I still remember the sinking feeling in September 2022. We had just completed a ‘simple’ cooling upgrade for a mid-sized colocation facility. The new CRAC units were in. The BTU calculations were perfect. The mechanical team was proud.
Then the first hot spot call came in. Then three more. Then the humidity sensor went haywire.
The client was furious. I was baffled. We had followed the playbook. Six weeks and $47,000 later, we were back to square one—tearing out half the new equipment and calling in the specialists I should have called first.
I’m a procurement coordinator who’s handled HVAC and cooling orders for about eight years. I’ve personally made—and meticulously documented—enough mistakes to fill a small handbook. This is the story of the biggest one, and the checklist I now use to keep our team from repeating it.
The Surface Problem: We Thought It Was Just About ‘More Cold Air’
Every data center manager knows the drill: your rack inlet temperatures start creeping above 75°F (24°C). Someone’s deploying higher-density blades. The old 20kW-per-rack average is now pushing 30kW. The obvious answer? Upgrade the cooling system. More tons, more airflow, more redundancy.
That was our thinking. The client’s requirement was straightforward: “Keep the IT load cool, don’t break the budget.” We spec’d out a set of chilled-water-based Computer Room Air Conditioners (CRACs) from a major brand. The numbers added up. On paper, we had 15% extra capacity.
But the paper lied. Or rather, we forgot to read the fine print on the real problem.
The Real Issue: We Ignored the ‘Dehumidification Trap’
Here’s something I learned the expensive way: cooling is not the same as climate control. Especially in a data center, where humidity swings can be more destructive than temperature spikes.
Standard CRAC units are designed to cool air. To do that, they pass warm air over chilled coils. This process naturally condenses moisture out of the air—that’s why your window AC unit drips water. In a data center, the problem is that this dehumidification is *uncontrolled*. The unit has to run its compressor to cool, and it can’t help but strip humidity while doing it.
Our mistake? We assumed the new CRACs would maintain the ASHRAE recommended humidity range (40-60% RH). They didn’t. When the IT load was light (nights, weekends), the cooling demand dropped, but the humidity control got worse because the units cycled on and off. Result: We had zones at 30% RH—dry enough to cause electrostatic discharge (ESD) risks—and zones at 70% RH—damp enough to risk condensation on cold surfaces.
Actually, let me be more precise: the worst zone we hit was 28% RH in one server row. That’s a Class 1 ESD risk. The client’s network engineer started wearing a grounding strap just to walk down the aisle. Not ideal.
What We Should Have Asked First
Looking back, the question we missed was simple: “Is the primary problem heat load, or is it simultaneous temperature and humidity control?”
For this client, the real issue wasn’t that the room was too hot; it was that the existing system couldn't keep up with the *latent* and *sensible* heat load together. The IT gear emits dry heat (sensible), but the building envelope and infiltration introduced moisture (latent). A standard CRAC copes poorly with mixed loads.
If I could redo that decision, I’d ask three things upfront:
- What’s your actual dew point? Not just the temperature. We should have measured dew point over a week to see the real moisture profile.
- What’s your air infiltration rate? Gaps in the raised floor, cable cutouts, door seals—these let in unconditioned air. Big source of latent load.
- Do you need precise humidity control, or just a range? For some colo clients, a 20% swing is fine. For others with tape libraries or high-density gear, it’s a disaster.
We skipped all three. Classic rookie move.
The Cost of Getting It Wrong
Let’s put a number on the mess. The original upgrade budget was $120,000 for four new CRAC units and installation. Here’s what we actually spent:
- $120,000 — the initial installation (planned)
- $8,900 — emergency service calls to rebalance refrigerant and check sensors (unplanned)
- $14,500 — retrofitting a ducted return air system to fix airflow distribution (facepalm moment)
- $6,200 — a humidity consultant’s fees to tell us what we should have known
- $47,000 — replacing two CRAC units with dedicated precision cooling units (the ones from Munters that we ultimately went with)
- ~$20,000 — client compensation for downtime and a 1-week service credit
Total: approximately $216,600. Nearly double the original budget. The client’s trust? Priceless and damaged.
