By inocta.io
A Harvard-led study estimated that 403 US data centres emitted about 25.7 million tonnes of CO2 in a year, without a separate figure for AI. Early evidence links data centres to coal plants retiring later. Climate TRACE modelled bigger cuts, and COP31 launched the voluntary Antalya AI Pledge.
The video ends with our simulator. To estimate your own team's AI use in trees, use the inocta.io energy simulator.
The sources behind the video
Each card matches one story in the video. Use the time stamp to jump to the claim.
-
Hyperscale data centers' energy consumption, carbon emissions newly estimated by researchers
Why a CEO cares: Roughly 54% of the electricity came from fossil fuels, for about 25.7 million tonnes of CO2 in a year. The data could not separate AI from other cloud work.
-
Early evidence links data centre exposure to coal power retirement timelines
Why a CEO cares: Coal plants exposed to data centres built since 2020 were retiring later. The study shows an association, not proof.
-
New Climate TRACE report quantifies the emissions impact of COP31 climate targets and identifies pathways with the greatest co-benefits
Why a CEO cares: Pairing electrification with new clean power could nearly double the cuts, to nearly 14% of global emissions. These are modelled savings, not results.
-
COP31's Antalya AI Pledge to tackle artificial intelligence's climate footprint
Why a CEO cares: Those who sign commit to cleaner power for AI and to reporting its impact. ITU and the COP31 presidency call on organizations to join, and give no signatory count.
Try the inocta.io energy simulator
Read the transcript
AI and the environment met at COP31 this week.
A Harvard-led study estimated the footprint of 403 US data centres over a year. Just over half their power came from fossil fuels, for about 25.7 million tonnes of CO₂. The study could not separate AI from other cloud services.
Early evidence links data centres to coal. Coal plants exposed to data centres built since 2020 were retiring later. It is an association, not proof.
Climate TRACE modelled electrification goals. Pairing them with new clean power could almost double the cuts, to nearly 14% of global emissions. Its data now answers plain-language questions. These are modelled savings, not results.
The COP31 presidency and the UN's telecom agency launched the voluntary Antalya AI Pledge. Those who sign commit to cleaner power for AI and to reporting its impact. A commitment, not yet a result.
Our simulator brings it to one team: 50 people, 20 AI requests a day each, on the Canadian grid. Their year of AI is about a third of what one tree absorbs. An estimate, not a meter.
The cost is already estimated. Most of the promise is still a model. Sources are on inocta.io, and our simulator shows how many trees your team could plant this year.
Get the next field note by email. Subscribe to The Signal







