Turn a PUE target into real numbers. Enter your IT load, pick a cooling approach and set your energy price to get facility power, cooling + distribution overhead, annual energy and cost — plus how much a better PUE saves against a legacy 1.6 air-cooled baseline.
PUE (Power Usage Effectiveness) is the single number that tells you how much of your utility bill actually reaches the servers. This tool turns an IT load and a cooling approach into facility power, overhead, annual energy and the money a better PUE saves.
Every step is a simple, labelled multiplier so you can see exactly where the numbers come from and match them to your site.
AI clusters run near-continuously at very high density, so cooling efficiency dominates the operating bill. At a 1.1 PUE a liquid-cooled hall spends almost every watt on compute; at a 1.6 PUE a legacy air-cooled room burns 60% extra on cooling and distribution before a single token is generated. Across a multi-megawatt cluster running around the clock, that gap is millions of dollars a year — and megawatts of grid capacity you have to secure up front.
The biggest lever on PUE is cooling. Moving from legacy air (≈1.6) to rear-door heat exchangers or direct-to-chip liquid cooling (1.1–1.25) removes most of the overhead. Higher coolant and supply temperatures, free cooling, efficient UPS and power distribution, and hot/cold-aisle containment do the rest. Because overhead scales with IT power, the savings compound as the cluster grows — which is why locking an efficient cooling design early is one of the highest-return decisions in an AI build.
This calculator shows the annual saving of each PUE step against a 1.6 baseline so you can prioritise the changes that pay back fastest, then take the target into a real mechanical and electrical design.
The rating bands the calculator uses and the cooling approaches that typically land in each. Figures are planning bands; your real PUE depends on climate, coolant temperatures, load factor and design.
| PUE | Rating | Typical cooling |
|---|---|---|
| ≤ 1.20 | Excellent | Direct-to-chip liquid (best-in-class) |
| 1.20–1.35 | Good | Modern liquid / hybrid |
| 1.35–1.50 | Fair | Efficient air-cooled |
| > 1.50 | Poor | Legacy air-cooled |
PUE (Power Usage Effectiveness) is total facility power divided by IT power, so 1.0 is the theoretical ideal. A PUE at or below 1.2 is excellent — typical of direct-to-chip liquid-cooled AI halls. 1.2–1.35 is good, 1.35–1.5 is fair, and anything above 1.5 (legacy air-cooled rooms run 1.6–1.8) is poor and expensive to run at AI density.
PUE = total facility power ÷ IT power. If your servers draw 1,000 kW and the whole facility draws 1,250 kW once cooling, power distribution and losses are added, your PUE is 1.25. The overhead — everything that is not IT — is IT power × (PUE − 1), so at 1.25 PUE a 1,000 kW IT load carries 250 kW of overhead.
AI clusters are dense and run near-continuously, so cooling efficiency dominates the bill. Direct-to-chip liquid-cooled AI data centers reach 1.1–1.25; a well-run air-assisted hall lands around 1.3–1.4. Above 1.5, cooling and distribution burn 50%+ extra on top of every watt of compute — which across a multi-megawatt cluster is millions of dollars a year.
The biggest lever is cooling: moving from legacy air (1.6) to rear-door heat exchangers or direct-to-chip liquid cooling (1.1–1.25) removes most of the overhead. Higher coolant and supply temperatures, free cooling, efficient UPS/distribution and hot/cold-aisle containment all help. This calculator shows the annual saving of each PUE step against a 1.6 baseline so you can prioritise.
Our solutions engineers size cooling, power and compute together to hit your PUE target, and confirm real lead times — no payment, no commitment, quotes back in about one business day.
Our engineers scope power, cooling and compute together. No payment, and quotes come back in about a business day.