Data centre power demand rises as agentic AI increases energy use

Traders Union
All News
Financial News
Data centre AI power demand

Highlights

International Energy Agency forecasts data centre power consumption rising 17 percent in 2025 and doubling by 2030, driven by agentic AI workloads.
A KAIST study finds agentic AI uses up to 136.5 times more energy per query than basic chatbots, with projected daily requests potentially increasing electricity demand by nearly 200 gigawatts.
Morgan Stanley estimates data centre annual emissions will quadruple to 1.2 gigatonnes by 2030, surpassing Japan's yearly emissions and heightening pressure on energy infrastructure.

Agentic AI raises data centre energy needs

As reported by Financial Times, the debate over AI's climate impact is sharpening as newer agentic systems begin to require far more electricity than basic chatbot interactions. Simon Stiell, the UN's climate chief, says energy-hungry AI is increasing pollution from coal, oil and gas, while calling on companies to run expanding data centres on renewable power rather than fossil fuels.

Former U.S. vice-president Al Gore argues AI emissions should be kept in perspective, saying their greenhouse gas effect remains far below that of the world's uncovered landfills. Still, both views align with the broader trend that data centre electricity use is rising quickly, even as hardware and software efficiency improve.

The International Energy Agency says data centre power consumption rises 17 per cent in 2025 alone and projects it will double from that level by 2030. A July paper from the Korea Advanced Institute of Technology finds that an agentic AI system comparable to current commercial services uses up to 136.5 times more energy per query than a basic chatbot session.

Zeke Hausfather, climate research lead at Stripe, calculates last month that his own use of Anthropic's agentic AI system, averaging about 20 prompts a day, consumes roughly a tenth of the electricity demand of an average U.S. household. He notes that he uses the latest AI tools more than most people, but corporate deployment at industrial scale and a push for mass consumer adoption could lift demand much further.

Emission outlook adds pressure on energy system

Meta's launch this month of its Muse agent across WhatsApp, Instagram and Facebook points to how quickly agentic AI may move into mainstream use. The company suggests users could delegate tasks such as selling personal belongings or processing medical records, indicating a path toward much heavier and more frequent computing workloads.

The KAIST team models a scenario in which daily AI agent requests rise to 13.7 billion, roughly matching the current number of daily Google searches. In that case, new power demand reaches just under 200 gigawatts, about half the average load of the entire U.S. power grid today.

Morgan Stanley analysts estimate this week that annual emissions from data centres will quadruple between 2025 and 2030 to 1.2 gigatonnes, exceeding Japan's annual emissions. Gore says that is a cause for concern, but also points to research suggesting advanced AI could eventually cut emissions through more efficient transport and progress in alternative proteins.

A paper published in Nature last year estimates those potential savings could reach as much as 5.4 gigatonnes, far above this year's forecast global data centre emissions. For now, however, the scale of any climate benefit remains uncertain, while the sector's rapid growth in power use and emissions is already taking shape.

In our earlier article on voter resistance to AI data center development in Texas, we examined how concerns about electricity demand, water use and higher utility bills are turning data center projects into a live political issue ahead of the 2026 governor and Senate races. We also noted that bipartisan pressure has prompted state officials to tighten oversight, including audits and pauses on some permits, as community backlash grows.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.