This wasn’t a vendor problem. It was a specification problem. We asked for ‘cooling,’ but we didn’t specify ‘dehumidification control.’ The standard CRAC units were fine at their job; we just hired a hammer for a screw job.
The Solution (Short and Boring, Because the Problem Was the Interesting Part)
After the dust settled, we replaced the two worst-performing units with Munters data center cooling solutions—specifically, their precision air conditioners with integrated dehumidification and heat wheel technology. The difference was night and day.
Why they worked:
- Dedicated humidity control. The Munters units can run in ‘dry mode’ without overcooling the space. They use a rotating desiccant wheel to absorb moisture, independent of the cooling coil. So you can keep the room at 70°F and exactly 45% RH without the wild swings.
- Heat wheel efficiency. For the same cooling load, the energy consumption dropped by about 35% compared to the standard CRACs. The client’s power bill saw an immediate improvement.
- Filters that actually work. This sounds basic, but the Munters units have better filtration and are easier to maintain. The maintenance cycle went from quarterly to semi-annual.
Honestly, the best part was the peace of mind. Once the Munters units were in, the hot spots vanished, the humidity stayed rock steady, and I stopped getting 2 AM alarms on my phone.
But—and this is important—I’m not saying Munters is the only solution. This approach worked for our situation: a mid-sized colocation facility with a mixed load and no dedicated humidification infrastructure. If you’re building a greenfield hyperscale data center, the calculus might be different. You might do fine with a well-designed airside economizer and standard CRACs. Or you might need a full chilled water system.
I can only speak to my context: retrofit projects where the existing cooling system is struggling with latent load and swing. If that’s your problem, precision cooling from someone like Munters is worth a serious look. If you’re dealing with something else, your mileage may vary.
One More Lesson: The Dehumidifier Parts Rabbit Hole
I also learned that munters dehumidifier parts are a thing you need to plan for. The desiccant wheels themselves are great, but they need replacement every 3-5 years. The bearings on the rotating assembly need annual greasing. And the control boards are proprietary—you can’t just grab a generic controller from Grainger. We had a brief scare in 2024 when a firmware update bricked a controller. That’s a story for another day.
Moral: Plan for the lifecycle cost, not just the sticker price. A standard CRAC is a commodity; you can find parts anywhere. A precision cooling unit is a specialty item. You’re trading maintenance ease for performance reliability. For us, it was worth it. For a small office server room, it might be overkill.
Final Thought: The Checklist I Use Now
After that disaster, I made a pre-purchase checklist for any data center cooling upgrade. It’s saved us from repeating the same error three times since 2023.
- Measure the load profile. Don’t guess. Put a power meter on the PDUs for a week. Know the peak, the trough, and the daily swing.
- Measure the humidity profile. Same thing. Use a data logger. If your RH swings more than 15% without any cooling change, you have an infiltration problem.
- Determine the criticality. Is this tier 1 mission critical, or tier 2 ‘nice to have’? That dictates budget for redundancy and precision.
- Consider the total cost of ownership. Not just the unit price. Maintenance, parts, downtime risk, energy consumption. A 30% more efficient unit pays for itself in 2-3 years in many markets.
- Ask the question: ‘Do we need simultaneous temperature and humidity control?’ If yes, you’re in the precision cooling zone. Don’t try to fake it with standard CRACs.
- Call a specialist. If this is your first time with precision cooling, talk to someone who’s done it before. Your local HVAC distributor might not be the right resource. I should know.
This approach worked for us, but our situation was a 10,000 sq ft colo space with 200 kW of IT load. If you’re dealing with a smaller server closet or a hyperscale facility, the calculus might be different. You know your environment better than I do.
One last thing: this was accurate as of early 2025. The HVAC market changes fast. Verify current pricing and lead times before you budget. Don’t make my mistake of assuming last year’s prices still hold.