water-based transfer fluid guidance for single-phase cold plate racks lists typical treated-fluid properties such as total suspended solids below 5 ppm, total dissolved solids below 1,000 ppm after treatment, conductivity below 1,500 µS/cm at 25°C after treatment, pH 8.0 to 10.5, total hardness below 30 ppm as CaCO₃, turbidity below 5 NTU, operating bacteria below 100 CFU/mL, and copper corrosion inhibitor guidance for azoles. (opencompute.org)

Those two examples illustrate an important decision point: a low-conductivity DI-style loop and an inhibited water-based coolant loop are managed differently. Conductivity that would be unacceptable in one program may be normal in another after additives are introduced. The goal is not to chase a generic number; the goal is to establish a verified baseline and control drift from that baseline.

pH: the first indicator of chemical stability

pH is one of the most important online measurements in data center water because it affects corrosion tendency, inhibitor performance, microbial control, and material compatibility. A pH that is too low can accelerate corrosion of metals. A pH that is too high can contribute to scaling, elastomer compatibility concerns, or additive performance issues depending on the formulation.

For liquid cooling, pH should be interpreted in context:

• Loop type: Facility water and technology cooling water may have different pH ranges.
• Metals: Copper, stainless steel, brass, aluminum, nickel plating, and brazed joints can respond differently to pH.
• Additive package: Buffers, azoles, molybdates, phosphates, glycols, and biocides may shift the desired pH.
• Temperature: pH readings vary with temperature, so measurement and compensation practices matter.
• Sampling method: Grab samples can change after exposure to air, especially in low-ionic-strength water.

From a sensor standpoint, pH should be measured where the sample is representative and where maintenance is practical. Inline or sidestream pH sensors should be installed with controlled flow, stable pressure, isolation valves, calibration access, and temperature measurement. In low-conductivity water, pH measurement can be more difficult because the sample has limited ionic strength; selecting a sensor designed for low-conductivity service can improve stability.

A strong pH program includes:

• A project-specific operating range.
• Warning and critical alarm limits.
• Calibration intervals based on drift history.
• Documentation of every chemical addition.
• Trending against conductivity, temperature, ORP, corrosion metals, and turbidity.
• A defined response plan for pH excursions.

If pH begins to drift, do not dose blindly. Confirm the reading, check calibration, compare with a lab or portable meter, review recent chemical additions, inspect makeup water quality, and evaluate whether air ingress, microbial activity, resin exhaustion, or incompatible materials could be changing the chemistry.

Conductivity: the fastest way to see ionic change

Conductivity measures water’s ability to carry electrical current, which increases as dissolved ions increase. The USGS notes that specific conductance is related to ion type and concentration and can be used to approximate dissolved solids content, making it a valuable real-time surrogate for changes in water chemistry. (usgs.gov)

In data center liquid cooling, conductivity is useful because it can quickly reveal:

• Makeup water contamination.
• Improper fill water.
• Ion exchange resin exhaustion.
• Leaching from hoses, seals, metals, or flux residues.
• Additive overfeed or underfeed.
• CO₂ absorption in low-conductivity water.
• Mixing between loops through a heat exchanger leak.
• Concentration changes after evaporation, bleed, dilution, or fluid replacement.

However, conductivity is not a complete water quality specification. ASHRAE notes that technology cooling system conductivity can rise over time due to ionic contaminants, CO₂ permeation, materials leaching, or the addition of corrosion inhibitors, biocides, or buffers; the guidance also emphasizes charting conductivity over time and understanding the root cause of deviations. (resourcecenter.ashrae.org)

This is where many monitoring programs fail: they treat conductivity as a single pass/fail number. A better approach is to define three values:

1. Fill or makeup water conductivity: The quality of water entering the system.
2. Post-treatment baseline conductivity: The expected value after inhibitors, buffers, glycol, or biocide are added.
3. Operating drift limit: The allowable change over time before investigation is required.

Sensor selection depends on the range. For low-conductivity or high-purity applications, a contacting conductivity sensor with the correct cell constant and temperature compensation is typically appropriate. For higher-conductivity, treated, or potentially fouling fluids, toroidal conductivity can be considered, although it may not be suitable for very low ranges. The important point is to size the measurement technology to the actual conductivity window rather than using a general-purpose probe.

Beyond pH and conductivity: the full water quality picture

pH and conductivity are core measurements, but they are not enough by themselves. Water quality for data centers should include physical, chemical, and microbiological indicators.

Temperature and dew point risk

Temperature is both a cooling parameter and a water chemistry parameter. Higher temperatures can accelerate reactions, influence biological activity, and affect sensor readings. Temperature also matters for condensation risk. ASHRAE recommends that non-CDU liquid cooling implementations regulate facility water temperature so it remains at least 2°C above the data center room ambient dew point, with alarms when supply water gets within that buffer. (resourcecenter.ashrae.org)

A complete monitoring strategy should include coolant supply temperature, return temperature, differential temperature, and—where relevant—server inlet air dew point. These values help operations teams distinguish a thermal performance issue from a chemistry issue.

