Artificial Intelligence and Environmental Sustainability: Addressing the Land Footprint of Al and Data - Action Dome, UNCCD COP17

Good afternoon, esteemed panelists, colleagues, and dear participants. Welcome to this UNCCD COP17 side event, Artificial Intelligence and Environmental Sustainability, which addressing the land footprint of AI and data centers. So my name is Amrita Satishan. I'm faculty of Environmental Law at Jamia Hamdard University, India. So let us introduce our panel for the session today in this land footprint of AI and data centers. Dr. Murli Thummarakudi, Director, G20 Global Land Initiative. And next, Dr. May Martin, Head Research and Training at the United Nations University Institute for Water, Environment and Health. Our next panelist is Sergio Rebekho, Director General, Nature and People Foundation, and Dr. Bilal Djemaoui, Deputy Director, ITU Telecommunication Standardization Bureau, and finally, it's Dr. Gayathri Nair, the Program Officer, Technology and Infrastructure for Grid Integration in IRENA. So welcome you all. Welcome to the session. Moving with the discussion, there is no doubt that the artificial intelligence offers significant opportunities to advance the sustainable development and strengthening the environmental decision making nowadays especially. But however, AI is often presented as an invisible technology to reality. The infrastructure supporting AI is really actually it's very physical. So every AI system depends on their data centers are there. They are using some advanced computer chips. So we have some critical minerals excavations and also and so it's happening, which is more associated with this AI data center. That means also has a significant water and land requirements there. So that is why the topic is particularly relevant to UNCCD COP17. Without the careful planning, and it may intensify the competition for land and place additional pressure on our ecosystem, so this is why today's discussion is both necessary and obviously it's very timely one our purpose. It's not a question the value of AI or to present technology and environmental sustainability as an opposing goal, never. Rather, our purpose is to examine how the benefits of AI can be achieved while its infrastructure is planned and governed responsibly. So the major focus of today's session is to understand The growing land footprint of artificial intelligence and its wider implication for sustainable land governance. So we will begin by looking at the emerging evidences, increasing competition for land and also as data centers and other AI infrastructure expand. Definitely, we need to consider their implications on essential land uses. And equally important concern is measurement and transparency as well. So how can we better measure? How can we better disclose and monitor AI's land footprint? And how can international initiatives contribute to the greater accountability for that? So from there, we will explore the governance dimension. and how AI infrastructure can be incorporated into the sustainable land use planning, and finally, the alternate solution to reduce the land stress through the AI data centers. So that will be our focus of our discussion. So with these introductory remarks, let us begin our discussion. It's my pleasure to begin this discussion with Dr. Murli Tumarikodi, Director of G20 Global Land Initiative. Dr. Murli, I believe, he will discuss the implication of AI infrastructure for sustainable land management and connect this emerging issue with the UNCCD priorities. Yes, Dr. Murli.

Yes, thank you, Chair. I must say I'm not sitting here as an expert, but clearly as an AI enthusiast. I have been working very closely with Dr. Bilal since 2017 when the ITU started AI for good and he will talk more about it. And then of course, ChatGPT came and AI became household terminology for everyone. And again, the first two years we were very excited about all the possibilities, all the prompt engineering and all that and suddenly, early this year, we find that there is one report which came out from the UN University on the environmental footprint of land. I have not seen that before actually, but then I found out that actually ITU had done a report before. The Secretary General in his London Climate Week address mentioned that the environmental impact of AI has to be studied and launched the AI environmental transparency initiative. So I thought when we have the COP17, we cannot not have AI in our agenda. As I was preparing for this, I found that there is another report which came from a joint think tanks from India, Russia and China, also on AI and land. And I'm very glad that you're here to discuss about that topic. And then we have International Renewable Energy Agency who also deal with Yeah, in a peripheral manner because the Secretary General announced that one of the commitment which the AI industry should probably do is to say that all the energy for AI should come from renewable by 2030. So clearly there's a role for ITU. I was very delighted and honored to have all these experts agreeing to join this session at a very short notice because we only started this after June. So we have tremendous expertise in this sitting here, and I'm here more to listen as much as you are, and I look forward to a tremendous discussion. Thank you.

Thank you, Murli. I think that was the framing intervention from the part of Dr. Murli. So now I move with Dr. Martin, actually, who will now lead the discussion with the evidence. I think he can discuss the topic from the evidences from his research studies as well. Yes, Dr. Martin.

