Johnson Controls has published a second AI Factory Reference Design Guide, this one centered on air-cooled chiller architectures for data centers and “AI factories” at industrial scale. The guide is aimed at designs scaling up to a 1 GW facility and is intended to be repeatable across regions where water availability, noise limits, and heat island effects can constrain traditional cooling plant choices.
The new AI Factory Reference Design Guide for Air Cooled Chillers follows Johnson Controls’ earlier water-cooled chiller guide released in February. Johnson Controls also says additional guides covering absorption chillers and direct-to-chip liquid cooling are planned.
Technically, the guide describes a “comprehensive thermal cooling architecture” built around high-efficiency air-cooled YORK centrifugal chillers, including YDAM and YVAM models. It also incorporates fan coil walls (FCWs) and coolant distribution units (CDUs) to support mixed environments where both air-cooled and liquid-cooled IT loads have to be served from the same overall thermal chain.
For engineers looking for concrete sizing and operating targets, the document includes sizing references for 220 MW compute clusters, along with recommended design temperatures and operating conditions across each stage of the thermal chain. Johnson Controls frames the intent as a blueprint for handling rising cooling-loop temperatures, the power draw required for cooling, efficiency losses associated with heat islands around air-cooled chiller plants, and limited water availability.
Several quantified outcomes are tied to the reference architecture. Johnson Controls states the design can support zero-water cooling by eliminating cooling towers, which it says would save more than 12 million gallons of water per day. It also reports a potential 32% improvement in annual energy consumption through “intelligent utilization of redundant chillers,” plus 20 MW of peak power savings by quantifying and mitigating heat island effects for air-cooled chiller plants.
Other outcomes Johnson Controls lists include returning up to 50 MW “to the AI factory” by implementing bifurcated loops for air- and liquid-cooling systems, and improving coefficient of performance (COP) by 30% while using 27% fewer chillers by raising chilled-water temperature to support warm-water Technology Cooling System (TCS) loops. Those are big numbers on paper, but they also imply tight controls integration and careful commissioning, especially when redundant chiller strategies and elevated water temperatures intersect with availability targets.
“At gigawatt scale, AI factories require a fundamentally different way of thinking about infrastructure,” said Austin Domenici, president, Johnson Controls Global Data Center Solutions. “The future requires designing integrated systems that can scale predictably, perform efficiently and adapt as technology evolves.”
Source: Johnson Controls