Turbidity, suspended solids, and particle control

Particles are a direct threat to cold plates and microchannels. Turbidity and total suspended solids provide early warning that corrosion products, biological material, installation debris, scale, or degraded elastomers are circulating. OCP recommends routine monitoring that includes water appearance and filter loading, and it calls for investigation if the fluid is not clear or filter differential pressure is high. (opencompute.org)

For the facility side, strainers may be adequate depending on heat exchanger design. For the technology cooling system, filtration should be matched to the smallest passages in the IT cooling equipment. ASHRAE recommends absolute filter ratings for TCS filtration because nominal ratings are too loose for protecting IT cooling hardware. (resourcecenter.ashrae.org)

Hardness, alkalinity, and scale tendency

Hardness is mainly associated with calcium and magnesium. In a cooling loop, hardness can contribute to scale formation when temperature, pH, alkalinity, and concentration conditions allow minerals to precipitate. Scale reduces heat transfer and can restrict flow.

Low hardness is typically desired for technology cooling systems, especially during fill and flushing. OCP’s water-based transfer fluid guidance treats high hardness as an indicator of poor makeup water quality and lists a fresh-fill desired hardness below 2 ppm, with treated-fluid total hardness below 30 ppm as CaCO₃. (opencompute.org)

Chloride, sulfate, and aggressive ions

Chloride and sulfate are important because they can increase corrosion risk, especially for stainless steels, copper alloys, and brazed components under certain conditions. OCP’s guidance includes chloride limits tied to stainless steel selection and notes that chloride limits are intended to reduce crevice corrosion risk where concentration mechanisms or pH depression may occur. (opencompute.org)

Because chloride and sulfate are not fully characterized by conductivity alone, periodic lab analysis is important. A stable conductivity trend does not prove that aggressive ions are absent; it only suggests that total ionic content is not changing dramatically.

Dissolved oxygen and corrosion potential

Dissolved oxygen can contribute to corrosion in closed loops, particularly where oxygen ingress continues through makeup water, permeable materials, poorly maintained expansion tanks, or repeated maintenance openings. Measuring dissolved oxygen is especially useful during commissioning, troubleshooting, and high-criticality operation.

ORP can also be valuable when a treatment program uses oxidizing or non-oxidizing biocides, although ORP must be interpreted carefully. It is not a universal biocide concentration measurement, but it can indicate changes in oxidation-reduction conditions that may correlate with treatment performance.

Microbiological activity and biofilm

Microbial contamination can lead to biofilm, under-deposit corrosion, plugging, and degraded heat transfer. ASHRAE notes that biofilms are naturally occurring in water systems and are controlled by stable water chemistry, minimized organic and ionic loading, good fluid design, avoiding dead legs, and minimizing low-velocity areas. (resourcecenter.ashrae.org)

For building water systems more broadly, CDC guidance emphasizes monitoring temperature, disinfectant residuals, pH, and slow-moving areas to reduce Legionella growth and spread. That guidance is especially relevant where data center water systems include cooling towers, potable makeup, or building water management responsibilities, even though a closed IT coolant loop has a different exposure profile. (cdc.gov)

Corrosion metals: copper, iron, zinc, nickel, and chromium

Corrosion byproducts are among the most decision-useful lab measurements. Rising copper can indicate cold plate, copper piping, brazed joint, or brass component attack. Rising iron can indicate steel corrosion. Zinc can suggest brass dezincification or galvanized material exposure. Nickel and chromium may indicate issues with plated components or stainless alloys.

OCP recommends independent laboratory QA testing that includes corrosion byproducts and metals, and it notes that careful monitoring of copper levels is recommended because standard corrosion coupons or linear polarization resistance probes may not represent cold plate conditions well. (opencompute.org)

Water treatment solutions for data center liquid cooling

The right water treatment approach depends on the loop. A hyperscale AI hall with many CDUs may require a different program than a small enterprise deployment with a packaged coolant loop. Still, most projects should evaluate these treatment building blocks.

Pretreatment for makeup and fill water

Makeup water is often the source of long-term problems. Depending on local water quality and system requirements, pretreatment may include:

• Sediment filtration.
• Carbon filtration when disinfectants or organics must be controlled.
• Softening to reduce calcium and magnesium.
• Reverse osmosis to reduce dissolved ions.
• Deionization or mixed-bed polishing for low-conductivity fill water.
• UV or sterile filtration where microbial control is critical.
• Controlled storage and handling to prevent recontamination.

