How to Calculate Dew Point in a Data Center — and Use It to Prevent Condensation

September 16, 2026 DC Build X ASHRAE, Cold Plate, Condensation Control, Data Center Cooling, Dew Point, Humidity Control, Liquid Cooling, Psychrometrics

Condensation is a temperature-difference problem, not a humidity problem. This guide works through the dew point calculation from first principles — the five-second rule of thumb, the Magnus inversion used in practice, and a fully worked cold-aisle example — then applies it to chilled-water surfaces, cold plates and CDU loops, with the ASHRAE envelopes and design margins that follow. Cited throughout.

A cold plate supplied at 12 °C in a room with a 10 °C dew point stays bone dry. The same plate in a room whose dew point has drifted up to 13 °C is now the coldest surface in the hall, and it is quietly raining on the tray beneath it. Nothing about the cooling loop changed. What changed is a number that most facility teams never actually compute.

Condensation is not a humidity problem. It is a temperature-difference problem. Water appears on a surface the instant that surface drops below the dew point of the air touching it, and it stops appearing the instant it rises above it. So the engineering question is never “is the room too humid?” It is “what is the dew point, and what is the coldest surface in the room?”

This article walks through how to get that dew point number properly — the fast estimate, the formula that is accurate enough to design to, and a fully worked example — then translates it into the condensation rules that matter in a modern hall, including liquid-cooled rows.

The moisture question that actually matters

Three quantities describe the water in the air, and only one of them tells you whether you have a condensation risk.

Relative humidity (RH) is a ratio: how much water vapour the air is holding, compared with the maximum it could hold at its current temperature. Because that maximum rises steeply with temperature, RH changes when the temperature changes even though not one drop of water was added or removed.
Humidity ratio (W) is the mass of water vapour per unit mass of dry air. It is an absolute measure, but it is not a temperature, so it does not directly answer “what will condense where.”
Dew point (t_d) is the temperature to which the air must be cooled, at constant pressure, for it to become saturated. It is an absolute measure of moisture expressed as a temperature — which is exactly why it compares directly against the temperature of a pipe, a cold plate or a containment door.

That last property is the whole reason dew point beat relative humidity as the industry’s control variable. The cooling industry’s own literature put it plainly: IT space humidity should be centrally controlled by maintaining dew point at the IT intakes, which is more cost effective than maintaining relative humidity at the return of the cooling units.^[8]

There is a second, more expensive reason. In a room with several cooling units, each controlling to its own return-air RH, the units disagree with each other by construction. Air at 32.2 °C and 50 % RH carries about 95 % more water by weight than air at 21.1 °C and 50 % RH.^[7] So a unit seeing warm return air adds far more moisture than the unit next to it, the room’s humidity climbs, and one of the units eventually flips into dehumidification. Half the plant humidifies while the other half dehumidifies. That condition has a name — demand fighting — and it is one of the primary causes of excessive energy consumption in IT environments.^[7]

Dew point control removes the disagreement, because the dew point of the air leaving a server is identical to the dew point of the air entering it. Air warms as it crosses the rack; its moisture content does not change.

Level 1: the five-second estimate

If you need a sanity check and not a design number, there is a well-documented shortcut from the atmospheric sciences literature: for moist air above about 50 % RH, the dew point sits roughly 1 °C below the dry-bulb temperature for every 5 % the relative humidity falls short of 100 %.^[6]

t_d ≈ t − (100 − RH) ÷ 5

t in °C, RH in percent. Valid as a rule of thumb for RH > 50 %.

At 24 °C and 50 % RH it gives 14 °C, and the correct answer is 13.0 °C — fine for a walk-around. At 24 °C and 45 % RH it gives 13.0 °C against a true 11.4 °C, and that 1.6 K of optimism is already the difference between “my 12 °C water is safe” and “my 12 °C water is condensing.” Use it to catch nonsense. Do not use it to set a setpoint.

Level 2: the formula you should actually design to

Going from a dry-bulb temperature and a relative humidity to a dew point is a three-step calculation. Steps 1 and 2 are the standard psychrometric chain; step 3 is the only one that requires care, because it is an inversion.

Step 1 — saturation vapour pressure at the measured air temperature. ASHRAE’s psychrometrics chapter gives a polynomial fit (attributed to Hyland–Wexler and adopted by ASHRAE) for the saturation pressure over water, valid from 0 to 200 °C, with temperature in kelvin and pressure in pascals:^[1]

ln p_ws = C8/T + C9 + C10·T + C11·T² + C12·T³ + C13·ln T
C8 = −5.8002206×10³ C9 = 1.3914993 C10 = −4.8640239×10⁻²
C11 = 4.1764768×10⁻⁵ C12 = −1.4452093×10⁻⁸ C13 = 6.5459673

Below 0 °C the calculation must switch to saturation over ice, which uses a different coefficient set and an extra T⁴ term.^[1] For a data hall it rarely matters — but it matters if you are checking a cold-climate economizer inlet.