Thank you, Chair. And thank you for everyone for joining this panel. This year, last month, we have published a report what Murali has mentioned. to show how AI and growth is actually impacting the footprint in different aspect. So we did some analysis not on the whole lifecycle of AI, but mainly energy consumption of AI and how that is impacting carbon, water, and land across the world. So what we wanted to show that when we use AI, We think that is something in software, but we want to show that AI is not a software, it's an infrastructure because to serve that chat and software, a huge amount of infrastructure actually behind it, which also has a big lifecycle and workflow. So we want to show that it is, we cannot treat it as a software, it's an infrastructure problem. And this infrastructure is data center, and which is very linked to the data center. And we have shown that data center is growing very fast. And the energy consumption of data center, if we put that whole data center as a country, see it would be the 11th largest consumer of electricity or energy. So, which means it's a huge amount of energy is consumed by data center. And then AI also in the data center, AI share is also growing. Currently, one fifth of the data center today would be by 2030, it would be two fifth of the data center energy consumption will be contributed or dedicated to AI. So this is actually huge phenomena and then we have to look into it. Then we can see that The current, when we talk about environmental footprint, this footprint is mainly used as mainly measuring carbon footprint, but we want to show that carbon only measure is not sufficient to show because when you generate energy, it also has a land footprint, also it has a water footprint. So we want to show that, so there is a pathway of that for energy footprint, For example, when we generate energy, so -- and we focus only on carbon, then we focus on green energy, but then what kind of energy mix we use? If we use hydro or solar, so they have lot of land and water footprint. So which means that this energy when consumed by green energy, consumed by data center, that can be utilized for other sector to grow for green energy transition. So we need to show not only for carbon aspect or perspective of carbon, we have to look into both water and land aspect as well. So which means all these three footprint are actually at scale and we have to show that trade-off between that. So there is a, and then there are two aspect of AI. So one is we also say that training, so when the big model and there is a growth of larger and larger model to grow. So there are a lot of training and there are a lot of news and buzzword about big, big parameter and big, big training. But we also wanted to show that training is only one part of the equation. the 90% of energy and consumption and service is going to actually inference when the training and model is developed, billions of people are interacting with AI model and there the consumption is huge. And then how we use that, for example, if we want to use a regular image like a 10 watt bulb running for 17 minutes, but then two tablespoon of water. But then if we have a genetic complex video, then same bulb would run for 42 hours and two days drinking water equivalent of a person. So which means we use it and how we use it is also important. And this is also a -- and that is going out very at scale and billions of people are using that. So we wanted to show in that report that We want to focus on both training and both inference as well. So in this process, we want to show that now how to address it. We want to make sure that we are promoting that first of all to address it, we need to understand what is going on. If we cannot measure it, we cannot manage it. So which means the data center and AI companies should first transparently disclose what is their footprint is about. And then government and international corporation and international organization, including UN, like UN Secretary General, initiated a global AI transparency initiative, actually that is a follow up of the report that we have published. So we also call for that transparency and then Based on that transparency, then government and international agencies and negotiators can find out a better solution how to take care of this. We are not against AI, but we are making it how to make it sustainable, how to make it more transition, more useful and more sustainable for the world. Thank you.

Thank you, Martin. Now I think I would like to invite our next speaker, that is Mr. Sergio Rebeko, Director General of the Nature and People Foundation. Actually, I'm expecting you will represent your findings from the white paper on the AI and environment especially. Thank you.