For low-conductivity systems, fill water handling matters as much as production. High-quality water can quickly absorb CO₂ or pick up ions from tanks, hoses, fittings, and open containers.

Cleaning, flushing, and commissioning

New systems often contain fabrication debris, flux residues, oils, particles, and installation contaminants. Existing systems may contain degraded fluid, corrosion deposits, incompatible additives, or microbiological growth. OCP recommends cleaning and flushing before final heat transfer fluid installation and provides startup practices such as using clean water, replacing filters as needed, circulating to suspend debris, draining, and refilling with the specified fluid. (opencompute.org)

Commissioning should create the baseline for the life of the system. At minimum, record pH, conductivity, temperature, turbidity, hardness, inhibitor level, bacteria, visual appearance, filter differential pressure, and lab metals after the system is stable.

Corrosion inhibitors and buffers

Corrosion inhibitors protect wetted metals, but they also change the chemistry being measured. Azoles are commonly used for copper and yellow metal protection, while other formulations may use different inhibitor packages. OCP’s water-based guidance highlights the importance of an inhibitor package designed for the materials present and specifically notes the importance of copper corrosion protection. (opencompute.org)

Never add inhibitors based on a generic recommendation. Verify compatibility with cold plates, CDUs, hoses, seals, pumps, valves, heat exchangers, and any glycol or biocide already in use.

Biocide and biological control

Biocide selection must account for materials, discharge requirements, safety, and treatment chemistry. OCP’s water-based transfer fluid guidance states that non-oxidizing biocide may be added if microbial activity is high and warns against halogens such as chlorine because of corrosion risk. (opencompute.org)

Good biological control is not only a chemical program. It also includes avoiding dead legs, maintaining adequate flow, eliminating stagnant bypasses, keeping fill water clean, minimizing organic loading, and monitoring bacteria through periodic lab or field testing.

Continuous monitoring vs. periodic sampling

Not every parameter needs an online sensor. The best program combines continuous measurement for fast-changing indicators with periodic laboratory analysis for parameters that require speciation or culture.

Continuous monitoring is well suited for:

• pH.
• Conductivity or resistivity.
• Temperature.
• ORP where relevant.
• Turbidity.
• Dissolved oxygen in selected systems.
• Flow, pressure, and differential pressure across filters.
• Leak detection and fluid level.

Periodic field or lab testing is usually better for:

• Bacteria and microbiological counts.
• Specific ions such as chloride and sulfate.
• ICP metals such as copper, iron, zinc, nickel, and chromium.
• Inhibitor concentration.
• Glycol concentration.
• TDS, TSS, and detailed fluid condition.

ASHRAE notes that online monitoring provides the best visibility into system health and stability, but cost and risk must be balanced. It also identifies conductivity, pH, and turbidity as parameters that can be measured online with basic instruments, while bacteria requires periodic sampling. (resourcecenter.ashrae.org)

OCP recommends more frequent testing during startup—weekly for a month, monthly for a quarter, and quarterly thereafter or until results stabilize—along with quarterly independent lab QA testing for the first year, with reduced frequency after stable results. (opencompute.org)

How to design a sensor package for data center water

A liquid analysis sensor package should be designed like a reliability system, not a collection of probes. For most liquid-cooled data center projects, we recommend a measurement plan that answers five questions.

1. What loop are we measuring?

A facility water loop may need robust sensors for higher conductivity, higher solids, treatment chemicals, and variable water quality. A technology cooling loop may need higher sensitivity, tighter temperature compensation, better low-flow sample handling, and compatibility with treated heat transfer fluid.

2. What are the control limits?

Sensors are only useful when the data has a decision attached. Define normal, warning, alarm, and shutdown or service thresholds. Where OEMs provide limits, use them. Where they do not, create site-specific limits based on treatment provider guidance and commissioning baseline data.

3. Where should sensors be installed?

Common monitoring locations include:

• Makeup water inlet.
• Post-treatment or polishing outlet.
• CDU supply to IT equipment.
• CDU return from IT equipment.
• Sidestream sample panel on the technology cooling loop.
• Heat exchanger inlet and outlet.
• Filter inlet and outlet for differential pressure.
• Expansion tank or reservoir where fluid chemistry may change.

A sidestream panel is often the most maintainable option for pH, conductivity, ORP, turbidity, and dissolved oxygen. It should include isolation valves, pressure regulation if needed, flow indication, drain routing, calibration access, and compatible wetted materials.

4. How will data reach operations teams?

For data centers, sensor data should integrate with the building management system, DCIM platform, CDU controller, or plant historian. Useful outputs include 4–20 mA, digital protocols, relays, and local display, depending on the architecture. Alarms should be actionable and should distinguish sensor fault, calibration due, high/low process value, and rate-of-change events.

