The International Monetary Fund has said that the economic gains of AI will outweigh the cost of its carbon emissions.

The IMF estimated that under current energy policies, AI’s needs between 2025 and 2030 will accumulate 1.7 gigatons of additional CO2 emissions. If renewables are adopted more heavily, then this could be reduced to a 1.3GT increase.

The 1.7GT is comparable to Italy’s energy-related greenhouse gas emissions over five years, the IMF said.

Even at its lowest estimate, this is still a bump in carbon emissions at an inflection point for the battle against disastrous climate change.

Regions – often the most deprived and least likely to benefit economically from AI – are already being ravaged by climate change-related disasters like floods, fires, and worsening natural disasters.

The IMF is comparing these carbon emissions to the projected boost in global output of 0.5%, saying that the “social costs” of AI’s added emissions is “minor compared with the expected economic gains from AI, yet is still adds to the worrisome buildup of emissions.”

The IMF said that the “costs” of these green house gas emissions is equal to $50.7 to $66.3 billion, or £39 per ton, which is smaller compared to the expected economic gains of AI.

Still, the IMF did point to some flaws in its own equation. Not everyone will benefit economically from AI, and climate change has devastating consequences that will not be directly reflected just economically: it will lead to rising sea levels, the extinction of species, the pollution of air, and is already directly contributing to people’s deaths through unprecedented heat.

Recommended reading

• IMF Warns GenAI Will Disrupt Jobs and Widen Inequality
• AI Energy Demand to Quadruple by 2030
• Data Centre Emissions Could Be 600% Worse Than Big Tech Claims

The use of £39 per ton as the economic cost of carbon emissions cited by the IMF is actually lower than many scientist’s and economics say is realistic, and it lacks the regional reality of these economic ‘costs’ and ‘gains’.

The IMF estimated that the global electricity consumption of AI could reach 1,500 TWh by 2030, which would be equivalent to India’s total demand. By 2030, this will result in a 1.2% cumulative increase in greenhouse gas emissions under current global energy policies.

It truly remains to be seen if big tech companies investing in mass data centres can mitigate the energy consumption of these servers sooner rather than later, and if AI can be deployed to improve the energy efficiency at such a degree as to mitigate its emissions impact and energy needs.