Thank you, Chair. It's a great pleasure and honor to be here today, and I want to thank the Secretariat of UNCCD. and the government of Mongolia for organizing this fantastic conference of the parties. This year is a super year for environment. UN will be hosting three special conferences of the parties and the water conference. And it is great that it all starts here in Mongolia at the 17th session of UN Convention to Combat Desertification with the main topic, restoring land, restoring hope. But for us, this year started in India in February at the India AI Impact Summit. The summit became the largest AI event in the global south, bringing together over 500,000 participants and representatives from over 100 countries. It was focused on healthcare, food security, agriculture, education, fintech, and especially thematic financing for AI development, and how and where it is currently being addressed in the current environment. But this summit also showed that environment was not a part of the technological agenda. As we say in Russia, If you're not in the menu, if you're not at the table, then probably you're in the menu. So India AI Impact Summit pushed us to prepare a white paper on AI environment. As Professor Sachs from Columbia University wrote in his foreword to this document, there are several advantages of this white paper. Its first great virtue is its timeliness. The infrastructure decisions that will determine as environment footprint for decades have to be done now. Our key message is that AI's environment impact is not a fixed cost, but a governance choice. Its second virtue is its comprehensiveness. The paper holds together in a single frame several topics that are mostly discussed separately. AI is both a growing source of environmental pressure and simultaneously among the most powerful tools we possess for providing sustainable solutions for environmental harms. The white paper follows AI across its full lifecycle, energy, carbon, water, critical materials, and electronic waste. It consistently examines domain by domain on AI's double-edged role in climate change, biological diversity, land degradation, and water resources, both as a potential threat and as a potential source of new solutions. The white paper's third great virtue is the distinctive partnership of authorship. The white paper is produced jointly by the Dynamic Coalition on Environment of the Internet Governance Forum, the Palai India Foundation, China Institute of the Fudan University, and the Nation People Foundation, thereby integrating the perspectives and policy experience of China, India, Russia, alongside the best known cases of United States and European Union. Quoting Professor Sachs, by treating AI governance as a genuinely multipolar endeavor it is, this paper offers analysis that simply doesn't exist elsewhere. Based on our research, we have identified three key challenges that AI is facing right now. The first problem is the concentration of technological power. Advanced chips, clouds, basic models, data, and capital are controlled by a limited number of countries and companies. This creates a risk where formal access to AI is combined with dependence on external supplies and a lack of influence over standards, prices, and data usage rules. The second problem is the gap between declaration and implementation. Investment promises, political declarations, financial commitments, and exhibition demonstrations vary in nature, but none of them by themselves demonstrates social impact. The third problem is the fragmentation of governance. The United Nations, the International Telecommunication Union, PAX Silica, the World Organization for Artificial Intelligence Cooperation, regional organizations, national regulators, and technical bodies can complement each other. But without a clear division of functions, they can duplicate work and make incompatible demands. Based on our white paper, we have also framed five principle stages performed by AI of a land governance process. The first one is identifying land degradation. This enables government to identify degradation most rapidly and systematically than through conventional field sampling alone. Second, determining where and how intervention should take place. AI supported systems combining data can help determine for what a particular land area is best suited for. Third, improving operational and construction efficiency. Drones, robotic systems, autonomous or semi-autonomous vehicles are the new cost efficiency solutions in the construction industry. Fourth, monitoring public expenditure and project quality. AI assisted monitoring contributes not only to ecological effectiveness, but also to public sector accountability and the transparency of environment investments. Fifth, connecting ecological restoration with sustainable business models. The most important contribution of AI is the creation of a more integrated decision making system that connects ecological science, public finance, infrastructure investment and long term economic activity. This July we presented this white paper at the Geneva Digital Week to the UN Secretary General and his team and at the World Artificial Intelligence Conference in Shanghai. We were asked by the Secretariat of the CBD Convention to present it at the official side event at CBD COP17 in Armenia in October, and we have received the support letter from Simon Steel, Executive Secretary of the UN Framework Convention on Climate Change. We also plan to present it at the UN Water Conference at the end of the year in December, and in Nairobi at the 21st meeting of the UN Internet Governance Forum. We are very grateful that we can present this white paper here today as part of this Magnificent expert panel. We have already begun working on the white paper in around 2027 supported by the Fudan University Internet Governance Forum Food and Agriculture Organization. We plan to present it next year at the second global dialogue on AI governance in New York City. Here with I want to thank very much all of you for your time and attention and wish all the participants and organizers a very successful conference of the parties. Thank you.

Thank you. Thank you so much. Actually, it was a wonderful presentation from your research and as well as some findings. So now we will move on to some global good practices and solutions from Dr. Bilal Jemoussi from the International Telecommunication Union.

