What If We’re Measuring AI’s Climate Impact Wrong?
About the session
The debate about AI and climate has settled into a familiar shape.
On one side: data centers consume enormous amounts of electricity and water, and training and running AI models carries a growing environmental footprint.
On the other: AI could help accelerate renewable energy deployment, optimize electricity grids, improve efficiency, and enable parts of the clean energy transition to happen faster.
Both matter.
But what if this debate is missing another emissions dynamic entirely?
When technology makes an energy system more productive, the resulting efficiency does not necessarily translate directly into lower total energy consumption. Greater productivity can lower costs, expand access, enable new economic activity, and ultimately increase how much energy gets used.
This is the rebound effect.
And it raises a much harder question for AI and climate: not simply whether AI consumes more energy than it saves, but whether the economic activity enabled by AI-driven productivity gains creates additional emissions that need to be included in the equation.
New peer-reviewed research published in npj Climate Action examines this problem by assessing the net climate implications of AI-driven productivity gains in energy systems.
Its implications could complicate some of the assumptions underneath today's AI and climate strategies.
Organizations are making responsible AI commitments. Technology companies are procuring clean energy for data centers. Investors are backing AI applications intended to accelerate decarbonization. Policymakers are beginning to design governance frameworks around AI's environmental impacts.
But much of that work focuses on AI's direct operational footprint and the emissions reductions its applications might enable.
What happens if the wider economic effects also matter?
This session will bring the research into the room and use it as the starting point for a broader conversation about how AI's climate impact should actually be measured.
Should responsible AI commitments account for enabled emissions as well as operational emissions? How should investors evaluate an AI application that increases efficiency but also stimulates additional demand? What does clean energy procurement accomplish if overall energy activity continues to expand? And which of these effects can we measure confidently today versus those where the evidence is still developing?
The goal is not to settle whether AI is “good” or “bad” for the climate.
It is to make sure we are asking the right question.
Discussion Group Leaders
Eric Firnhaber is Senior Director, Global Engagement at Digital Green, advancing global partnerships and engagement that strengthen technology-enabled solutions for smallholder farmers and rural communities.
Will Alpine is Sustainable Software Design Leader at Schneider Electric, advancing more sustainable AI and software through carbon-aware technology, industry standards, and responsible innovation.
What to expect
A research-into-practice conversation beginning with the analytical gap and moving quickly toward what it means for decisions being made now.
Participants will engage with the core findings of peer-reviewed research on AI-driven productivity gains and energy systems before exploring their implications for responsible AI commitments, clean energy infrastructure, investment, governance, and climate advocacy.
The conversation will bring different disciplines into the same room: researchers studying the effects, engineers and technology companies building the infrastructure, investors allocating capital, policymakers developing governance frameworks, and advocates translating technical questions into public policy.
Importantly, the session will distinguish between what the evidence can already tell us and what remains uncertain. The purpose is not to replace one simplistic AI-and-climate narrative with another, but to build a more complete analytical frame.
Who this is for
This session is for AI and climate researchers, technology companies making net zero and responsible AI commitments, clean energy investors and developers, climate advocates, policymakers working across AI governance and climate policy, and anyone who has participated in the AI and climate debate and wants to examine whether we are measuring the right things.
What you will get out of it
A clear explanation of the rebound effect and why AI-driven productivity gains complicate conventional assessments of AI's climate impact
An introduction to peer-reviewed research examining the net climate implications of AI-driven productivity gains in energy systems
A framework for thinking about enabled emissions alongside AI's operational footprint and potential emissions reductions
Practical questions for responsible AI commitments, clean energy procurement, AI investment, governance, and climate advocacy
A clearer distinction between what we can confidently measure today, what remains uncertain, and what organizations should start paying attention to now