Johnson Controls has unveiled a new reference design guide aimed at data centre operators, focusing on the conversion of waste heat into functional energy. Facilities currently lose 57% of generated energy as heat, but this new approach redirects that output into cooling systems. By shifting to absorption chillers, operators can power their cooling cycles using exhaust from on-site generators rather than relying on electricity. The company estimates this method could cut cooling-related electrical demand by as much as 44%.

The potential for scaling is significant. A 1GW AI Factory could potentially support nearly 100MW of additional compute capacity through this method without requiring extra power generation. For a standard AI factory model in the United States, this recovery of capacity is valued at approximately US$18 billion in additional revenue over the facility's operational life. This strategy addresses one of the primary constraints in the industry, which is the reliance on limited grid power availability.

Beyond capacity gains, the architecture supports environmental targets. The design achieves a power usage effectiveness rating of 1.23 and operates without on-site water use, while also cutting carbon emissions associated with cooling by up to 43%. Johnson Controls has built this system on over 65 years of experience with absorption technology, previously deployed in naval and manufacturing settings. The modular design is scalable, allowing it to apply to campuses ranging from 100MW to gigawatt-scale operations.

Management emphasizes that the industry must move beyond simply seeking more power. By treating waste heat as an asset rather than a liability, operators can maximize the energy they already possess. This pivot allows data centres to meet the intense power requirements of modern AI workloads while reducing the immediate pressure on electrical grids. As infrastructure constraints continue to shape development timelines, this methodology offers a path to increase compute density within existing physical footprints.