Aug 6, 2026 · 34 min · 9 segments
This episode of Think Change explores the growing environmental footprint of artificial intelligence – from rising energy demand and emissions to the vast quantities of water needed to power and cool…
Joaquim LevyGuest
Divij JoshiGuest
Fiona MacklinGuest
Sara PantulianoHost
Our work here is also pointed out to what some call the green digital divide, basically pointing out that the benefits of AI are concentrated while the environmental and the resource costs fall disproportionately on the global south.
So how should AI governance frameworks evolve to address this distributional and climate justice concern?
I think there's definitely been a lack of consideration of the distributional impacts of our current AI economy and the way in which governance conversations have happened and are continuing to happen.
I think a lot of the way in which developments are taking place rest both the control or the kind of ownership of control over these systems and of innovation trajectories, as well as the distribution of value in a handful of countries and corporations.
We're seeing an enormous concentration of power and an enormous concentration of wealth and an unequal distribution that way.
I think in terms of the green digital divide, what we're seeing is, and going back to a roundtable we held with some experts, a really interesting phrase that came up called the gentrification of green energy, right? That while we're focusing on increasing additional capacity to the grid or increasing additional green energy capacity, a lot of this is being bifurcated and going to the development of AI data centers or the running of data centers, right? And this is happening without much of a democratic conversation around whether this is the priority that we need to kind of be fulfilling right now.
And this also means that our long-term kind of infrastructure is also energy infrastructure is also being shaped towards fulfilling the needs of particular kinds of AI usage.
Now, whether that AI usage is itself appropriate for large parts of the world is unclear.
Going back to what Joachim already said, It's not entirely apparent that the kind of AI we need, particularly in the global south or in different parts of the world, is these large frontier models that are incredibly energy intensive, but also incredibly resource extractive in different ways, right? Not just energy, but there's data extractivism that goes into the development of these models, meaning that you're taking data from populations in the global south, but not really giving value back to them.
So who's actually responsible in making these systems and the labor that goes into it? And so maybe one of the things I do want to pose is like, Maybe governance needs to consider what is the appropriate technology for places like that and how do they actually ensure these questions of sovereignty that often come down to making sure both that there's control within jurisdictions of the kinds of innovative trajectories that they want using AI systems, which may mean a large language model, but it could also mean a small language model.
It could also mean a more focused, something like a transformer model used in protein discovery, which has been one of the most high impact things.
How do sovereignty conversations actually ensure that there's democratic and participatory control over what kind of AI we want? As well as from a global governance angle, how do we ensure that considerations along the supply chain of labor, of data extractivism, how do they start factoring into how we distribute value across AI trajectories?
Our work here is also pointed out to what some call the green digital divide, basically pointing out that the benefits of AI are concentrated while the environmental and the resource costs fall disproportionately on the global south.
So how should AI governance frameworks evolve to address this distributional and climate justice concern?
I think there's definitely been a lack of consideration of the distributional impacts of our current AI economy and the way in which governance conversations have happened and are continuing to happen.
I think a lot of the way in which developments are taking place rest both the control or the kind of ownership of control over these systems and of innovation trajectories, as well as the distribution of value in a handful of countries and corporations.
We're seeing an enormous concentration of power and an enormous concentration of wealth and an unequal distribution that way.
I think in terms of the green digital divide, what we're seeing is, and going back to a roundtable we held with some experts, a really interesting phrase that came up called the gentrification of green energy, right? That while we're focusing on increasing additional capacity to the grid or increasing additional green energy capacity, a lot of this is being bifurcated and going to the development of AI data centers or the running of data centers, right? And this is happening without much of a democratic conversation around whether this is the priority that we need to kind of be fulfilling right now.
And this also means that our long-term kind of infrastructure is also energy infrastructure is also being shaped towards fulfilling the needs of particular kinds of AI usage.
Now, whether that AI usage is itself appropriate for large parts of the world is unclear.
Going back to what Joachim already said, It's not entirely apparent that the kind of AI we need, particularly in the global south or in different parts of the world, is these large frontier models that are incredibly energy intensive, but also incredibly resource extractive in different ways, right? Not just energy, but there's data extractivism that goes into the development of these models, meaning that you're taking data from populations in the global south, but not really giving value back to them.
So who's actually responsible in making these systems and the labor that goes into it? And so maybe one of the things I do want to pose is like, Maybe governance needs to consider what is the appropriate technology for places like that and how do they actually ensure these questions of sovereignty that often come down to making sure both that there's control within jurisdictions of the kinds of innovative trajectories that they want using AI systems, which may mean a large language model, but it could also mean a small language model.
It could also mean a more focused, something like a transformer model used in protein discovery, which has been one of the most high impact things.
How do sovereignty conversations actually ensure that there's democratic and participatory control over what kind of AI we want? As well as from a global governance angle, how do we ensure that considerations along the supply chain of labor, of data extractivism, how do they start factoring into how we distribute value across AI trajectories?
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.
