A peer-reviewed study shifts the debate: what the big cloud providers sell to the fossil fuel industry weighs much more than the electricity for their servers.
The debate on the climate footprint of artificial intelligence focused on electricity consumed by data centres. A peer-reviewed study, published in Nature Group's npj Climate Action magazine and relayed on September 9, 2026, proposes to look elsewhere: AI tools that Google, Amazon and Microsoft design tailored for oil companies such as Chevron, Exxon or BP, in order to improve oil extraction.
According to the authors, the climate effect of these tools is between three and thirteen times that of the data centres themselves. In other words, the most significant emission item would not be the electric invoice for the calculation, but the additional hydrocarbons that this calculation helps to get out of the ground.
This scoping has a practical impact on how to assess the climate commitments of major AI players: a company can display a decarbonated electricity supply for its infrastructure while marketing services that increase fossil energy production. The two columns do not appear in the same balance sheet.
Read the full article on the original source
Go to the source →