Original title: AI worsens climate change, integrated assessment shows
Huiying Ye, Richard S. J. Tol, Fangzhi Wang · August 2026
Artificial intelligence affects the planet in complicated ways, but researchers lacked a single model to measure the total impact. To fix this, scientists added AI's effects on pollution, economic growth, and climate damage into a major climate-economy model. They compared two possible paths: one where AI acts like normal information technology, and one where it transforms society like the Industrial Revolution. They discovered that developing AI is net pollutingCauses more pollution overall than it prevents or avoids. overall. With today's weak climate action, tech-like AI adds 0.1 degree C of warming by 2100, while revolutionary AI adds 0.8 degree C. These climate harms cancel out about one-fifth to one-quarter of AI's economic gains. Cutting emissions protects future AI investment, showing climate action helps tech grow.
Imagine you are deciding whether to buy a more efficient delivery van for your small business. The van saves you time and money, but it still burns gas, and you worry about the extra pollution. You think the fuel cost is small compared to the savings, but it adds up over years. Now suppose you switch to cleaner fuel for all vehicles. That fuel costs a bit more, but it lets you keep the van, and the pollution drops so much that your total climate bill is lower. In this paper, the AI is the van, the fuel is energy use, and the clean fuel is emissions abatement. The researchers show that cutting emissions lets you keep AI's economic gains without wrecking the climate.
Policymakers and investors now have a tool to weigh AI's economic benefits against its climate costs. The finding that cutting emissions actually allows more AI investment — not less — changes how we plan for both. Anyone deciding whether to fund AI or climate action sees that they are complements, not rivals.
The abstract doesn't say explicit limitations. It does cover that the results are calibrated to evidence and depend on distinguishing two AI types, so real AI may not split so cleanly, and the temperature and cost figures are model estimates, not measurements. It also does not address other AI impacts on climate, such as AI used for climate science.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
AI worsens climate change, integrated assessment shows — Huiying Ye, Richard S. J. Tol, Fangzhi Wang
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.