5. How will sensors be maintained?

Every sensor needs a maintenance plan. For pH, that means calibration, reference electrolyte considerations, cleaning, and replacement intervals. For conductivity, it means verification against standards and inspection for fouling or bubbles. For turbidity, it means optical cleaning and bubble control. For dissolved oxygen, it means membrane or optical cap maintenance depending on technology.

Troubleshooting common water quality excursions

When water quality moves out of range, the response should be systematic.

If pH drops

Possible causes include acidic contamination, CO₂ absorption, biological activity, inhibitor depletion, incompatible chemical additions, or corrosion reactions. Confirm the reading, check calibration, review recent work orders, test conductivity, and send a lab sample for ions and metals.

If pH rises

Possible causes include overfeed of alkaline buffer, concentration of treatment chemicals, makeup water changes, or chemical incompatibility. Confirm pH at temperature and compare against inhibitor and hardness levels to assess scaling risk.

If conductivity rises

Possible causes include ion leaching, chemical overfeed, poor makeup water, heat exchanger leakage, concentration, or contamination during maintenance. Compare conductivity rise with chloride, sulfate, hardness, TDS, and metals to identify whether the increase is benign treatment chemistry or harmful contamination.

If conductivity falls

Possible causes include dilution, underfeed of inhibitor, incorrect makeup, polishing system changes, or sensor error. A falling conductivity trend can be as important as a rising trend if it indicates loss of corrosion inhibitor.

If turbidity or filter differential pressure increases

Possible causes include installation debris, corrosion products, biological growth, scale, degraded elastomers, or incompatible fluids. Inspect filters, review startup cleanliness, test metals, and evaluate whether cold plate performance is changing.

If copper, iron, or zinc rises

Possible causes include active corrosion, dezincification, plating wear, oxygen ingress, low inhibitor concentration, pH excursion, chloride contamination, or incompatible wetted materials. Verify inhibitor residual and consider a focused materials review.

If bacteria counts increase

Possible causes include contaminated fill water, stagnant zones, low biocide residual, organic contamination, dead legs, or inadequate flushing. Confirm with repeat testing, review biocide compatibility, and avoid repeated dosing without understanding the root cause.

Decision guide: when to invest in online water quality monitoring

Manual sampling may be acceptable for small, non-critical, vendor-packaged loops with stable chemistry and strong service support. Online monitoring becomes more valuable when:

• The system supports high-density AI or mission-critical workloads.
• Multiple CDUs or many racks share a loop.
• Cold plates have very small channels.
• The site has variable makeup water quality.
• Warranty compliance requires documentation.
• The facility is remote or lightly staffed.
• Fluid treatment includes inhibitors or biocides that must be controlled.
• Past issues include corrosion, fouling, bacteria, or unexplained thermal degradation.
• The operator wants predictive maintenance rather than reactive sampling.

For many facilities, the best starting package is pH, conductivity, temperature, turbidity, and filter differential pressure on the technology cooling loop, plus conductivity and pH on makeup water. Larger or higher-risk systems may add ORP, dissolved oxygen, corrosion monitoring, automated sampling, and tighter integration with the BMS or DCIM environment.

Best practices for long-term water quality control

A strong liquid cooling water program should include:

• A loop-by-loop water quality specification.
• A complete wetted materials list.
• Approved water treatment solutions and chemical compatibility documentation.
• Defined fill, flush, and commissioning procedures.
• Baseline chemistry after startup.
• Continuous monitoring for fast-changing parameters.
• Periodic lab testing for ions, metals, bacteria, inhibitor, and glycol.
• Alarm limits tied to response actions.
• Calibration and sensor maintenance procedures.
• A record of every chemical addition, filter change, fluid replacement, and abnormal event.
• Regular review of trends, not just individual sample results.

The most important mindset shift is to treat water quality as operational data. pH, conductivity, turbidity, oxygen, temperature, and metals are not isolated lab values; they are early signals of system health.

Building a safer liquid cooling strategy with the right measurements

As liquid cooling adoption expands, the difference between a reliable system and a problematic one will often come down to water quality discipline. pH and conductivity are the foundation, but true protection requires a broader view: particles, hardness, chloride, sulfate, oxygen, microbial activity, corrosion byproducts, inhibitors, temperature, and trend behavior.

Our role is to help data center owners, engineers, CDU suppliers, and water treatment partners measure those variables with confidence. Whether you are designing a new AI-ready facility, commissioning your first direct-to-chip deployment, or improving monitoring on an existing loop, the right liquid analysis sensors can turn data center water from a hidden risk into a managed asset.

If you are defining water quality requirements for an upcoming liquid cooling project, contact our team to discuss sensor selection, sample panel design, monitoring points, and integration options for your facility water and technology cooling loops.