Flipping the switch: carbon-negative and water-positive data centers through waste heat utilization

Abstract

Artificial intelligence (AI) growth poses major electricity, emissions, and water challenges. Globally, AI data centers are projected to demand Gigawatts of electricity, leading to Gigatons of carbon dioxide emissions and trillions of gallons of water consumed per year. With increasing deployment of high efficiency chip cooling, which in turn raises the waste heat temperature, data center waste heat could become a Gigawatt-scale energy resource. In this perspective, we analyze the various options for using data center heat from a thermodynamic, revenue, and emissions perspective. We show that direct air capture and thermal water purification are highly promising due to their ability to efficiently capture/avoid CO₂ while producing a valuable product. Using data center heat for other purposes such as heating, cooling, electricity conversion, or atmospheric water production, are shown to have lower potential for emissions reduction and economic benefit. We then discuss the advantages of waste heat-powered direct air capture and water production compared to incumbent carbon capture and desalination approaches. Importantly, we highlight key technological and scientific opportunities that can enable these impactful end uses. Lastly, we propose a new data center metric, the Energy Use Efficiency (EUE), which incentivizes waste heat reuse and shows that data centers with heat utilization can be carbon-negative and water-positive, addressing major sustainability challenges of AI.