Yes. Thank you very much. Very good afternoon to all of you. Thanks, Murali, and UNCCD for your kind invitation to this forum. It is actually my first time in Mongolia. How many of you are here for the first time in Mongolia? Great. Well, thank you very much, Mongolia, for hosting us. I think it's a fantastic opportunity. And what I would like to talk about today is a balancing act between AI and the environment. Morali mentioned that in 2017, we launched the first AI for Good Global Summit. because we saw the potential of artificial intelligence to solve humanity's biggest problems. Healthcare, for example, transforming healthcare by supporting earlier diagnostics, more personalized care, improving health systems, planning and faster response to public health emergencies. In agriculture, AI can help improve crop monitoring, facilitate the efficient use of water and other resources. In climate action, AI can analyze complex environmental data to optimize energy grids, track global carbon emission, and accelerate the development of clean technology to combat climate change. In disaster management, AI can strengthen forecasting systems, improve risk mapping and help communities prepare for floods, wildfires, and other natural hazards. In cities, AI is enabling the cityverse using AI, spatial intelligence, digital twins to help cities better plan their infrastructure, anticipate risks, improve resilience, and make more sustainable decisions. So these are the powerful applications of AI for good that are helping us solve humanity's biggest challenges. But we need to look at the aspects of the implication to the environment. Behind every AI system are data centers, telecommunication infrastructure, compute platforms, electricity, cooling, water systems, material and land use. So as AI scales, and Murali mentioned the journey with AI, initially machine learning models, then ChatGPT came and everyone heard about AI or generative AI. Then you have embodied AI in robotics, and today we talk about agentic AI. So all of these applications of AI are demanding more and more data centers. So as we scale, we need to look at the infrastructure that supports this artificial intelligence. And this brings us to the fundamental question, how can we properly assess both the environmental impact of AI and the environmental benefit that it enables us? And that's where ITU published the report on measuring what matters, closing the gap in assessing AI's environmental impact, reviewing existing approaches to measure AI's environmental footprint. It showcases the importance of greater transparency, standardized metrics, reporting frameworks, and data sharing. Further, ITU standard recommendation ITU-T L.1801, and this is publicly available, you can download it, provides a first of its kind international methodology for assessing the environmental impact of AI systems across the entire lifecycle, from hardware and infrastructure to training, development, operation, and end of life management. The standard does not only assess the impact of AI systems, it also enables the assessment of the environmental benefits. And this is where the balancing act comes in, that AI can deliver across the sector, helping stakeholders better understand the net environmental implications of AI development. And this is important because sustainable AI is not solely about measuring AI's footprint, it is equally about understanding where AI can contribute to reducing the emission, improve resource efficiency, and support environmental objectives. And methodologies alone are not enough. And this is where we are encouraging stakeholders to implement the ITU standard L1801 and share data. Because as part of the implementation, there is a data reporting and sharing activity, implementation results and lessons learned. This will help us strengthen the evidence based on sustainable AI and support broader efforts to improve transparency and deepen our understanding of AI's environmental impact. And the sustainability of AI depends heavily on the sustainability of the infrastructure it enables. And this is the area where ITU has been working on for many years, uh, in terms of developing environmental standards, measuring the ICT impact. I see that the time signal is up, so I'll be wrapping up quickly. The work spans guidance on designing, building, and operating sustainable data centers, improving energy and resource efficiency, optimizing cooling systems, reducing emissions across the lifecycle of digital infrastructure, and integrating sustainability considerations into procurement, operation, decommissioning, and recycling practices. Thank you.

Thank you. Thank you so much. That was some practical solutions also we are searching through to your presentation. So finally, we will search for some alternate solutions. That's why Dr. Gayathri Nayak is here. She will share her thoughts on meeting the growing energy demands of data centers through renewable energy. Yes, Dr. Gayathri.

Thank you. Thank you, Amrita. And thank you, UNCCD, once again for this kind invitation. I'm very happy to be part of this very insightful discussion. So those of you who do not know about IRENA, we are an intergovernmental organization. I'm sorry, I think I should. So we are an intergovernmental organization. We support countries in their transition to renewable energy and a sustainable future. And we work to provide capacity building technical advisory services to individual country members. And we have been, we have around 172 countries and the EU as our member, we who decide on the strategy and the direction of the work we do. So in this presentation, I think I'm summing up all what has been said by the panel here, so I'll be very quick. Why do we need data centers? I think you've heard from all of us here that data is very, very critical going in the future. We need data for every system. I'm maybe talking only from a power system perspective, but I think every other system operator needs data for its futuristic operation. And if I talk from a power system perspective, data allows you to forecast demand, generation, provide information and insights on what kind of contingencies can happen so that we can provide you with reliable and affordable energy. So data is going to be an important part of our lives going forward. It has been actually. So data centers, like I think Dr. Martin also said, We have conventional data centers and we have AI data centers. So the usage of AI data centers started emerging after the AI tools came into play. Conventional data centers, while we were dabbling with social media, everything was being processed somewhere as a hard infrastructure, where data centers were the hard infrastructure which is processing it. And with AI data centers, it has increased multiple times, or multi-fold, I would say. And he also mentioned about training and inference, where the training is basically to train these large models to come up with more and more accurate results. And inference is where you would be the user and inference, it would give you the right answer at the right time. So basically, training is something, and why I'd like to make a distinction here between training and inference, is while you are training a model, the energy requirement could be or possibly or potentially shifted. So you can train a model maybe in the middle of the day when renewable energy is at its peak, whereas inferencing is something you need to be real time, almost real time. So that means you cannot shift that usage to a later time, and that has great implications on the grid. So if you are taking power from the grid, then you also need to be capable of providing a bit of flexibility, which is a keyword. It's a buzzword if anybody wants to check it out. It's one of the key reasons why the grids are not functioning the way they should. And training and inference can actually offer that flexibility. So data centers can be a very flexible load going forward. So in addition to data centers demanding a lot of land, as in data centers themselves need land, plus they need land if they are dependent on renewable resources, they also have need for water resources, they also can create a lot of noise pollution depending on what kind of generation sources they are using, and of course the stacks also generate a certain amount of noise. In case of water resources, why do they need it? Because they create a lot of heat and that has to be cooled. Again, here you will also have an implication on energy because if thermal generation units are being used to power the data center, it means you're also generating a lot of heat in that sense. So if you're using renewables, it is implicit that you're reducing the amount of heat generated because of the data center application. So what are the kind of data centers that we have right now? We have enterprise, co-location, co-hosting, AI, edge, and hyperscale, all based on the different kind of organizations that use them, the size of them, and where they are sited. So hyperscale, I've provided the definition on the slide. It's one of the key kind of data centers that's coming into play from these large organizations. And one key aspect of citing is latency, where the data centers try to provide information as quickly as possible, and network communication is-- internet communication is very, very accurate, or I can say very good. So that's one aspect. And IRENA has been working to define something called as a green data centers. This is basically to-- identify if and going forward when data centers claim that they are green data centers, how accurate that is. So how is the combination of resources they are having? Is it hybrid? Is it from renewables or not? And what kind of certifications can be provided? So we are working with a lot of industry members on this and we'll be shortly coming out with a lot of other information on this as well. Lastly, but not least, how to cite. So there are a lot of points that has been discussed here, but for me, one of the key concerns is about the supply or the generation or the power that can be supplied to it. So expanding, there are a lot of points in the power sector that can be leveraged to supply data centers. And one key aspect is to optimize the existing infrastructure that you have. So instead of going on deploying renewables more and more, rather you should take a look at your own system and understand if the current infrastructure can be optimized somehow to continue providing the power that is needed for the data centers. And I explain more in the ensuing discussion and questions on succession as well. Thank you.

