UN urges AI firms to come clean on their environmental costs

The United Nations wants artificial-intelligence companies to stop treating the environmental bill as somebody else’s problem.

In a call amplified this week by Secretary-General Antonio Guterres, the organisation is pressing the firms behind the AI boom to disclose the carbon, water, and land their systems consume, and to switch their data centres and supply chains to clean power before the costs land on the communities least able to absorb them.

The disclosure demand rests on a fresh evidence base. A June 2026 report from the UN University Institute for Water, Environment and Health (UNU-INWEH) warns that AI’s environmental footprint is expanding rapidly across all three dimensions, and calls for urgent action to keep the technology inside what it terms planetary limits.

The report reframes a debate that has largely fixated on electricity as one about water and land as well.

The headline number is the power. Data centres are projected to draw about 945 terawatt-hours of electricity a year by 2030, the report estimates, a figure it sets against the combined annual electricity use of Pakistan, Bangladesh, and Nigeria, three countries whose populations together exceed half a billion people.

The AI estate, on that trajectory, would consume nearly triple what those three nations use put together.

Water is the quieter cost. Cooling the servers and generating the power to run them gives AI what the authors call a water footprint, and they project that AI-related water consumption could match the basic annual domestic water needs of 1.3 billion people by the decade’s end.

The land footprint, tied mostly to the space and infrastructure that power generation requires, is the third leg the report adds to a picture usually drawn with one.

None of this is presented as a reason to abandon AI, but as a reason to build it differently. The report sets out a framework for what it calls a responsible AI ecosystem, resting on transparency, efficiency by design, equity, lifecycle responsibility, global cooperation, and sustainable use.

The first of those, transparency, is the one Guterres has seized on, because almost everything else depends on companies first being honest about what their systems actually consume.

That honesty is currently in short supply, which is part of the point. Operators rarely publish granular figures for the water a given data centre draws or the emissions tied to a particular model, leaving researchers to estimate.

The push for disclosure is, in effect, an attempt to convert a field of educated guesses into something regulators and the public can actually audit.

The UN’s own framing stresses that the costs must not be quietly shifted onto vulnerable communities, the places where cheap land and lax oversight tend to draw data centres in the first place.

The warning lands amid a building of AI infrastructure already running at extraordinary scale.

US utilities are bracing for a $1.4tn capital surge to feed the data centres planned by 2030, and projects such as Meta’s gas-powered Hyperion campus in Louisiana show how readily the grid’s shortfalls are being met with fossil fuels rather than the clean power the UN is asking for.

Regulators have begun to respond, with the EU telling Big Tech to align its data centres with climate goals and a wave of startups chasing ways to curb their energy use. The report stops short of prescribing hard caps or naming companies, leaving the harder politics of enforcement to governments.

Its contribution is the framework and the figures, a way of measuring an industry that has grown faster than the tools to track it.

The UNU-INWEH report, available in full from the institute, gives those efforts a common yardstick. What it cannot supply is the disclosure itself, which remains, for now, in the gift of the companies.