Data Center Cooling Solutions
Fan arrays, chilled water coils and N+1 redundant design built for high-density racks and zero-downtime facilities.
Data center cooling has one job that matters more than any other: keep every rack within its temperature and humidity spec, at every hour of every day, with no single point of failure. Mainstream builds the fan arrays, replacement coils and redundant system designs that data center and colocation facilities use to do exactly that—read our full breakdown of data center cooling methods for the deeper technical comparison, or keep reading for how we approach high-density and redundant designs specifically.
Whether you run a single-suite colocation facility or a hyperscale campus, the same design questions come up: how much spare cooling capacity do you carry, how do you avoid hotspots as rack density climbs, and how fast can a failed component be replaced without an outage. The sections below walk through how Mainstream answers each one with fan arrays, custom coils and redundant system design.
High-Density Data Center Cooling
Not every rack needs the same cooling approach, and the right answer changes with power density:
- Under ~5 kW/rack — standard air cooling (CRAC/CRAH units or fan arrays) handles the load comfortably.
- ~5–15 kW/rack — air cooling still works, but airflow management (hot/cold aisle containment, redundant fan capacity) becomes important to avoid hotspots.
- 15–30+ kW/rack — typical of dense AI/HPC deployments — usually needs direct liquid cooling or immersion at the rack, often paired with chilled water for the facility-level heat rejection.
Mainstream’s IntelliCUBE fan arrays are built with motors up to 30 HP and total static pressure up to 14″, and IntelliWALL configurations can move roughly 500,000 CFM of total airflow—enough headroom to keep air-cooled designs viable well into the higher end of that range before a facility needs to move to liquid.
Density alone is not the whole story—how well the room manages airflow matters just as much. Hot and cold aisle containment keeps supply air from mixing with exhaust heat before it reaches the next rack, and it is often the difference between a facility comfortably running at 10 kW/rack on air cooling and one fighting hotspots at the same density. Fan arrays help here too: because capacity is split across many smaller fans instead of one large unit, airflow can be tuned zone by zone rather than as a single all-or-nothing setting.
Chilled Water Coils for Data Centers
Facility-level heat rejection still comes down to a coil moving heat out of water or refrigerant and into air. Mainstream builds custom-sized chilled water and DX coils as exact-match replacements for existing air handlers, with field measuring to guarantee fit and fast lead times for facilities that can’t afford an extended outage to get a coil swapped.
Chilled water systems use a central plant and water-cooled heat exchangers, which scales well for larger facilities running multiple air handlers off the same loop. Direct expansion (DX) systems use refrigerant-based, more self-contained units, which can be simpler to deploy for smaller or edge sites. Either way, a coil that is even slightly off-spec from the original forces a facility to compromise on airflow or temperature control—which is why Mainstream field-measures before building a replacement rather than shipping a close-enough part.
Redundant Fan Arrays (N+1)
Because a data center can’t tolerate a single fan failure taking down a rack row, cooling capacity gets designed with spare capacity built in—commonly described as N+1 or N+2. Fan arrays are what make that practical: instead of one large fan that is either running or failed, capacity is split across many smaller direct-drive fans, so losing one still leaves the array at or near full capacity. See our full explanation of N+1 and N+2 redundancy for how the math works.
Redundant capacity only protects uptime if it is actually maintained and tested. Best practice is to rotate which units are running lead versus lag on a schedule, so spare capacity does not sit idle long enough to fail quietly, and to monitor run hours and alarms across the array rather than just the units currently in service. That operational discipline is what turns an N+1 design on paper into real protection against downtime.
Why Mainstream
- AMCA Certified Fan Performance — published ratings you can design against, not estimates.
- 30+ years in business — engineering fans and coils for commercial and industrial HVAC since the early 1990s.
- Expedited lead times — built for emergency and time-critical retrofit projects.
- Made in America — manufactured and supported domestically.
Every fan array and coil we build is sized to the specific static pressure, airflow and space constraints of the facility it is going into—not sold as an off-the-shelf part. For a data center where an oversized or undersized system means either wasted energy or a hotspot risk, that fit is the whole point.
If you are planning a new build, a retrofit, or just trying to get ahead of rising rack density before it becomes a hotspot problem, the right starting point is a real assessment of your airflow, redundancy and coil condition—not a generic spec sheet. That is what the form below is for.