Thank you. Thank you, Dr. Gai. I think I know that the esteemed panelists, you're only six minutes for presenting your findings and evidence. But anyway, we are planning, we need a very critical analysis as well as a very great discussion about all these findings and research. That's why now this floor is open for all the question and answer session. Okay. Yes, I think we have volunteers. Yes, you will receive the mic. The microphone. Yes. Yes, please.

Thank you. So it's to my understanding that AI relating to analyzing the environment and satellites is low compute, which is great. However, in the program and budget for 2027-28, it was noted that the Secretariat is now using or planning to use AI for data analysis as well as decision making, which to me sounds more like potentially LLM use, which could be higher compute. So how are you, what is the best way to determine if an AI use is generating a net positive impact at the UNCCD and how when you're seeing the effect on paper, how do you make sure that something that might seem like a small effect isn't actually devastating a vulnerable community?

Thank you. I assume that question is directed to us, but as much as to other organizations as well. How do we ensure that our data use of AI is in a way net positive? That's a question. I must say, like all of us, the organization of UNCCD is also learning to use AI in an ethical and responsible manner. As of now, I was at the ITU AI for Good session in July. More than 50 UN organizations are already using AI for different purposes. And there's no unified standard on what is an ethical use, what's a sustainable use, et cetera. UN has come out with some guidance, but each of the agencies is still trying to prepare its own guidance, learn from each other, et cetera. So as of now, I would not claim that we have a clear answer to that. But the next one year, We'll be working both within UNCCD, but also with other UN partners and others, including looking at the new European Union legislation, new legislation emerging in other parts of the world so that we could come out with better guidance to our own staff, but also globally as to what would be a more ethical, sustainable use of AI. Thank you.

Yes.

Yeah, perhaps I could add to that. I think it's an excellent question because it goes along the potential and the application of AI to make efficiencies in organizations. In the ITU, we've been using AI for enhancing the translation of documents for the past five to six years, and our efficiency of translating documents has increased by the translator from 20 pages a day to 25. We use AI to generate videos. So in the past, we used to have humans record videos in one language, but now we use avatars with video generation in the six languages of the UN. So these are amazing applications of AI that reduce energy use in other ways. So I think we need, as I mentioned in my opening remarks, we need to look at the balance. There is a lot of potential, there is a lot of optimization. AI is used to generate reports quickly. We use it for captioning of meetings. So as we have this meeting, an AI captioning is working and we can get the report generated. almost instantly after the meeting is done with all the action items. So these are all the uses in the secretariats across the UN system. In fact, we have a report published by the inter-agency working group on artificial intelligence in the UN showing more than 800 different applications of AI across the UN system. This is publicly available and I can share with you the database.

Just a short comment from my side. So we see that there's a big trade-off between economics and environment. You know, it's always a trade-off. And we definitely think that new technological solutions, including AI, not only AI, not only just focused on AI, but all technological solutions can be a balance between environment and economy. So that can help, that can help to balance at last these two things.

