AI, Energy and Climate: India AI Impact Summit: Arunabha Ghosh

That's the music for this video. There's no better place than right here on the DSR network.

“And there's no better way to enjoy the DSR network than by becoming a member.”

Members enjoy an ad-free listening experience, access to our discord community, exclusive content, early episode access, and more. Use code DSR26 for 25% off discount on sign up at the DSR network.com. That's code DSR26 at the DSR network.com/buy. Thank you and enjoy the show. Welcome to the AI Energy and Climate Podcast.

A special series from the DSR network hosted by David Sandalot and Nagarro Fellow at Columbia University Center on Global Energy Policy.

Join us as we talk with leading experts to explore the intersection between these critical issues that will impact the future of each and every one of us.

“Hi, I'm David Sandalot. This is the AI Energy and Climate Podcast.”

I recently attended the AI Impact Summit in Delhi. I thought it was quite an amazing event with tens of thousands of people, hundreds and hundreds of subs and obsessions, and speeches by more than 20 heads of state as well as dozens of tech company leaders, including Sam Altman of Open AI, Dairy Oameda, and Tropic, Dennis Hossabas of Google DeepMind, and many others. This was the fourth in a series of global summits on AI. The first was in the United Kingdom in 2023 and Bletchley Park outside London.

The second was in Seoul, South Korea in 2024.

The third was last year in Paris, and the fourth was this year in Delhi. This was the first of these AI summits in a developing country, and a lot of the discussion at the summit, centered around the role of AI in developing countries. And some of it was part of Prime Minister Modi's work to put India on the map when it comes to AI, and a lot of people spoke about how the average age of participants in the summit was quite young, probably in the mid-20s. While I was in Delhi, I had the chance to visit one of the best organizations working on energy and environment in India and my experience.

The Council on Energy and Environment and Water, which is widely known by its initial CEW. It's run by my friend Aranabagosh, who is consistently thoughtful and insightful on energy and climate issues. I learned in the course of our conversations that he and his team are doing very thoughtful work on the role of AI in the energy and climate system. And so I was delighted when Aranabagosh said he'd be available for a conversation on our podcast. Both Aranab and I had flights to New York on the day of the big blizzard.

I was returning home, he was coming to visit. Both are flights for canceled, along with thousands of other flights. And it wasn't easy for either of us to get to New York City. But by the end of the last week, we both managed to get to New York and had a chance to record this conversation. I hope you enjoy it. Aranabagosh, welcome to the AI Energy and Climate Podcast.

Delighted to be on this podcast, David. Thanks for having me. Well, it's great to see you and it was great to see you in Delhi last week, where we both attended the AI Impact Summit. I thought it was quite a remarkable event. There were more than 35,000 people there, hundreds of substantive sessions of vast exposition hall. I guess more than 20 heads of state spoke, including McCron and Lula, Sunches of Spain, and of course Prime Minister Modi. And then all technology leaders, including Sam Altman, Dario Amidade, Dennis Hassebus, underperchai of Google.

“What are your reflections? What are your thoughts about the summit?”

Well, my first thought is what the summit was titled.

It was called the AI Impact Summit, and it's the fourth and a series that was the AI Security Summit in the UK. There was the AI Soul Summit, which also continued with Security and Safety in France. There was the AI Action Summit, but I think the first summit on AI occurring in a developing country with a focus on impact.

And what's the best way in which it could be deployed, which is inclusive and impactful and in a way, you know, democratic.

“So I think just nudging the conversation in that regard was to me the biggest takeaway.”

Of course, a lot more, a lot more nuances in that statement, but I'm glad that it occurred. And I'm glad 20 heads of state, 50 odd ministers. And I think by some count, well over 100,000 people showed up. Well, and you had an important advisory role for the government generally in this area.

“And I think for the summit as well, could you say a word about that?”

Yes, I was appointed as the chair of an AI and climate expert engagement group. So the way the summit was structured was that there would be working groups of the member states of or rather of country delegations that would debate and discuss various aspects of AI. But these expert engagement groups were designed to bring in the external expertise into the conversation. And AI and climate was a very important one, especially when we think about not just AI done sustainably, but how AI could support sustainability.

These were the two big themes that we tackled. And I was privileged to bring together a group involving representatives from a private sector, from Microsoft, from Google, from MIT and Georgia Tech, from IIT and India, civil society organizations. So we had a very rich and deep conversations on many aspects. And we put out a report before the member states consideration at the summit.

