Stay informed with free updates
Simply sign up to the Technology sector myFT Digest — delivered directly to your inbox.
AI-driven productivity gains in the oil and gas industry will substantially increase global emissions, according to a report led by two former Microsoft employees, outweighing any potential gains from clean energy.
Upstream oil and gas extractors can use AI to reduce drilling costs, find new deposits and improve recovery rates — making them cheaper and more abundant.
It can also help solar and wind operators produce more renewable energy by forecasting how much sunlight and wind will be available, doing predictive maintenance and assessing the best time to dispatch battery storage.
But the peer-reviewed report published in the journal Nature this week finds these emissions savings would be offset by AI expanding fossil-fuel supply.
If AI adoption spread across all energy sectors, it could add between 0.5bn and 1.8bn tonnes of carbon dioxide to the atmosphere every year, according to the authors of the report, who are affiliated with Enabled Emissions Campaign, a climate advocacy group.
While AI use in renewables could avoid up to 500mn tonnes of CO₂ annually, each 1 per cent productivity gain in fossil fuels would require a 4 to 5 per cent productivity increase in the renewables sector in order for net emissions to stay neutral, they concluded.
Some content could not load. Check your internet connection or browser settings.
“AI-driven productivity gains enable more emissions than they avoid — reinforcing fossil incumbency rather than displacing it,” the report said.
Much of the discussion around the environmental impact of AI revolves around the data centres and their heavy use of natural gas and coal to meet computing energy demands.
Utilities and Big Tech are undertaking a huge build-out of gas-fired power plants. In 2025, the US nearly tripled its gas-fired capacity in development, according to Global Energy Monitor. Data centre developers are also using on-site gas and diesel engines, citing long waiting times for plants to be built and connected to the grid.
The Trump administration has extended the lives of six coal plants that were scheduled to be retired.
But the researchers say this misses the impact of AI use by some of the largest fossil-fuel players, which outstrips the direct environmental impact of data centres.
The emissions impact of energy companies using AI is 2.8 to 10 times higher than the International Energy Agency’s estimate last year of data centre emissions, at 0.18 gigatonnes of carbon dioxide.
AI-assisted oil and gas production could unlock 470bn to 1tn extra barrels of oil, according to consultancy Wood Mackenzie, while Goldman Sachs estimates that AI use in shale could expand reserves by 8 to 20 per cent and reduce drilling costs by 30 per cent.
According to an industry study by IBM, 44 per cent of upstream oil and gas companies are already using AI in exploration and another 45 per cent plan to adopt it in the next three years.
Energy production boosts for solar and wind are potentially significant, with the researchers modelling that AI may increase output by as much as 30 per cent.
But there is limited evidence of industry adoption and gains, and AI cannot stop the biggest barriers to renewables deployment: interconnection queues, permitting delays and curtailment due to weak demand or transmission constraints, say the analysts.
Productivity gains in oil and gas also produce larger emissions ripple effects, as their use spreads across the whole economy in sectors like transport, manufacturing and chemicals.
The report recommends that climate and AI policy frameworks take into account “enabled emissions” rather than solely looking at direct energy use by AI, and use carbon pricing along with limits on AI-enabled fossil-fuel expansion.