Thank you to all of the panelists and the chair. So my question is open to all of you whoever would like to take. We were discussing how AI can be used to reduce the emissions. This was, you mentioned it. You see that in the scientific community, we are researching on quantum computing right now and how people are estimating that maybe within the next decade, maybe before or maybe even after, it is going to aid AI and maybe AI emissions might decrease. But here is the caveat that quantum computing might also require huge data centers and huge amount of power and huge emissions. So it's a general question on the global level. Has there been an eye on quantum computing? Have people been looking at it? How it's going to aid or maybe, you know how when ChatGPT came, next day everyone had it on their phone. So with quantum computing, if it just comes out, you can't stop technology from spreading. So is there an eye on it? And is there an eye that we can help AI with it? So that's my question open to all of you. Thank you.

Maybe I could start. Absolutely. We have been looking at quantum information technologies in the ITU. When I say the ITU, it's not the secretariat. It's really 194 member states and about 1,000 private sector entities and universities. Since 2019, we had a focus group on the uses of quantum. And there are three areas where quantum is useful, computing, metrology, and communication, or secure communication. And in terms of your question on quantum, in the past two AI for Good summits, we had a quantum for good session to see how quantum is helping also solve humanity's biggest problems, including the environmental impact. Of course, quantum computers are very efficient in solving certain problems. And AI processing computing is also quite intense. So we've been looking in these Quantum for Good sessions on how quantum computing can help solve the compute use of energy, because quantum computers are more efficient than traditional computers, but they're not there yet. There are only niche applications of quantum computing today to generalize it for broader use of AI is still an ongoing process, both in the research and the development. Many companies are bringing quantum computers to the industry that will be addressing the AI compute element.

Thank you. add to that point and say that, you know, Urwa is a computer engineer as well as somebody who is interested in communication. Now, one of the challenges with quantum computing is that there are very few people who actually understand quantum computing. Everyone talk about it, but not very few understand it. I personally don't understand much of it as to how it will aid one way or the other. I would go back to Dr. Jamosi's point that unless we know How much impact it has, of course we cannot manage it, so the starting point would be to understand what exactly are its contribution to AI or otherwise, and whether it's positive or negative, then we can start to address it. I actually did not know that AI for good actually have a quantum session. I think I actively avoid quantum computing, but clearly the fact that they are keeping an eye on it is a very good thing. And I'm sure institutions like UN University may also be looking at this. Right now, I'm not aware of it, but at least ITU is on the board and maybe there are other actors as well, but certainly something which is worth following. Thank you.

Yes, Gayatri.

So, I mean, as Irina, we also don't work on quantum computing per se, but when you mention about how much power it is going to need or it's similar to AI, the increasing influence of AI. So I think I also mentioned in the previous, during the presentation about looking at how to optimize the present structure of the power sector and not going for implementing more and more renewables or others and also looking at solutions like microgrids which allow you to Have access to multiple sources at the same point, so it's not only looking at wind or solar at one place; maybe it's possible to have a biogas plant, or there are a lot of microgrid applications that are coming up for AI data centers. And one aspect is, so either you can shift from renewable base, which is land based renewables to non land based. So maybe looking at offshore is a good solution. Offshore has been around for a long time, but application for AI per se is something which has to be explored, I guess. And another example I recently read is from China, where they have started deploying data centers under sea. So that is one option where it serves a dual purpose of reducing heat as well. So possibly, yeah, space as well. So yeah, these are some of the solutions for this thing. But all in all, I still would say, Application of AI to power sector is equally important because it reduces the implication of generation sources. You can kind of optimize how much resources you have and reduce that implication if you have a very good forecasting service by AI. So that's one of the uses, yeah.

Yes.

Yes. Again, a short comment from my side. So quantum, yes. Photon, definitely. You know, so there are a lot of other types that we probably don't know yet how it will develop. So what we say from our side is be smart. When you start spotlighting of the country, you know, development plan, how long term and high level of the country, the region has to be developed, be smart about where the data centers must be done. So be smart from the beginning because you don't know where the quantum will be on next or photon or something else. We don't know, but be smart from the beginning.

Yes.

Uh, this, uh, thank you so much for, uh, giving us very nice, uh, you know, opportunity, uh, to discuss on, uh, these issues, uh, to learn and also, you know, uh, some of the questions are very nice as well. My question is, uh, to any of you that, uh, the standards that ITU has developed, is it, uh, targeting, Are the standards targeting the public sector or the private sector? If it is public sector, then the government regulatory mechanisms and the policy frameworks that they will include and they will make sure that the private sector is following all the standards while developing their data centers. And if it is targeting, you are targeting to the private sector, then different set of indicators will be there and the strategy will be there. So my question is, are you targeting public sector? the government policies and institutions to follow the particular, uh, you know, standards to incorporate into their existing, uh, environmental impact assessment rules and guidelines. Uh, or if you just, uh, focusing on the private sector, then they will incorporate it in their ESG, uh, frameworks. So, uh, those are two different issues and we are at the, not in the middle of the journey, it is just, uh, we are in the infancy. So in such a, an, uh, stage, Uh, what should be the direction or you are just trying to mix, uh, both the way private and public sector? That is, uh, the question. Okay.