“But I think the best part of having a group like that is we take the conversation and the actions forward from the summit.”

So it's more about what is to come that is more exciting than just what we did. Well, so what did you recommend? Well, the two themes, as I was saying, are one, how do we make AI more sustainable?

And second, how do we use AI to drive sustainability? On the first, the issue here is not just about, you know, a race to bigger and better compute path, but to do it with intentionality.

So transparency in the environmental footprint, citing decisions being taken in with the environmental resource constraints in mind. So, for instance, locations where there's more access to clean energy, good AI, I could data centers be situated there. Similarly, locations where water might be more readily available could those, then, you know, more water intensive cooling technologies be deployed. So having some intentionality to this. And then embedding metrics such as power use efficiency, water use efficiency carbon intensity, all these are ways in which we can have a more intentional approach towards making the AI story a lot more sustainable.

But the second pillar, and this is why this was an AI impact summit, was how do you use AI for driving sustainability actions better, whether it's in disaster resilience, whether it is in optimizing for clean electrons, whether it is in driving sustainable agriculture and so forth.

But underlying both these pillars, we basically argue that we need four principles to guide the development of AI and its intersection with climate.

The first principle has to be the planet, the second has to be the purpose, what is this for, the third, the process, the transparency, the inclusivity with which this is done, and forth, keeping people at the center. So creating the infrastructure and its use cases with communities, so that it has the greatest kind of social benefit, but also is is inclusive by design. So these four fees were our core principles, planet purpose, process, and people.

Our number, this tracks very closely with some work we've done in the past couple of years is so interesting to hear what you're saying because we've done a sustainable data centers roadmap, which identifies many of the issues you just talked about one of our recommendations was that sighting is absolutely key for the sustainability of. And we need to pay attention to the power source and to water availability and then we've actually done two additions of an artificial intelligence for climate change mitigation roadmap that pulls out many of the things, although at your four fees are just a great way of summarizing this.

But about development of data centers sustainably in India in particular, one of the things I heard over there last week was emphasis on using renewable energy for data center development in India.

And I heard people express concern about water limitations and water scarcity in parts of India.

“What do you think is the potential for developing data centers sustainably given the principles that you're in your committee identified?”

See, at one end there is going to be a huge growth in the data center infrastructure. The latest budget in India has made it very clear that data centers above five gigawatts will be considered as infrastructure. And secondly, that there's going to be about 20 year or more than 20 year long tax holiday to attract investments. We think that the power demand will itself grow to about six and a half to eight and a half gigawatts by 2030 upwards from where it is now about one and a half gigawatts of power demand for data centers.

About 150 billion litres of water being used to cool the good data centers we think that'll at least double.

“So the question that is how do we make this or do this sighting in a more rational way?”

We've identified and we're similar to your roadmap, which I had a look at David. We identified and report between CW and systemic, the way in which the capacity was coming up.

Basically, Delhi, Mumbai, Bangalore, Chennai, Kolkata, these five cities regions account for about 60% of the data center development are great.

We don't have a national level in sentence, we don't have a national level policy framework. But what you do find in India is there about 15 states, 15 of our 28 states that have some kind of AI/data center policy framework to attract investments as part of their industrial policies. However, only five right now have explicit sustainability metrics embedded in.

“So I think the opportunity here is to see that if this is going to be a major investment and infrastructure play.”

How do we optimize not just at an individual state level, but how do we optimize at a national level? So for instance, if there is a particular part of the country that is well suited for clean energy generation. Let's generate more clean energy and create the direct of take for data centers, but you could potentially cite them in areas where water is more easily available. Similarly, when you're thinking about water, let's not only think about water withdrawals, but actually think about water consumption, two different metrics.

It's the water consumption that matters. So how do we create a closed loop system, whether water consumption reduces? And on the withdrawal side, let's promote a circular economy of treated wastewater, which can then serve as the cooling source. So this this optimality cannot occur, just state by state or city by city. So once we start looking at it at a much larger regional scale, I think the opportunities become much greater because the sunshine somewhere, the water and the hydropower flows somewhere else and the wind blows somewhere else.

Start optimizing for this, you get a much more sustainable pathway for very large scale data center development. You can find your own landscape with Shopify and business. And to be honest with you, with the checkout with the world, because you can improve better. That's right, the checkout with the world that can improve better. The legendary checkout from Shopify is simply the shop of your website, a bit to social media and over everything else.

Its editorial content is completely independent, and the views expressed are exclusively those of participating experts.

It is presented live without editing.

