UN Chief Demands AI Companies Disclose Power and Water Footprints
Guterres' call at London Climate Action Week puts pressure on tech giants as US data centre electricity demand heads toward 38% of net grid consumption by 2037.
António Guterres used London Climate Action Week on Tuesday (2026-06-23) to call on every major artificial intelligence company to disclose its energy and water consumption, framing AI's expanding infrastructure as a direct obstacle to the global energy transition.3
The demand came with stark numbers. Data centre electricity consumption in the United States is projected to grow 300% over the next ten years and account for 38% of net electricity consumption through 2037, according to the National Electrical Manufacturers Association. Guterres warned that AI could by 2030 consume more power than all but five countries on Earth — and enough water annually to meet the basic needs of all 1.3 billion residents of sub-Saharan Africa.3
For energy markets, the implications are not marginal. Ireland offers a near-term illustration: data centres now consume a fifth of the country's total electricity, more than its urban homes use, in a country with one data centre for every 42,000 people. That concentration has already strained grid procurement and forced state planners to throttle new connection approvals.2
Guterres tied the disclosure push to a broader emissions target. Global greenhouse gas emissions must peak immediately, fall steeply this decade, and reach net-zero by 2050, he said, calling on G20 nations to lead the transition away from fossil fuels. The AI sector's rising power appetite complicates that arithmetic: the more electricity hyperscalers draw, the harder it becomes to retire coal and gas plant on schedule without adding equivalent firm renewable capacity first.3
The disclosure call is framed as a prerequisite, not a solution. Guterres proposed a framework for AI companies to report impacts as part of the UN's plan for global energy independence, but stopped short of a binding international reporting standard or a compliance mechanism, which limits the immediate market impact of Tuesday's (2026-06-23) statement.3
Without verified figures on energy sources, cooling loads, and water withdrawal, grid operators and investors cannot accurately model the demand curve that AI infrastructure will place on power systems. Voluntary disclosure frameworks in corporate sustainability have historically produced inconsistent and non-comparable data — the same risk the UN's call implicitly acknowledges.3
Power developers and equipment suppliers have moved to position in front of the buildout. The broader energy infrastructure sector has been repricing as hyperscalers secured long-term power agreements, including natural gas contracts and nuclear supply memoranda, to guarantee firm baseload for campuses that need uninterrupted high-voltage supply.1
ICE Endex TTF front-month gas was trading at €40.60 on Thursday (2026-06-25) — a market already sensitive to any signal that industrial load growth could accelerate faster than renewable build rates absorb. If US data centre demand does reach 38% of net consumption by 2037, American gas burn for power would face structural upward pressure, feeding back into global LNG pricing and, through the Atlantic arbitrage, into European hub dynamics.3
Whether disclosure requirements translate into any constraint on AI power consumption is a separate question. Tech companies have argued that efficiency gains — more computing per watt — could offset absolute demand growth. But making models less energy-intensive may not reduce overall environmental impact if lower cost per query simply encourages more queries, a rebound effect with no clear ceiling.2
The shape of that demand curve matters more than Guterres' speech itself. If power planners are working from projections of 38% US net consumption share by 2037, the current pace of transmission infrastructure permitting and renewable project development does not close the gap.3