Recent research by Alex de Vries-Gao, a PhD candidate at the Vrije Universiteit Amsterdam Institute for Environmental Studies, suggests that artificial intelligence could use close to half of all the electricity consumed by data centers globally by the end of 2025. According to the researcher, AI accounts for up to a fifth of the electricity that data centers use.

As AI usage increases, its appetite also grows in parallel despite efficiency gains and at a fast enough clip to warrant more scrutiny. He says, “We see these big tech [companies] constantly boosting the size of their models, trying to have the very best model out there, but in the meanwhile, of course, also boosting the resource demands of those models.” 

Estimation Technique 

In light of the limited information available regarding energy consumption by companies regarding AI chip usage, the researcher uses a method called “triangulation.” Under this, using publicly available device specifications, analyst estimates, and company earnings calls, the researcher estimates the scale of AI hardware production and its associated energy use. He found that Taiwan Semiconductor Manufacturing Company (TSMC), which produces AI chips for Nvidia and AMD, more than doubled its AI chip packaging capacity between 2023 and 2024. 

To read about the latest reveal of MetaAI's AI training Manual- Permitting Flirty but Non-sexual Prompts, click here! 

Based on this growth and the power consumption of AI hardware, he concluded that AI systems likely consumed as much electricity in 2024 as the entire Netherlands. By the end of 2025, he projects that AI’s energy demand could rise to 23 gigawatts, comparable to the electricity usage of the UK.

More Transparency Needed from Tech Companies 

It’s a mystery that could be solved if tech companies were more transparent about AI in their sustainability reporting. “The crazy amount of steps that you have to go through to be able to put any number at all on this, I think this is absurd,” de Vries-Gao says. 

To read about Perplexity's upcoming agentic browser, click here!  

“It shouldn’t be this ridiculously hard. But sadly, it is,” he adds. Someone using AI tools to promote a fundraiser could generate nearly twice as many carbon emissions if their queries are processed by data centers in West Virginia rather than California, due to differences in the local energy mix, the article on the issue reads. 

Focus on efficiency and pushing consumption. 

Even as models become more efficient, their overall electricity consumption is expected to rise simply because more people will use the technology. In any case, tackling the issue will be difficult without first measuring its true scale. Data centers could use up to 9% of total electricity generated in the U.S. by the end of the decade, more than doubling their current consumption, as technology companies pour funds into expanding their computing hubs, a Reuters report references the Electric Power Research Institute saying so in May 2024.