Nowadays, it’s hard to find digital tools and services that don’t incorporate generative artificial intelligence (AI). Between Google urging you to “write with Gemini,” ChatGPT use becoming increasingly commonplace, and individuals entering prompts in AI programs to generate silly images, AI is everywhere. And huge private-sector investment means the technology is here to stay.

Unfortunately, there’s potentially a pretty big downside: AI requires huge amounts of electricity and water.

One ChatGPT query uses 10 times more electricity than a standard web search, and when this demand is met with fossil fuels, it adds up unbelievably fast. A recent study found that ChatGPT produces carbon dioxide emissions equivalent to taking 260 flights from New York City to London.

And that’s on a monthly basis. Yikes!

Water usage is also nothing to sneeze at: In 2023, the United States alone used 66 billion liters to cool data centers, which is more than 346,000 American households use in a year. Current projections show that 2027 AI demand could result in water use equivalent to six times Denmark’s annual water withdrawals.

Double yikes. But what makes it so resource-intensive in the first place?

Running AI tools and programs requires a lot of computing power, as does training their complex underlying models. The data centers built to serve as hubs for these computationally intensive processes require a lot of energy and use a great deal of water to cool hardware and prevent overheating. Because AI is consistently fine-tuned with additional prompts and inputs, the energy and water demands persist long after a model is deployed.

While the major companies investing in AI have made strong commitments to achieving net-zero and water-positive operations, their AI ambitions are currently derailing their progress. In 2024, both Microsoft and Google reported major emissions increases in their annual sustainability reports, with Microsoft’s up 29% from 2020 levels and Google’s up 48% from 2019. Both companies’ water withdrawals have also consistently been on the rise.

We’re not under any delusions here: so long as it’s profitable for companies and the excitement around AI persists, it will continue to grow in the foreseeable future. And who knows? Some of this growth could eventually benefit the climate—after all, AI can help map deforestation, predict climate emergencies, and support a slew of other climate solutions.

In the meantime, tech companies must meet the electricity demands of data centers with clean energy and center water tracking and management in data center operations. A little transparency could also reduce demand—AI tools are convenient and easy to use, but the public is largely unaware of their environmental impact.

Let’s do what we can to push for a better status quo!