Step 2 — the actual vapour pressure. Relative humidity is the ratio of the actual vapour pressure to the saturation pressure at the same temperature, so:

p_w = RH × p_ws(t) (RH as a fraction)

Step 3 — invert to get the dew point. The dew point is defined as the temperature at which the saturation pressure equals that actual vapour pressure: solve p_ws(t_d) = p_w.^[1] ASHRAE does this inversion either iteratively against the saturation table, or directly using the IAPWS-IF97 saturation-temperature equation for dew points at or above 0 °C.^[1] The older closed form in the same chapter is the Peppers correlation, which is accurate from 0 to 93 °C.^[1]

For engineering work in the 5–25 °C band, the most practical option is the Magnus form with the coefficients re-fitted by Alduchov and Eskridge. It inverts algebraically, so the dew point falls straight out of the equation:

p_ws(t) = 610.94 · exp[ 17.625·t / (243.04 + t) ] p_ws in Pa, t in °C

γ = ln( p_w / 610.94 )
t_d = 243.04 · γ / (17.625 − γ) the dew point

Alduchov and Eskridge optimised those two constants (17.625 and 243.04 °C) against accurate vapour-pressure references; the resulting fit stays within about 0.4 % of the reference values between −40 and +60 °C, which is well inside the uncertainty of the sensors that feed it.^[5]^[6]

Worked example: 24 °C and 45 % RH in the cold aisle

Take a typical contained cold aisle: 24 °C dry bulb, 45 % RH, sea level.

1. p_ws(24 °C) = 2985 Pa
2. p_w = 0.45 × 2985 = 1343 Pa
3. γ = ln(1343 / 610.94) = 0.7874
t_d = 243.04 × 0.7874 / (17.625 − 0.7874) = 11.4 °C

The same moisture content, expressed three ways: a humidity ratio of 8.4 g of water per kg of dry air; a dew point of 11.35 °C; and a relative humidity that depends entirely on where you measure it.

That last point is worth a table, because it is where RH-based specifications fall apart. Hold the moisture constant and watch what happens as the air moves through the hall:

Measurement point	Dry bulb	RH at the same moisture	Dew point
Cold aisle / IT inlet	24 °C	45 %	11.4 °C
Mid-hall (mixed)	27 °C	38 %	11.4 °C
Above a rack	30 °C	32 %	11.4 °C
Hot aisle	32.2 °C	28 %	11.4 °C
CRAC return	35 °C	24 %	11.4 °C

One moisture condition, five different RH readings. This is why an RH setpoint means different things at every sensor in the room — and why the dew point is the same number everywhere.

Now put a surface in that room. A chilled-water valve body, coil header or uninsulated pipe carrying 7 °C water sits 4.4 K below the 11.4 °C dew point. It will condense, and it will keep condensing no matter what the room’s RH reading says, because the room’s moisture is what it is. Sweating chilled-water components are the oldest condensation failure in the industry, and they are entirely predictable from this one comparison.

The design rule in one line: find the coldest surface temperature in the space, and require it to stay above the dew point by a margin. Everything else in this article is either how to compute the dew point or how to choose the margin.

If you only have humidity ratio or wet bulb

Sensor and DCIM exports do not always hand you an RH. Two conversions cover most cases.

From humidity ratio W (kg water per kg dry air), the vapour pressure follows from the partial pressures:

p_w = p · W / (0.621945 + W) p = local barometric pressure, same units as p_w

Then invert as before. Because W is absolute, this route has no temperature dependence at all — which is why it is the more robust of the two if you are reconstructing conditions from logged data:

Humidity ratio	Dew point	Comment
6 g/kg	6.5 °C	Dry hall; comfortable margin against cold surfaces
8 g/kg	10.7 °C	Typical well-run contained hall
10 g/kg	14.0 °C	Approaching the ASHRAE recommended ceiling of 15 °C dp
12 g/kg	16.8 °C	Above the Class A1 allowable maximum dew point of 17 °C — no cold-water surface is safe
16 g/kg	21.3 °C	Out of envelope for every air-cooled class by dew point

From a measured wet bulb you can get to humidity ratio through the adiabatic-saturation relation in the psychrometrics chapter, and then to dew point as above.^[1] Note the chapter’s own warning: a psychrometer’s wet bulb is not exactly the thermodynamic wet-bulb temperature, and a wet/ice-bulb reading is imprecise around 0 °C.^[1] For dew point control, measure the dew point (or RH plus temperature) rather than inferring it from a sling psychrometer.

