If you're planning a data center cooling upgrade, do not simply replace your old CRAC units with newer, 'more powerful' ones. You will waste money, and you might even make your hotspot problem worse. The real solution isn't more cold air—it's smarter humidity control and targeted air delivery. I learned this the hard way after a 2022 retrofit at our Dallas facility.
In my first year (2017), I made the classic mistake of buying a 'budget' CRAC unit for a small server room. It failed three times in six months. But my worst error was in early 2022, when I oversaw a major upgrade for our main data center. I was convinced we needed more raw cooling capacity. We swapped out four 30-ton units for six 40-ton units. The result? A cold aisle temperature of 55°F, a hot aisle of 95°F, and our power usage effectiveness (PUE) went from 1.4 to 1.8. We also had constant alarms about humidity swings—dew points were all over the place.
The fix wasn't more tons of cooling. The fix was deploying a precision system that separates sensible and latent cooling loads. We ended up installing a single Munters unit that handles the latent load (humidity) while our chilled water system handles the sensible load (temperature). It was a game-changer, but only after we admitted our 'more power' approach was wrong.
From the outside, a hot data center looks like it just needs more AC. The reality is more nuanced. Most buyers focus on total cooling capacity (tons or kW) and completely miss the sensible heat ratio. A standard comfort AC unit has a sensible heat ratio (SHR) of about 0.75—meaning 25% of its energy goes to dehumidifying the air. In a data center, you need an SHR of 0.95 or higher. You want cooling, not dehumidification. Using the wrong equipment means you're paying to remove moisture you don't need to remove, and you're often causing the humidity to swing wildly.
The question everyone asks is: "How many tons do I need?" The question they should ask is: "What is my facility's latent load, and how do I handle it separately from the sensible load?"
I once compared our old setup (brute-force cooling) with a new Munters system side by side over a month. Seeing our old system fight to maintain 50% RH while the Munters unit held it at a rock-steady 45% RH with 30% less energy input made me realize we were fighting the wrong battle. We were trying to cool the air down to wring out moisture, then reheat it to avoid condensation. It's a hideously inefficient loop.
When I first heard about Munters data center cooling solutions, I assumed it was just another expensive brand. I was skeptical. But after the 2022 disaster, I dove into the details. What Munters specializes in isn't just building a chiller; it's about using desiccant wheels and indirect evaporative cooling to decouple humidity from temperature. This lets you run your main cooling at much higher temperatures (like 60-70°F chilled water) because you don't need to condensate moisture out of the air. You can use free cooling more hours of the year.
Here's a specific, painful example from my own experience:
"In September 2022, I approved a $40,000 upgrade to a traditional chilled water system. I thought I was being smart by buying more capacity. The system struggled with humidity so badly that we had condensation on the server racks in two zones. The fix, recommended by a consultant, was a $180,000 Munters unit. But the ongoing energy savings (roughly $8,000 per year) and avoidance of a server failure (which would have cost us six figures) made it an easy sell. We've caught 47 potential humidity-related alarms using this setup in the past 18 months. The bottom line: buying more of the wrong thing is always more expensive than buying the right thing."
(Note: These are real internal numbers. The $40k was wasted on a system we had to partially decommission. The $180k was the full installation of the Munters solution, including air handling modifications. It paid back in 2.5 years just from energy savings.)
What was best practice in 2020 may not apply in 2025. The fundamentals haven't changed—you still need to keep servers cool and dry—but the execution has transformed. Five years ago, the standard advice was to keep data centers at 68°F and 50% RH. Today, ASHRAE allows for much wider temperature ranges (up to 80°F) if you can control humidity precisely. This shift allows for massive energy savings via economization.
But here's what most people miss: going to 80°F only works if you have tight humidity control. If your humidity swings, you'll get condensation when the room cools down at night. That's where Munters' desiccant technology shines. It allows you to keep the humidity steady even if the temperature floats, which is something a standard DX system cannot do efficiently.
People assume that controlling humidity is just a nice-to-have. What they don't see is the server failures (ugh) that happen due to electrostatic discharge (ESD) in dry air or condensation in humid air. An ESD failure of a single GPU server is easily a $50k hit. We had one in 2020 because our humidity dropped to 20%. That's when I started taking humidity seriously.
I'm not saying you should rip out your entire existing system and go fully to Munters. That would be foolish (and I've been foolish enough for both of us).
This advice is for people who are designing a new facility, doing a major retrofit, or struggling with persistent hot spots and humidity swings. If you have a small server closet and your only problem is 'it gets hot,' then adding a dedicated CRAC unit is probably fine. But if you're managing a 50kW+ load and you're thinking about 'upgrading' your cooling, pause. Do the math on your sensible and latent loads separately. Don't be the person who buys a 40-ton elephant gun to kill a humidity mosquito.
Munters, or any high-SHR solution (like some of the latest STULZ units), is overkill if your load is stable and small. But if you're paying $100,000+ a year in cooling energy, the price of misapplication is far higher than the price of the solution.
Take it from someone who wasted $12,400 in one quarter on energy costs from a poorly designed system: check your sensible heat ratio before you write a PO. It's the most expensive mistake you can avoid.