“For further information about the UAE's efforts in the areas of artificial intelligence and technology, go to the website of the embassy at www.uay-emBC.org”

And search for UAE-US tech cooperation. We thank them for their support. We thank everybody who is supporting this podcast for their support, and we look forward to it developing and growing over time because the issue is so important. It's so interesting here in which my perspective here in the United States, because there are some significant similarities between what's happening in the United States and India.

But we did a study last year looking at what's happening in state governments in the U.S. on policies.

And many states have incentives to attract data centers into the state. There are numbers of new laws passed in 2025 on that topic. Sounds like the same thing is true in India, but much fewer states in the United States have any type of sustainability guidance with respect to those data centers.

“I think in the United States, there may be a trend towards more of that.”

There's starting to be local opposition, particularly based upon water usage. It's strengthening the United States, but there are some real similarities there. And in your point about the circularity and water use is really important. For those listeners around familiar with this people often use water withdrawals to me and just how you how much water you take out of the ecosystem, but consumption. But then if you return it to the ecosystem, it's not considered consumption. So consumption is how much you take it out of the ecosystem.

And the good news is that we have now more and more systems for cooling data centers that are closed loop is you're suggesting that the don't consume a lot of water. So so that's very that's very promising. Thanks for clarifying that David and then if I could just add one more point to you, which is that we are seeing the trend that you're describing whether it's in the United States, whether it's where even another developing countries. I sort of tongue-in-cheek, call it sort of the formal approach to data-centered development.

So everyone's got the fear of missing out, so you know, let's throw in everything to attract the investment in building that infrastructure.

“But once we start thinking about the resource footprint, and we start thinking about the end use cases, then I think not just the infrastructure development can be more optimal.”

But even the provinces and cities or countries can decide what kind of incentives they offer to attract what kind of infrastructure development, which makes the most sense for them at an economy level, but also at a social and political level, get that social license to operate accordingly.

But turning to the second bread topic that you say your committee was looking at, which is how do we use AI to promote sustainability? What do you see as the major opportunities in India in that way?

So one of the biggest opportunities we see is increasing resilience. 80% of Indians today are living in areas that are already highly vulnerable to hydromanological disasters of subkind of the other. How do we build resilience for communities, citizens, communities, and the economy as a whole? While not slowing down the economy, this is one of the world's fastest growing major economies. So deploying AI to build resilience is absolutely a lower-hanging fruit.

I'll give you some particular use cases, early warning systems and disaster response. An AI powered flood forecasting system can provide early warning seven days in advance. And thereby significantly reduce the lot potential loss of lives or a main infrastructure before the disaster hits. Similarly, AI-based models for extreme weather forecasting can increase accuracy by up to 15% outperforming the traditional forecasting methods. That means you're able to get the information out to a small hold of armor when or how intense the rainfall will be.

Similarly, if you have digital twins of physical infrastructure, and then use AI to map, say, the heat risk.

“We at CW have been developing the first high-resolution climate resilience Atlas for India.”

Our heat action plans, now in over 100 cities, are at a neighborhood scale. And we are using machine learning tools and AI to plug in, at least 70 to 80% of the core inputs that are needed to develop those plants. All this helps to very rapidly increase the nature of risk assessments that then opens up the opportunity for more precise salient insurance cover.

That's one kind of use case. The second use case is to optimize on the electrons.

As I was saying earlier, when the sun is shining or when the wind is blowing, that's when you use the electrons better.

“But predictive accuracy using AI helps us to manage not just the grid, but can help to manage more spot trading more intraday markets for clean electrons.”

Which therefore means that your data centers, but other infrastructures, etc, can also use the cleaner electrons more readily, without them getting wasted. And your grid is also more resilient.

And a third use case that we've identified in a report of ours is on sustainable agriculture.

In a country like India, 60% of the farmers are small and marginal, actually even more. And they're all dependent on grain fed agriculture. If we need to move them towards more sustainable agriculture practices, there are different government schemes, etc.

“That can be rationalized to get money out of the door into the hands of the farmers.”

Another challenge is that we find is that the verification becomes problematic. So we've seen that using sort of an AI intelligence layer for an Agree stack can help combine the farmer, the farm plot, the crop variety, the crop practices into a much more credible farming or sustainable farming registry. Where you get verification at the plot level based on agricultural land classification. And therefore the payout from the government to the farmers can be quicker and encourage adoption of more of these practices.

