Fears About Skyrocketing Energy Use By AI Are Unfounded, New Report Finds

WASHINGTON—Surging interest in artificial intelligence (AI) has sparked a wave of critical speculation that it will consume massive amounts of energy, which will accelerate carbon dioxide emissions with a potentially devastating environmental impact. But as with similar panics around past technologies, many of the early claims about AI’s energy consumption are inflated and misleading, according to a new report by the Center for Data Innovation.

“Groups that oppose AI, whether from honest misunderstandings of the evidence or intentional cherry-picking of the facts, continue to push the narrative that AI’s energy footprint is growing out of control,” said Daniel Castro, director of the Center for Data Innovation and author of the report. “It turns out the evidence behind those dire claims is vastly overstated. Given the enormous opportunities to use AI to benefit the economy and society—including transitioning to a low-carbon future—policymakers and the media need to carefully vet wild claims of its environmental impact.”

The Center’s report reviews frequently repeated claims in the debate over AI’s energy usage, highlighting early missteps that some researchers have taken in their analyses and tracing how false conclusions have shaped the policy conversation on this issue.

For example, a 2019 study purportedly found that training a large language model (LLM) generated 626,155 pounds of CO₂ emissions, roughly equivalent to a traveler taking 300 roundtrip flights from the East Coast to the West Coast. Yet the report proved to be incorrect: The actual emissions were 88 times smaller than the sensational estimate. However, alarming findings like this one often continue to be repeated long after they are disproved.

The Center’s report highlights many other false narratives, detailing where energy forecasts tend to go wrong, such as ignoring that the energy use of AI is limited by economic considerations or that the rate of performance improvements in AI will decline over time.

The Center’s report concludes with recommendations that policymakers address concerns about AI’s energy consumption by taking the following steps:

• Develop energy transparency standards for AI models.
• Seek voluntary commitments on energy transparency for foundation models.
• Consider the unintended consequences of AI regulations on energy use.
• Use AI to decarbonize government operations.

“Just as early predictions about the energy footprints of e-commerce and video streaming ultimately proved to be exaggerated, so too will the dire estimates about AI likely be wrong,” said Castro. “It’s reasonable to take stock of AI’s energy use, but policymakers should be careful not to overreact to misleading narratives about it. There are measured steps policymakers can take to ensure AI is part of the solution, not part of the problem, when it comes to energy use and the environment.”

Read the report.