Data center cooling infrastructure and airflow

Data Center Cooling & CFD: Designing Airflow That Works

Spetia Engineering R&D·January 30, 2026·9 min read
Key takeaways
  • 01Cooling is where data centers most often underperform; CFD simulation predicts airflow and hotspots before a rack is installed.
  • 02Hot-aisle/cold-aisle containment prevents hot and cold air mixing, the single biggest efficiency win in air cooling.
  • 03PUE (Power Usage Effectiveness) measures how much overhead cooling and power add; good design drives it toward 1.2 or below.
  • 04Rising rack densities push facilities toward liquid cooling, which changes the mechanical design fundamentally.

You can specify plenty of cooling capacity and still get hotspots, tripped equipment, and wasted energy — because cooling is about airflow, not just tonnage. Air takes the path of least resistance, hot and cold streams mix, and a rack in the wrong spot bakes while another is over-cooled. CFD (computational fluid dynamics) lets you see and fix all of this in simulation, before the facility is built.

Containment: stop the mixing

The foundational move in air cooling is hot-aisle/cold-aisle containment: arranging racks so cold supply air and hot exhaust air are physically separated, and containing one or the other so they can’t mix. Mixing is the enemy — it forces you to over-cool to compensate, wasting energy and still risking hotspots.

What CFD reveals

  • Hotspots: racks or zones that won’t receive enough cold air at design load.
  • Bypass and recirculation: cold air short-circuiting back to the units, or hot air recirculating into intakes.
  • Effect of failures: what happens to temperatures when a CRAC unit drops (the redundancy scenario).
  • Optimisation: floor-tile placement, containment, and setpoints tuned for efficiency before commissioning.

The shift to liquid cooling

As AI and HPC push rack densities far beyond what air can handle, facilities are moving to liquid cooling — rear-door heat exchangers, direct-to-chip, and immersion. This fundamentally changes the mechanical design: coolant distribution, manifolds, and leak management become central. Designing for a liquid-cooled future is now part of forward-looking data center engineering.

Cooling proven before construction

Spetia Engineering designs data center cooling with CFD verification and coordinates it in BIM, so airflow, efficiency, and redundancy performance are demonstrated in design rather than discovered on site.

Frequently asked questions

What is CFD in data center design?+
CFD (computational fluid dynamics) simulates airflow and temperature in the data hall, predicting hotspots, air recirculation, and the effect of equipment failures before the facility is built. It lets engineers optimise containment, floor-tile placement, and setpoints so the cooling actually performs at design load.
What is PUE and what is a good value?+
PUE (Power Usage Effectiveness) is total facility power divided by IT power — a measure of cooling and distribution overhead. A PUE of 2.0 means overhead equals the compute load; well-designed modern facilities push toward 1.2 or below, and CFD-driven airflow design is a major lever for achieving it.
Why are data centers moving to liquid cooling?+
AI and high-performance computing are driving rack power densities beyond what air cooling can practically handle. Liquid cooling — rear-door heat exchangers, direct-to-chip, or immersion — removes heat far more effectively, but it fundamentally changes the mechanical design, adding coolant distribution, manifolds, and leak management.