So whether it's an early warning and resilience against disasters, whether it's an optimizing for clean electrons, whether it's in supporting the poorer farmers of which there are tens of millions. We see very clear use cases in the here and now, which is why just the compute bar game is not the game that we should be playing or even ought to be playing. In fact, there's a very different story to be told from country like India. It's such a thoughtful answer on of and it really underscores the many ways that AI can contribute to number different areas including is you're highlighting climate resilience mitigation of climate emissions.

I wonder have you thought about what it will take to make those use cases real. And what barriers need to be overcome to actually transfer those ideas into things that are happening in the real world.

The first thing in any AI model deployment is actually getting better data.

You know, it's what you're feeding in which is what the model is learning and self learning. So the prior art here has to be in using all resources available, whether it is ground proof thing, whether it is satellite based imagery. Where it is geospatial, whether sensors to get better data, whether it's on on heat stress precipitation patterns extreme weather impacts etc. Or even data on electrons produced or not transparent platforms for that really can help democratize. The access to this information but also feeder into models that can then use that information better.

The second step David is we also have to let's not underestimate the investment that is needed in the capacity to interpret the information.

And you know, whether you are a small holder farmer or whether you are an administrator in a small district. Just saying that there is a model that's predicting, you know, the solar radiation that's predicting the wind flow or it's predicting the next disaster is insufficient.

“If you're unable to interpret that result and use it for what actions you need to take.”

Does that mean you need to, you know, send in some first responders to deal with an impending disaster, natural disaster or does it mean that you have to delay, for instance, the date on which you saw your seeds in the farm. So that you know, intense rainfall does not, you know, kill it, kill the plant and plantation before it's germinated. So the interpretive capacity is highly underestimated and I think this is worth an investment for countries to do.

The third thing I would say is be a little bit more open and creative on the public private partnerships, right now everyone claims their.

The model is the best or their tool is the best, etc. And I think we need what I would call fidgetal sandboxes a combination of physical and the digital sandbox which is open and transparent. So different model developers and different use case developers can try things out transparently.

“Because you see the market, none of us really knows the kind of scale that the use cases will take.”

So rather than think of it as a fixed pie that one model developer versus another has to fight over. We are still at early stages of this this story. And if we did it more transparently in these physical sandboxes, we can a lot more confidence on the known nodes, but also the unknown unknowns and and drive the next generation of innovation that is needed in the models. But also needed in the business cases and the use cases that I can be useful. Save, how does this story go?

Now it's time for the first time I've heard the term "visitual sandbox" that's very interesting.

“Maybe in populace I think I've heard it said save of what is the physical component of the sandbox. What do you mean by that?”

The physical component is the infrastructure, the policy apparatus from the ground. The point here is the digital story can't be a solution looking for a problem. It has to follow the messiness of the real world and be designed accordingly. Even language. There are startups and India. There are developing models. We've got 1500 dialects in the country. Hindi alone. Every 80 kilometers the dialect changes. English is not an official language, but it's a language people speak and a combination. So the point I'm trying to make is that real world is complex and it's nuanced. The policy world, the administrative world is complex, nuanced, undercapacity to sometimes and yet has a lot of resilience that is not accounted for.

So when we combine the digital tools of which AI is only one end with the physical infrastructure, the community, the administration, then you gather actual real case of what is called a digital sandbox.

Take your big. It's a small question. It's a follow up by her discussion less. We can India about all the languages in the country, which of course are multiple as you say in many of them are spoken by large numbers of people.

Have you read or heard anything about how effective the AI is in some of the minor languages and how good the tools are with some of the languages.

“I'm forgetting the name of the startups so pardon me for that. But I think there's a few Indian startups that have gotten a lot more accuracy in their models by making the models learn on on these kind of nuances in our in our linguistic heritage.”

So you know, I don't know how they have trained the models or what kind of compute power has been needed for it and whether they can with the models can do everything on that son. But they've prioritized and privileged the importance of linguistic diversity.

Maybe the expense of comprehensiveness on every question that you can ask. So again, I think there will be an interplay in different types of nodes.

“So if you then ask a question around, you know, what's the best way in which, you know, I could grow rice.”

Today, if you type that question in, you might get one type of answer based on known systems of agriculture. But if you've not factored in indigenous knowledge and and traditional practices and the linguistic traditions that drive that not just in India, but elsewhere as well. Then I think you'll miss out on on on some of the key insights that would be necessary. So it's just an example of again why just building the next largest model might not be the most appropriate way to go down. And one race to run, but mirroring the complexity of human civilization and its interaction with an even more complex planetary system might be a more challenging and yet more exciting race to be part of.

