AI economic gains likely to outweigh emissions cost, says IMF

(Reuters) -Economic gains from artificial intelligence will boost global output by around 0.5% a year between 2025 and 2030, outweighing the costs of rising carbon emissions by the data centres needed to run AI models, the International Monetary Fund said on Tuesday.

An IMF report released at its annual spring meeting in Washington nonetheless noted that those output gains would not be shared equally across the world, and called on policymakers and businesses to minimise costs to broader society.

"Despite challenges related to higher electricity prices and greenhouse gas emissions, the gains to global GDP from AI are likely to outweigh the cost of the additional emissions," it said.

"The social cost of these extra emissions is minor compared with the expected economic gains from AI, yet it still adds to the worrisome buildup of emissions," it said in the report titled "Power Hungry: How AI Will Drive Energy Demand".

Takeup of AI is seen driving a surge in demand for energy-intensive data processing power in coming years, even as the world struggles to keep promises on reducing carbon emissions.

The IMF report noted that the space dedicated to server-filled warehouses in northern Virginia, which has the world's largest concentration of data centres, was already roughly equivalent to the floor space of eight Empire State Buildings.

It estimated that AI-driven global electricity needs could more than triple to around 1,500 terawatt-hours (TWh) by 2030 - about the same as India's current electricity consumption and 1.5 times higher than expected demand from electric vehicles over the same period.

The carbon footprint of that rise will in part depend on whether tech firms can keep promises to slash emissions from data centres by increased use of renewables and other means.

COULD AI LEAD TO ENERGY EFFICIENCY GAINS?

The IMF estimated that strong takeup of AI would, under current energy policies, mean a global cumulative increase of greenhouse gas emissions of 1.2%, between 2025 and 2030. Greener energy policies would limit that increase to 1.3 Gt, it estimated.

Using a figure of $39 per ton to quantify the social cost of those emissions, it put that extra cost at $50.7 to $66.3 billion - smaller than the income gains associated with the 0.5% point annual boost to global GDP it said AI could yield.

Independent analysts say the economic and environmental impact of AI will depend to a large extent on how it is put to use - and notably whether it can lead to efficiency gains in energy use or more sustainable overall consumption patterns.

The Grantham Research Institute on Climate Change and the Environment said it could even lead to an overall reduction in carbon emissions if it accelerated advances in low-carbon technologies in the power, food and transport sectors.

"But market forces alone are unlikely to successfully drive AI's application toward climate action," said Grantham policy fellow Roberta Pierfederici.

"Governments, tech companies and energy companies must play an active role in ensuring AI is used intentionally, equitably and sustainably," she said, citing the need for R&D funding and policies to address inequalities exacerbated by AI advances.

(Writing and reporting by Mark John in LondonEditing by Ros Russell)