Power Usage Effectiveness (PUE) is the industry metric for a data center's overhead: total facility energy divided by the energy actually consumed by IT equipment. A PUE of 1.0 would mean every watt goes straight to computing; real facilities sit higher because cooling, power conversion and lighting all draw extra power. This sim recomputes PUE live from three real levers.
P_IT = baseline × load
P_cool = P_IT × coolingFactor(tech, T_out)
PUE = (P_IT + P_cool + P_fixed) / P_IT
- Cooling technology — traditional air-cooled CRAC units fight heat with brute-force airflow and are the least efficient. Direct liquid cooling pulls heat away at the chip with coolant, cutting overhead sharply and staying efficient regardless of the weather. Free-air economizers pump in outside air directly — nearly free when it's cool outside, but they switch back to mechanical cooling once ambient temperature crosses a threshold, so their overhead jumps sharply in hot weather.
- Outside air temperature — hotter outside air makes every cooling method work harder, but the free-air economizer is by far the most sensitive to it.
- IT load — more computing work means more heat to remove, but PUE itself is mostly load-independent since both IT power and cooling power scale together.
- Renewable share — the fraction of total facility power drawn from the on-site solar array and wind turbine instead of the grid. It doesn't change PUE, but it changes the carbon footprint per watt — the other half of what makes a data center "green".
Real-world relevance: hyperscale operators report PUE near 1.1–1.2 by combining liquid cooling with free-air economization in temperate climates, then offset the remaining grid draw with power-purchase agreements for wind and solar — exactly the two dials this simulation lets you turn independently.