Where condensation actually bites

Every one of these failures is the same equation: a surface temperature below the room’s dew point. What differs is which surface.

Location	Why it condenses	What to do
Chilled-water pipes, valves, coil headers	Water at 7–12 °C against a dew point that is often higher	Insulate and vapour-seal; raise chilled-water supply temperature; drip trays under valves
Underfloor plenum	Cold air plus humid infiltration through slabs and cable entries; the coldest surfaces in the building	Seal penetrations; keep the plenum above dew point; monitor plenum dew point, not just room RH
Cold aisle containment doors, frames, blanking panels	Uninsulated glass-and-steel assemblies sitting directly in supply air	Specify insulated or thermally broken containment; check the frame temperature, not the air temperature
CDU secondary loop, manifolds, quick-disconnects, hoses	Uninsulated by design (most in-rack loops are), and often run cold for chip reasons	Hold supply above the room dew point by a margin; leak detection at the low point of each manifold
Cold plates on non-CDU loops	Direct building chilled water, typically 4–6 °C, is far below any data-hall dew point	Do not use building chilled water directly; put a CDU or heat exchanger between the loops
Direct evaporative / adiabatic cooling intakes	Evaporative cooling adds moisture to the air stream, raising the dew point of the supply	Model the dew point of the supply air, not the outdoor air, when checking downstream surfaces
Immersion tank walls and plumbing	Coolant at low temperature against a humid room	Same margin rule; keep the tank exterior above room dew point or enclose it

The industry’s current answer to most of this is not dehumidification — it is to keep the cold surfaces warmer. ASHRAE’s guidance for liquid-cooled servers is explicit: where a CDU can raise the secondary coolant temperature above the dew point, use it; in non-CDU implementations the facility water must be regulated so it is always maintained at least 2 °C (3.6 °F) above the data centre’s room ambient dew point, with that buffer covering sensor uncertainty.^[4]

Hardware documentation turns that into a number. One rack-level liquid cooling product specifies that the water in the supply hose, return hose and heat exchanger must be kept above the dew point of the location — and gives 18 ± 1 °C as the minimum for a room to ASHRAE Class 1 (maximum dew point 17 °C), and 22 ± 1 °C for Class 2 (maximum dew point 21 °C).^[10] The same document warns that ordinary building chilled water is typically too cold for this, because it can be as low as 4–6 °C.^[10]

The wider liquid-cooling consensus is moving the same direction. ASHRAE’s liquid-cooling classes are named for their maximum facility-water supply temperature — W17 being the coldest, with warmer classes above it — precisely because running water hotter than the room’s dew point removes condensation from the risk register and unlocks chiller-free cooling.^[9]^[2] The Open Compute Project’s Advanced Cooling Facilities workstream frames it as a design preference: operating above the dew point is preferred where possible, with condensation protection, detection and collection as the fallback rather than the default.^[13]

What ASHRAE actually specifies

These are the numbers your dew point calculation is checking against. The recommended envelope is where a facility should run for reliability and efficiency; the allowable envelope is what equipment is tested to function within — a statement of functionality, not of reliability.^[2]

Class	Allowable dry bulb	Allowable moisture	Max dew point
Recommended (A1–A4)	18–27 °C	−9 to 15 °C dp, and 70 % RH (or 50 % RH where corrosive gases are present)	15 °C
A1	15–32 °C	−12 °C dp & 8 % RH up to 80 % RH	17 °C
A2	10–35 °C	−12 °C dp & 8 % RH up to 80 % RH	21 °C
A3	5–40 °C	−12 °C dp & 8 % RH up to 85 % RH	24 °C
A4	5–45 °C	−12 °C dp & 8 % RH up to 90 % RH	24 °C
H1 (high density)	5–25 °C, recommended 18–22 °C	−12 °C dp & 8 % RH up to 80 % RH	17 °C

Equipment environment specifications for air cooling, 2021 guidelines. All values are at the IT equipment inlet.^[2]

Three things about this table are easy to get wrong.

