The Expansion of AI: Potential Energy Costs and Benefits

Artificial Intelligence (AI) has taken the world by storm ever since the launch of ChatGPT in 2022. Along with its unprecedented energy demand due to the rise in data centers, concerns are mounting regarding such demand’s environmental impacts. But can the benefits of AI outweigh the costs?

ChatGPT, one of many language processing and generative AI applications, is expanding its reach across industries and driving the expansion of data centers, the facilities doing the heavy lifting to store, process, and power services for such technologies. Their energy cost is quite high. According to the IEA, data centers require exorbitant amounts of energy to function — one data center can consume almost the same amount of energy as 100,000 households combined — and research suggests that the largest data centers under construction today could consume up to 20 times as much energy.

This increasing demand threatens extensive environmental costs. The UN’s Digital Agency found that the world’s top tech companies saw a 150 percent average rise in emissions between 2020 and 2023 due to AI developments and the expansion of data centers. Furthermore, from available data in 2023 for 166 companies, they emitted a combined 297 million tonnes of CO2 per year — the equivalent of Argentina, Bolivia, and Chile’s emissions combined.

Despite the negatives, there are many potential benefits to the expansion of AI that outweigh the initial costs. A review in the Journal of Environmental and Energy Economics suggests that the expansion of AI technologies could improve energy efficiency in the long term. If fully utilized throughout the industrial sector, AI technology provides the possibility for industries to lower their carbon emissions through the optimization of production, streamlining processes, and identifying issues humans may have missed. AI may be able to foster innovation as well in industrial sectors by generating new solutions to efficiency challenges.

AI could provide the opportunity for efficiency in tech production and other energy improvements leading to cuts in everyday carbon and water use, beyond the industrial sector. In fact, one estimate shows that household CO2 consumption could be reduced by up to 40 percent with AI technology.

The Journal of Environmental and Energy Economics review most notably suggests that if the expansion of AI is done hand in hand with green energy initiatives, environmental benefits will be much greater than if data centers expand through utilizing natural gas. This is a feasible reality. Over half of the global demand for new data center projects is planned or being met by renewables. Furthermore, S&P Global reports that nuclear energy makes up 43 percent of the estimated 47.6 GW energy increase over one year in the U.S., demonstrating the U.S.’s increasing clean energy capacity.  Goldman Sachs estimates that U.S. data center power growth through 2030 will be met by 60 percent natural gas and 40 percent renewable energy sources.

In order to promote these possibilities, policy can help regulate the AI boom and promote energy efficiency.

One challenge facing the renewable energy sector is the permitting crisis. Industry professionals at the first day of this year’s Energy Imperatives Summit in Washington D.C. expressed their frustration at the timelines for energy project approval. To allow clean energy expansion in order to power expanding data centers, permitting times must be streamlined and made more efficient, in a way where environmental regulations are still followed. Revising the permitting process will be up to our government officials and lawmakers.

In addition to improving permitting timelines, government can invest in or incentivize renewable innovation. Increasing renewable energy accessibility and affordability to power data centers would allow AI to reduce global carbon emissions rather than increase them. This could be seen in the form of clean energy tax credits, performance-based regulatory objectives, or other economic incentives. Lawmakers could further incentivize firms across the energy industry to utilize AI in their everyday operations (with appropriate safeguards) to increase energy efficiency, advance greener technology, and decrease our carbon footprint by enhancing emissions reduction strategies.

Further, Tyler Norris recently released his research with Duke University regarding load flexibility, explaining that a majority of data centers run at their maximum potential power rating 24/7, though their utilization rate is on average only about 50 percent. This means that data center power consumption could be optimized, and AI could prove to be a powerful tool in doing so. Policymakers could work to incentivize data centers to increase their load flexibility and increase efficiency of these centers to prevent wasting energy.

As of now, in the U.S. there is scarce energy policy regarding the expansion of AI. Many laws concern only matters of security and not the looming energy crisis in part due to expanding data centers. Though the Federal Energy Management Program within the Department of Energy provides best practice guides and recommendations for data center efficiency methods, there is no specific law in the U.S. to regulate energy policy regarding AI expansion. In the future, the right policy will be crucial in ensuring the AI boom is navigated sustainably and ensure economic success over an energy crisis.

In conclusion, the rapid expansion of data centers and AI technology is daunting. Though it presents a challenge, it is a challenge that is worthwhile, due to its countless possible benefits. Now we must ask ourselves how to expand the AI and energy industry strategically, so that the benefits outweigh the costs.

Written by Emma Kelliher, Public Policy Intern