Energy demands will be a growing concern for AI technology

Digital and artificial intelligence (AI)-driven sustainability management is an emerging trend that sees AI implemented to, for example, integrate climate risk analytics with incident management data for better management of extreme weather events.

More regulators are mandating real-time emissions data disclosure, pushing companies to adopt automated monitoring solutions. Industrial firms are leveraging digital twins for energy efficiency and climate risk scenario modeling. Overall, AI and advanced analytics are being integrated into sustainability reporting, carbon footprint tracking, and supply chain optimization.

But there’s a downside to the use of AI and digital solutions, which is the energy data centers consume to store cloud data and power artificial intelligence. Data centers with hundreds of thousands of computers need to be kept cool with electricity and sometimes chilled water, putting a strain on energy and water resources.¹ AI uses so much electricity that Google reported its greenhouse gas emissions jumped 50% since 2019.² Each ChatGPT query is routed to a data center and fed into an AI system, using as much electricity as could power a lightbulb for 20 minutes.

Tech companies Google, Microsoft, and Meta have pledged to reach at least net-zero carbon emissions by 2030, and Amazon has pledged to reach net-zero emissions by 2040. All four companies have pledged to be water positive by 2030, meaning they'd replenish the environment with more water than they consume for operations.³ Microsoft has taken an extra pledge to be carbon negative by 2030, but in their sustainability report released in May 2024, they revealed emissions increased by 29% from 2020 mainly due to data center construction to support AI workloads.⁴

In a breakout session on sustainability at the recent Wolters Kluwer Enablon Sustainable Performance Forum (SPF) 2025 event in Chicago, Stuart Neumann, Vice-President of Advisory Services at Verdantix, said that, while these tech firms have ambitious carbon reduction targets, the arrival of AI has increased their emissions trajectories. Neumann added there is no immediate solution to this issue, and the next two to three years will likely see increasing emissions for these firms. He emphasized the difficulty of supporting AI data centers with renewable energy alone and said nuclear power might be necessary to achieve the required scale.

Using nuclear power to generate enough energy to feed data centers and AI workloads presents another burden on natural resources, as nuclear power plants (NPPs), like data centers, typically rely on water for cooling purposes. A typical NPP consumes 20% to 80% more water than coal-fired energy facilities with similar capacity, resulting in daily water consumption of between 35 to 65 million liters.⁵ In response to global water shortages, NPPs are looking for more sustainable and less resource-intense ways to cool facilities while still maintaining (and even enhancing) operational efficiency.

No matter how you slice it, AI eats and drinks a lot of energy and water, and producing more energy to power AI also requires a lot of natural resources.

Most of us may not have anticipated how ubiquitous AI would quickly become and how easy it would be to utilize chatbots for assistance with research, homework, coding, or to create videos and images.  To keep up with the demand, Apple recently announced plans to spend $500 billion over the next four years to construct new data centers. Google anticipates spending $75 billion on AI infrastructure in 2025 alone.⁶

Currently, 4.4% of all energy in the U.S. goes toward data centers. And given the current trajectory of AI development and accessibility, our AI usage is the lowest it will ever be. Projections from the Lawrence Berkeley National Laboratory in December 2024 predict that by 2028 more than half of electricity going to data centers will go toward powering AI, meaning that AI alone could consume as much electricity annually as 22% of U.S. households.⁷

The looming question is how do we establish a balance in the use of AI and cloud-based solutions to help all types of companies (not just tech firms) achieve their sustainability goals, while also using AI to assist individuals in their everyday lives?