euronews.com web signal

UN Chief Guterres Demands AI Firms Disclose Environmental Costs

ai infrastructure climate ai-infrastructure transparency

TL;DR

  • UN Secretary-General Guterres called for AI companies to disclose carbon, water, and land use at London Climate Action Week.
  • Data centers consumed more electricity than all but 10 countries in 2025, with coal supplying roughly 30% of that power.
  • Data center electricity use is projected to nearly double by 2030, with renewables expected to meet only half the increased demand.

Global data centers consumed more electricity in 2025 than all but 10 countries, and the UN's top official wants the AI industry to stop treating that as background noise. At London Climate Action Week, Euronews reported, Secretary-General António Guterres called for a new transparency framework requiring AI companies to disclose their carbon emissions, water usage, and land consumption, and to commit to powering their facilities with renewables by 2030. "No more hidden costs," Guterres said. "It is time to come clean."

The framing draws on a UN study released earlier in June 2026 projecting that data centers could exceed the electricity consumption of all but 5 countries by 2030, rising from roughly 1.5% of global electricity use today to nearly 3%. Water, energy use, and pollution associated with AI are expected to double within four years.

The energy mix behind that demand is the uncomfortable part. International Energy Agency figures cited in the reporting show coal currently supplies roughly 30% of data center power, while renewables account for only about 27%, with natural gas at around 26% and nuclear at 15%. Renewables are expected to meet only half of the increased demand over the next five years, meaning even aggressive clean energy commitments may not prevent the net fossil fuel load from growing in absolute terms.

What the reporting does not give you is much on enforcement. Guterres framed this as a call for voluntary disclosure and commitment, not a binding regulation. The specific companies targeted are not named, and whether the 2030 renewable commitment refers to actual clean generation or certificate purchasing goes unaddressed.

For practitioners building on cloud AI infrastructure, the direction is worth watching regardless. If consumption and associated pollution double within four years as projected, the pressures will eventually reach API pricing, data center availability, and regulatory requirements in energy-sensitive markets, whether or not formal disclosure rules materialize first.