The dew point figures are inlet conditions. A sensor on the CRAC return can read happily inside the envelope while the air entering the top of a rack is outside it.
The low-moisture limit is a compound condition — the higher of −12 °C dew point and 8 % RH — not a simple floor.^[2]
The fifth edition changed the humidity rules. The low end was extended to −9 °C dew point on the strength of electrostatic-discharge research, the upper limit moved to 70 % RH but only where gaseous pollution is low (50 % otherwise, based on an ASHRAE research project on corrosion run by Syracuse University from 2015 to 2018), and a new high-density class H1 was added.^[3]^[12]

A quick-look dew point table

For day-to-day operations, this grid removes the arithmetic. Rows are dry bulb, columns are relative humidity, cells are dew point in °C. Cell shading flags the recommended-envelope ceiling of 15 °C dp and the Class A1 allowable maximum of 17 °C.

Dry bulb	30 % RH	40 % RH	45 % RH	50 % RH	60 % RH	70 % RH
18 °C	0.2	4.2	5.9	7.4	10.1	12.5
20 °C	1.9	6.0	7.7	9.3	12.0	14.4
22 °C	3.6	7.8	9.5	11.1	13.9	16.3
24 °C	5.4	9.6	11.3	12.9	15.8	18.2
27 °C	8.0	12.3	14.1	15.7	18.6	21.1
30 °C	10.5	14.9	16.8	18.4	21.4	23.9
32 °C	12.3	16.7	18.6	20.3	23.3	25.8

green = inside the 15 °C dew point ceiling · amber = above 15 °C, outside the recommended envelope · red = above 17 °C, beyond the Class A1 allowable maximum dew point. Values computed with the ASHRAE saturation-pressure fit at sea level; expect small shifts at altitude.

Read one row and the practical consequence is obvious. At 24 °C, moving from 45 % to 60 % RH raises the dew point from 11.3 °C to 15.8 °C — that is 4.5 K of condensation margin given away, and every chilled-water surface below 15.8 °C in the hall is now a wetted surface.

Turning the number into a design decision

Once you can compute the dew point, the design sequence is mechanical:

Pick the moisture setpoint as a dew point, not an RH. Something in the 5–12 °C band gives you real margin against cold surfaces while staying inside the recommended envelope.
Find the coldest surface in the space. Not the air temperature — the surface. Chilled-water valves, coil headers, CDU hoses, manifold bodies, containment frames, plenum pipework.
Require surface ≥ dew point + margin. ASHRAE’s liquid-cooling guidance uses 2 °C above the room dew point for non-CDU facility water, explicitly to absorb sensor uncertainty.^[4] For uninsulated in-rack components, treat 2 K as the floor and think harder about anything that carries water at single-digit temperatures.
Where the margin cannot be met, change the surface, not the room. Raise the coolant temperature (CDU, bypass loop, or warmer facility water), insulate and vapour-seal, or isolate the loop. Dehumidifying an entire hall to protect one valve is the expensive answer, and it is usually the wrong one.
Let the coils run warm. Cooling coils that run below the room dew point dehumidify by accident: moisture is wrung out of the air, and the humidifier adds it back a few metres away. Lawrence Berkeley National Laboratory reported a 28 % energy reduction in a data centre when its humidity controls were simply switched off, and notes that raising coil temperatures above the dew point eliminates the condensation and increases both efficiency and effective cooling capacity.^[9]
Control at the IT inlet, with a wide dead band. Centralised dew point control at the intakes, with a band wide enough that small swings do not trigger humidification or dehumidification, is what stops units fighting each other.^[8]^[7]
Alarm on the margin, not on the absolute number. The useful alarm is “coldest supply water − room dew point < 2 K”, computed continuously from two live sensors. A dew point alarm alone tells you nothing about whether a surface is at risk.
Check the air you are introducing. Direct evaporative and adiabatic systems add moisture to the supply stream. Model the dew point downstream of the humidification step, because that is the number the cold surfaces will see.^[11]

The part that is easier to get wrong than the maths

The formula is not the weak link. The sensors are.

A capacitive RH sensor with a ±2 % RH accuracy translates into roughly ±1 to 3 K of dew point uncertainty depending on the operating point — and the same sensor, uncalibrated and drifting, will happily report a comfortable RH while the actual moisture sits two degrees of dew point higher. A chilled-mirror hygrometer is far more accurate but is a laboratory instrument, not a hall instrument.

