Energy demand debate turns to wasted heat and data center efficiency
Kathleen McGinty argues that efficiency and heat recovery can unlock energy capacity faster than building new supply alone.
By Daniel Okafor · Business Editor
3 min read
Rising electricity demand, including from AI data centers, is sharpening attention on energy that is already produced but lost before it does useful work. Kathleen “Katie” McGinty, Johnson Controls’ vice president and chief sustainability and external relations officer, argues that waste heat and building efficiency should be treated as core parts of energy supply.
In a Fortune commentary, McGinty wrote that energy planning has long centered on adding power plants, pipelines and transmission. She said new supply remains necessary, but argued that efficiency upgrades can add usable capacity faster than large infrastructure projects that require permitting, financing and construction.
Waste heat as unused capacity
The U.S. Department of Energy says industrial systems lose 20% to 50% of their energy as waste heat. McKinsey has estimated that more than 3,000 terawatt-hours of usable waste heat goes uncaptured around the world each year.
McGinty compared that figure with U.S. electricity use, citing Energy Information Administration data, and said the untapped energy is roughly three-quarters of annual U.S. electricity consumption. She described that unused heat as a large resource already inside the energy system.
Heat accounts for nearly half of final energy consumption globally, according to the International Energy Agency. McGinty argued that once heat is generated, companies and policymakers too often fail to count it as an asset that can be recovered or reused.
AI data centers add pressure
The IEA projects that data center electricity demand will more than double by 2030 to about 945 terawatt-hours, which the agency says is close to Japan’s current electricity consumption. S&P Global, citing IEA projections, reported that data centers will account for more than 20% of electricity demand growth in advanced economies this decade.
McGinty said much of the debate over AI’s power needs assumes that new generation is the main answer. She argued that part of data center demand comes from supporting systems, particularly cooling, rather than computing itself.
Average power usage effectiveness across the data center industry is about 1.5 to 1.6, according to data cited by McGinty. That means about one-third of total energy use goes to non-computing functions, she wrote.
McGinty said some operators have shown they can meet annualized PUE targets below 1.3 while removing evaporative water from cooling. She said that level of performance can cut non-compute energy use by half and reduce reliance on local water resources.
Faster than new construction
McGinty pointed to absorption chillers as one technology that can use heat rather than electricity for cooling. In data centers and similar facilities, she wrote, those systems can reduce chiller electricity use by 90% while putting waste heat to work.
She also cited estimates that waste heat recovery could save tens of billions of dollars annually worldwide. Heat pump technology, she wrote, has been shown to lower energy bills by 32% and reduce emissions by 60%.
The central argument is timing. McGinty said power plants and grid expansions can take years or decades, while efficiency improvements and thermal integration can often be installed in months.
She said the energy system still needs more generation and transmission to meet long-term demand. Her case is that efficiency, thermal recovery and smarter system design should move from secondary measures to priority resources because they can free capacity already being produced.
This story draws on original reporting from Fortune.