This research challenges my understanding of energy consumption by data centres and increasing use of artificial intelligence (AI). On the plus side, it’s always good to have your ideas challenged and, in turn, to challenge the challengers!

This news was released twice. (Presumably, it didn’t get as much attention as hoped for the first time around.) A December 6, 2025 news item on ScienceDaily (rerun with a changed headline as a March 18, 2026 news item) provides a startling (to me) piece of information,

Artificial intelligence is often blamed for driving up energy use and worsening climate change, but new research suggests its overall impact on global emissions is surprisingly small. The findings even point to potential environmental and economic benefits as AI continues to expand.

Researchers from the University of Waterloo and the Georgia Institute of Technology analyzed data from across the U.S. economy alongside estimates of how widely AI is being used in different industries. Their goal was to understand what might happen to energy use and emissions if AI adoption keeps growing at its current pace.

…

A November 12, 2025 University of Waterloo news release (also on EurekAlert), which originated the news item, offers more information about the study,

…

According to the U.S. Energy Information Administration, 83 per cent of the U.S. economy is powered by petroleum, coal and natural gas, all of which contribute to climate change when burned. The study authors found that while power usage from AI in the U.S. equalled the energy consumption for all of Iceland, the amounts were not noticeable on a global or national scale.

“It is important to note that the increase in energy use is not going to be uniform. It’s going to be felt more in the places where electricity is produced to power the data centres,” said Dr. Juan Moreno-Cruz, a professor in the Faculty of Environment at Waterloo and Canada Research Chair in Energy Transitions. “If you look at that energy from the local perspective, that’s a big deal because some places could see double the amount of electricity output and emissions. But at a larger scale, AI’s use of energy won’t be noticeable.”

While this paper did not examine the effects on local economies where the data centres are located, the researchers found some encouraging results.

“For people who believe that the use of AI will be a major problem for the climate and think we should avoid it, we’re offering a different perspective,” Moreno-Cruz said. “The effects on climate are not that significant, and we can use AI to develop green technologies or to improve existing ones.”

To reach their conclusions, environmental economists Moreno-Cruz and Dr. Anthony Harding examined different sectors of an economy, the jobs within those sectors, and what portion of them could be done by AI.

Moreno-Cruz and Harding plan to repeat the study for other countries to measure the impacts of AI adoption in other parts of the world.

…

Here’s a link to and a citation for the paper,

Watts and bots: the energy implications of AI adoption by Anthony R Harding and Juan Moreno-Cruz. Environmental Research Letters, Volume 20, Number 11 (Environ. Res. Lett. 20 114084) DOI 10.1088/1748-9326/ae0e3b Published 11 November 2025 • © 2025

This paper is open access.

Acknowledgement of research shortcomings

The authors, as good academics do, critique their own work in the Discussion portion of the paper. Some of the (1) data is quite old and, in some cases, they had (2) only one source. Also (3), from the paper,

Funding disclosure

A R H and J M C were supported by funding from Google. The funder had no role in the design, analysis, interpretation, or writing of this paper. All findings and conclusions are solely those of the authors.

Google, eh? Maybe add an extra grain of salt when reading the research.

Finally, all of the data is focused on industrial and occupational use, from the paper

…

Future research could address some of the limitations identified here, incorporating dynamic effects, exploring industry-specific AI impacts, and investigating the interplay between AI-driven productivity gains and energy efficiency improvements. Such work would further refine our understanding of AI’s role in shaping future energy demand and environmental outcomes. Additionally, as more data becomes available on the real-world impacts of AI adoption across different industries, researchers can update and refine the estimates presented in this study.

It is also worth considering the potential for AI itself to contribute to solutions for energy efficiency and emissions reduction. AI technologies could play a significant role in optimizing renewable energy sources, such as wind and solar power. AI can also optimize industrial processes, increasing overall efficiency and reducing waste. Future studies should consider these aspects to provide a more comprehensive view of AI’s environmental impact.

5. Conclusion

In conclusion, our study provides a valuable starting point for understanding the broader energy and environmental implications of AI adoption across the economy. We hope to stimulate further research and informed discussion on this important topic by highlighting both the potential impacts and the areas of uncertainty. As AI continues to evolve and reshape various aspects of our economy and society, ongoing analysis and monitoring of its energy and environmental impacts will ensure sustainable development of these transformative technologies.

Perhaps, the authors also could have acknowledged that the impact of personal AI usage should also be considered for future research and when data is available.

In light of the paper’s focus on industry and occupational use, perhaps its title “Watts and bots: the energy implications of AI adoption” should have been more specific.

Thought experiment (sort of): what about personal usage?

Thanks to the authors and their paper for stimulating some new thoughts about AI and data centres, namely, that personal usage could become very important not only culturally, but, also in terms of energy consumption..

For example, research from the University of British Columbia (UBC) “The AI Genie Phenomenon and Three Types of AI Chatbot Addiction: Escapist Roleplays, Pseudosocial Companions, and Epistemic Rabbit Holes” by M. Karen Shen, Jessica Huang, Olivia Liang, Ig-Jae Kim, and Dongwook Yoon who presented it at the 2026 CHI Conference on Human Factors in Computing Systems suggests to me that personal usage (whether it’s considered an addiction or not) may constitute a higher percentage of use than is currently imagined or considered in figures in the data considered by the University of Waterloo researchers.

Note: For anyone interested in the UBC paper, it’s open access and there is an April 27, 2026 UBC news release (also on EurekAlert) for anyone who wants a brief overview of the work.