Excellent question. It's actually both. So the development of the standard itself is done by the public and the private sector. So the study group that develops the standard has members from the private sector that is building data centers. telecommunication infrastructure, cloud computing, all of the elements that are using energy and using resources. But it also has the government, the representatives around the table. So the development of the standard is a joint activity. And then the published standard is a voluntary one. And governments can use it and say that we would like you, if you're going to, we are going to procure a data center to comply to this standard. And so that we can, we are able to measure the environmental impact to look at the entire life cycle. We talk about the digital product passport that looks at the life cycle of the, of the product. So the implementation of the standard. Because most of the products come from the private sector, it's really by the private sector. What the public sector does is use the standard as a way to measure whether the private sector deploying is keeping up to the limits that the governments wish to see. And as I mentioned, the standard looks at both aspects, not just the consumption of resources and energy, but also how AI can help save energy by through its application. So the equation is basically the overall environmental impact is what resources are you using minus what resources are you saving. And that's what the standard data collection is about, to look at that equation and how to balance it.

If I could just add to that point. We in this panel as we are putting together, we also wanted to have private sector representation, but the time available was too short and the private sector, you know, has many internal barriers to joining such a session at this point of time where the industry is, you know, a little bit of under pressure on this topic, but definitely what we have to work with the private sector. So from the UNCCD side, we are working on two, activities to engage the private sector. The public sector we are already engaged. One is that the next environmental impact assessment, International Association for Environmental Impact Assessment meeting, which is in Christchurch, New Zealand, we are having a specialized session for the private sector on how do we assess the environmental impact of AI and data centers, because data centers are coming all over the world, including in Mongolia and many of those Those who are governing actually don't know how to exactly do the EIA, especially on life cycle analysis. And the consultants who do the EIA also don't know how to do it. So therefore, we are trading on that. We are also planning to work with ITU on the next AI for Good Summit, which he will actually talk about in the end, to bring more privacy actors into the discussion so that we have a more rounded session. Thank you.

Any other questions? Yes, here, please.

Good afternoon. I'm Karen Sudmire with the G20 Global Initiative. Thank you for this very enlightening discussion. I'm also a prolific user of AI, but I'm concerned about its environmental costs. Have there been any initiatives by governments to recuperate some of the external costs? tax? There is, I would say, good AI and less good AI, like for example, Bitcoin. I'm not really into Bitcoin. I'd like to tax Bitcoin because I think there's lots of costs related to that kind of AI use. Are there any governments that are taxing private sector companies? Since you're not here, we can talk about that, right? So.

A few months back, I had an opportunity to meet with Dr. Monique Barbot, who is the environmental minister of France, who came to visit UNCCD. And I asked her this question, that there is this huge AI usage and the debate coming up. There's a question of AI and environmental sustainability. But also there is this question of AI going to destroy a lot of jobs that a lot of people may not have jobs in the future. There's a fear. Whether it happens or not, we don't know, but there's a fear. So I asked her, have they ever thought of something like a token tax? You know, AI consumption is measured in terms of the number of tokens you use. And in advance, Bill Gates proposed that if you are using a robot to replace a human being, you could have a robot tax imposed on the robot and then give that money to the people who are displaced. So similarly, could we have a token tax on AI? And she mentioned that, no, we don't have one, but she said, of course, France is very good at taxing, and therefore, if there's a possibility, we'll definitely find a way to do it. So at the moment, I don't know of any country who actually do a token tax, but this is being widely debated, not just for environmental cause, but also from the employment angle. I don't know if any, and one of the panelists are aware of any country who actually use it or in an advanced stage of discussion. Thank you.

Yes. Any other queries and discussions? Yes, yes, please.

Again, quickly from my side, so we are from environment. As I've told, we got nothing to do with technological team and we are very glad to be here. So for us, it's a great honor to be here at this table and to discuss this question right now. When I was in India, for example, I've heard only one person speaking about environment and AI. It was Yunus Ji and he said climate change. Nothing about biological diversity loss, nothing about land degradation. It was like one topic. When I was at AI for Good, I've heard another person speaking about two topics. He said, Climate change, biological diversity loss. I said thank you. And right now we hear like land degradation. So I mean, we are moving. Plus to that, UN SG have launched in London, I think, the affordability and transparency and affordability initiative. So right now, United Nations got a vision of where to go. And from our point of view, when we are talking about the AI regulation, future of AI, It's all about United Nations. That's what we think, and because United Nations can do good things and know precisely well how to make them happen. That's what ITU showed in us with all these trends. I mean, they got legislation, and then they put it on the local level. They know how to do it. No one else in the world knows how to make it right.