Two practical consequences follow. First, the 2 K buffer in ASHRAE’s liquid-cooling guidance exists for exactly this reason; treat it as covering measurement error, not design sloppiness.^[4] Second, place sensors where the envelope is defined — at the IT inlet, distributed across the hall and up the height of the racks — rather than relying on a single unit-mounted reading at the return.^[8]

Common mistakes

Controlling to relative humidity. RH moves with temperature; dew point does not. Controlling RH means chasing a number that changes when nothing has happened.
Specifying RH limits in a project brief. “Keep the hall at 45–55 % RH” is unspecifiable without a temperature, and it produces demand fighting the moment there is a temperature gradient — which there always is.
Checking the air, not the surface. The air cannot condense. Only surfaces can.
Using the rule of thumb as a design value. At 45 % RH it over-predicts the dew point by 1.6 K, which is precisely the margin people need.
Forgetting the underfloor. Supply plenums hold the coldest air and the coldest pipework in the building, and they are the last place anyone installs a dew point sensor.
Running direct evaporative cooling without re-checking downstream surfaces. The dew point of the supply air is the one that matters after humidification.
Sizing a liquid-cooling loop from a different edition of the guideline. The 15 °C recommended ceiling, the 17 °C Class A1 maximum and the Class H1 addition all changed between editions. Check which edition the specification references.^[2]^[3]

The bottom line

Dew point is one calculation, done from two measurements, and it converts the vague risk of “condensation” into a specific comparison between a temperature you control and a surface you know about. Compute it at the IT inlet. Compare it against the coldest surface in the room. Hold the difference above two degrees. Warm the water where you can, insulate where you cannot, and stop moving moisture around a hall that never needed it in the first place.

References

ASHRAE. 2021 ASHRAE Handbook — Fundamentals, Chapter 1: Psychrometrics (saturation pressure over water and ice, dew-point definition and calculation, numerical calculation of moist air properties). ASHRAE, Atlanta. handbook.ashrae.org
ASHRAE. 2023 ASHRAE Handbook — HVAC Applications, Chapter 20: Data Centers (Table 1, 2021 Thermal Guidelines equipment environment specifications for air cooling; liquid-cooling classes). ASHRAE, Atlanta. handbook.ashrae.org
ASHRAE Technical Committee 9.9. Thermal Guidelines for Data Processing Environments, 5th edition (Datacom Series Book 1, 2021), including the TC 9.9 Reference Card.
ASHRAE Technical Committee 9.9. Water-Cooled Servers: Common Designs, Components, and Processes (white paper, 2019) — design considerations for CDU and non-CDU implementations, including the requirement to hold facility water supply at least 2 °C (3.6 °F) above the data centre room ambient dew point. ashrae.org (PDF)
Alduchov, O. A., and R. E. Eskridge. 1996. “Improved Magnus Form Approximation of Saturation Vapor Pressure.” Journal of Applied Meteorology 35(4): 601–609. DOI: 10.1175/1520-0450(1996)035<0601:IMFAOS>2.0.CO;2
Lawrence, M. G. 2005. “The Relationship between Relative Humidity and the Dewpoint Temperature in Moist Air: A Simple Conversion and Applications.” Bulletin of the American Meteorological Society 86(2): 225–234. DOI: 10.1175/BAMS-86-2-225 — amsjournals.org (PDF)
Schneider Electric Data Center Science Center. Humidification Strategies for Data Centers and Network Rooms, White Paper 58 Rev 3 — dew point control versus relative humidity control, demand fighting, and moisture content at differing dry-bulb temperatures. apc.com (PDF)
Schneider Electric Data Center Science Center. Optimize Data Center Cooling with Effective Control Systems, White Paper 225 Rev 0 — centralised humidity control by dew point temperature at the IT intakes. apc.com (PDF)
Lawrence Berkeley National Laboratory. Realizing High-Performance Buildings (Data Centers chapter) — cooling coils running below the dew point cause unintentional dehumidification; 28 % energy reduction from disabling humidity control in an LBNL data centre. eta-publications.lbl.gov (PDF)
Lenovo. ThinkSystem Heavy Duty Rack Cabinets — Water specifications for the secondary cooling loop: coolant must be kept above the location’s dew point; minimum 18 ± 1 °C for ASHRAE Class 1 and 22 ± 1 °C for Class 2; typical building chilled water at 4–6 °C is too cold. lenovofiles.com
AIRAH. 2021. “Moderating the Impact of Integrating Water-Cooled Servers Into Data Centres.” Ecolibrium, October 2021 — supply water must be above the room dew point; two control strategies for maintaining the margin; ASHRAE Class A1 maximum dew point of 17 °C. airah.org.au (PDF)
Uptime Institute Journal. 2021. “New ASHRAE guidelines challenge efficiency drive” — fifth-edition changes to the humidity envelope (70 % RH conditional on low gaseous pollution, 50 % otherwise), the corrosion research behind it, and the new Class H1 high-density envelope. journal.uptimeinstitute.com
Open Compute Project, Advanced Cooling Facilities (ACF) subproject — discussions on condensation protection and the preference for operating liquid-cooling loops above the dew point, with drip trays historically recommended where it was not. ocp-all.groups.io