A ChatGPT search consumes 10 times more energy on average than a Google search, according to Goldman Sachs. According to Joule, OpenAI’s GPT-3 model took 1,287 MWh of energy to train, the rough equivalent of the annual energy consumption of about 130 U.S. households.
Artificial intelligence (AI) requires much more energy than traditional computing, a factor often overlooked by users.
Many people have taken to using ChatGPT as a replacement for Google. Aside from some obvious concerns about using AI, known for often generating false information, as one’s primary source of information, people must consider the energy costs of their searches.
As businesses rush to implement AI in every aspect of their workflows, it becomes necessary to consider if the often only incremental efficiency improvement from AI is worth the costs to the business and the environment.
The vast majority of the world’s energy is still produced by burning oil and coal. According to Our World in Data, energy consumption accounts for three-quarters of global greenhouse gas emissions.
Climate change and other environmental concerns are of tremendous relevance in the U.S. after newly inaugurated President Donald Trump signed an executive order on Jan. 20 to withdraw the U.S. from the Paris Agreement.
Another step in the wrong direction came with Google’s 2024 Environmental Report. Google stated a 48% increase in greenhouse gas emissions since 2019 despite having a net-zero emissions goal by 2030, citing increased data center energy consumption as the primary cause.
Data centers are facilities that house the infrastructure necessary for a large AI model’s computational, storage, and cooling demands.
In the same report, Google announced, “As we further integrate AI into our products, reducing emissions may be challenging due to increasing energy demands from the greater intensity of AI compute, and the emissions associated with the expected increases in our technical infrastructure investment.”
It is evident that businesses that have attempted to implement AI wherever possible have to accept the energy tax that comes along with advanced computing.
For the businesses that are a little late to the game, take Google’s energy and emission struggles as a warning. AI is costly, and disrupting the existing workflow with a flashy AI tool may not be the most promising solution.
Individuals who regularly use AI in their daily lives, at work, or at school must also consider the repercussions of increased energy usage. The complete elimination of AI from one’s life may be a bit drastic. Still, if people were more intentional, only using ChatGPT when Google would not suffice could bring the world closer to a future of energy efficiency and net-zero emissions.
*This editorial reflects the views of the Editorial Board and was written by Audrey Finigan. The Editorial Board voted 12 in agreement and 1 somewhat in agreement.