“Well, our job is time to start to wrap up. I want to be sure that our listeners know about the terrific organization that you run, which has the initial CEW, it's the, I think I've got to write the council on energy environment and water.”

And you were good enough to invite me out there for lunch last week. I appreciate that that's not the reason I think it's terrific. When relying upon your reports and reading your work for for years now, it's with many others, but we would just say a word about CEW and what it does and how our listeners can confine the resources that you're putting out. Sure. Well, a thanks David for for your kind words CEW or the council on energy environment and water is an independent policy research institution where 15 and a half years old is a rowdy teenager.

Which we founded in 2010. We've been in the past ranked among the 20 best climate think tanks in the world, but I would say our work is no longer just about climate. We really earn a think tank that bridges development and sustainability. So our work currently has a lot of focus on integrating more clean energy into the grid, on industrial decarbonization with a focus on materials circularity. We have a lot of focus on energy security on the fuels of the future, the critical minerals, the components, etc., going to clean tech. And a huge amount of work around the quality of life, you know, the air we breathe, the water we drink, how we become more resilient against shocks, but the biggest theme currently in our strategic plan is around building a green economy.

It's extremely exciting because a green economy is not just the energy transition, it's a nested egg by the energy transition nest within the circular economy, it's never nest within the bioeconomy.

And we estimate that in India alone that is a 4 trillion dollar investment opportunity over the next two decades. For us at CW, we like to look over the horizon, on policy issues that are not yet policy issues. And we've done this from the days we began. We've been intimately involved in the growth of India's renewable story, it's green hydrogen play, it's critical minerals play, it's announcements around net zero, working now with 20 state governments, hundreds of cities.

You can get more information at CW dot iron.

“Well, thank you for your time, Arana, but we always wrap up this show with two questions for our guests and the first one is how are you using AI these days?”

Well, I am still training myself to be honest. We've built out a AI team at CW across all our 15 research areas, etc. But I personally, I'm training myself to get better at the kinds of prompts I give.

And also I think with the development on the quant site, you know, they're not just the, what chat GPD does for instance, but what clouds, opportunities are, etc.

I would like to see a lot more efficiency there, including on administrative tasks than running an organisation as large as us. I would like to make those things simple. Yeah.

“At terrific. And final question, could you recommend three books reports or articles for our listeners? Could be anything, a favorite from your childhood or something you're ready yesterday?”

Well, something I started reading yesterday actually on my flight over to the United States was, is a recent book by Jimmy Wales founder of Wikipedia. It's called Seven Rules of Trust. And I find quite fascinating because, you know, this is a co-created platform, right, in the internet age. And it makes me wonder, what is the co-created platform we'll create in the AI age? And one of his principles is transparency.

“And I think we all have to be radically a lot more transparent about our, as I said earlier, our environmental footprints, our resource footprints, our intentionality, etc.”

So that's one because it's top of my mind and it's still on my bedside table. And article, I would urge people to read maybe as a set plug is I wrote a paper in foreign affairs a couple of years ago called can India be a green superpower. And I would urge more not your audience to read this because not because it's kind of a flag waving pitch. But there is a lot that's going on in a large emerging economy like India that is not, not know and just for that knowledge worthwhile.

And a third book I would think about is, it's in my top 10 sort of books ever.

And it's a complete, it's a completely different context, it's a decolonization era in Nigeria. But why it comes to my mind right now when you ask this question is because a lot of what we do as intelligent people, as I mean form people, as presumably rational people, is work on the back assumptions of what society, around us is structured as. And if I think about that book, things fall apart, it's not just that decolonization was happening. But how social change was occurring for the good for the bad and how received wisdom about the structure of society, the structure of politics began to disrupt, began to change from the lens of an individual.

I think we are entering and having a way entered that kind of a world within our countries and globally. So we should ask ourselves, not what is rationality within our received assumptions, but what are the new assumptions itself that we have to first explore and encounter before we determine what rational and perhaps even human action should be. Yeah, and well, if I could make one more pitch, my next book is on water and that'll be out on world water day next month, it's called water nature, progress.

So welcome you to read it, but as I said, you know, things are falling apart, but it's also an opportunity to re-build together. Oh, are an up-a-gosh. Thank you for the thoughtful recommendations I'm looking forward to reading your new book, that's very exciting. And I always learn every time we talk with you very grateful that you made the time to join us today.

Thank you David, it was a pleasure chatting with you and look forward to seeing you again in person very soon.