Thank you. My question is actually mostly directed towards Irina. Would it be the idea of a sustainable data center is very compelling, but would it be possible for an organization like the UNCCD to ensure that all of their AI use is only going through that more ethical data center?

It is possible if the definition of what we are coming out with is green data centers with regards to renewable energy usage. So if We have a definition which says that the data center that you're coming up with needs to have a more sustainable use of land, or there is a definition that is, you know, you need to certify. There is a certification process which includes the use of land, then it is possible to have it. But I'm not sure how many organizations have such a process in place, but it also needs to come from the highest regulatory levels. And I guess it will come up in due course of time because the impact of data centers is not less in terms of land, right? So it will start. Yeah.

Yes. Any other questions? With your permission, actually, I have a doubt because I have gone through it, the research by Dr. Mir Martin. So I think I have a doubt. Actually, your research actually shows some clear trade-off between carbon, water, and land footprints, really. So I want to know that what would be the best way for the policy makers to weigh these trade-off when choosing energy pathways for AI infrastructure?

Thank you. I would take the question in different way because there is a trade-off, but there is no single trade-off that you can apply at a very large scale. For example, when you do a trade-off at a country level and you say this is the best way to do this, this is the energy pathway, that will not work because For example, the carbon target for a country is a global target, but then when you data center actually impacting at a local scale where the data center is being established and the electricity grid that are connected to that data center and then generation happens wherever it is. So which means yes, we need to go to trade off, but we also need to be careful that what process we apply for getting the trade-off. So which means in terms of AI data center, the trade-off has to be measured or analyzed on the sighting decisions. For example, where do you do the data center? For example, a carbon friendly data center But if it is a electricity generation is a very land impactful, then it's not working in that case. And at the other end, like a data center on a water stress basin, and then you have a green energy, which is also very water impactful is not working. So, which means we want to also say that You need to do the trade-off, but trade-off calculation and decision processing has to be done at the sighting decision, at the local scale. You cannot have a very global trade-off that, okay, this is the best way to do that at the country or at the regional level. So that will not work. But we have to do the trade-off, but we also need to say where we apply the trade-off. Thank you.

Thank you. And also anyone from the panel can contribute there on my behalf.

I want to add to that point. I think it's very, very important. that normally when you say AI or data centers should be using renewable energy, you think it's a very good idea. But when you put in a different context where there's not enough renewable energy and you have to convert agricultural land, to put up solar panel so that the local data centers can be powered, then it's a very bad idea. So it has to be very context specific and putting a mandatory national target might often lead into unintended consequences. Very important. Thank you.

If I could add to that, there is a project in ITU by some of our members, instead of bringing power long distance from the power generation to the data center, which is very costly because there is a lot of leakage of electricity in the transport, you try to have the data center as close as possible to the power generation site and then use fiber optic to transmit the data, which is a little more efficient. Fiber optics are not very hungry when it comes to energy use, so this is one of the standards that we're working on in the ITU to bring data centers and power closer, but then use fiber optic for the communication to the user and to the, the training material.

Exactly, thank you. So, I think this is the time frame for our discussion throughout the panel. Now, I would like to present a video presentation which is done by Doctor Bilal Jemausi from the part of ITU. Please.

I asked you a question. What will you do with your AI agency? More than 12,000 of you from over 170 countries, well, you answered.

The title of this summit contains a question.

Good for whom? A technology is good when the human beings whose work made it possible are not erased by.

At IQ events on AI, we always discuss three top priorities, solutions, skills, and standards.

Time is running out, but it's not too late to unmask AI and protect what is human in a world of machines.

To invest in artificial intelligence is to secure our future and that of the generations coming after us.

We are very pleased to announce that we are launching a focus group in ITU. AI is a global matter, and I'm hopeful that if enough citizens understand the level of the stakes, things will change, governments will move.

When I talk to people about AI back then, they'd say, what's AI? And I tell them that AI is going to make computers as intelligent or more than humans.

But how do we actually explain AI for good? AI for good is the prompt, and we, us humans, are the model.

You show the world the true meaning of AI for good.

Thank you, Dr. Bilal and your team. I think it's Dr. Bilal, please, a few words.

Thank you very much, so that was a wrap up of the 2026 edition of AI for Good that several of our panelists were there and, and mentioned today. And as you can see from the video, there are so many applications of AI for good. There are so many world leaders. We had three presidents speak at the opening, many heads of governments, ministers, many youth, as you saw. We looked at the skilling of youth from the age of 10 to 18 with the robotics competitions, the challenges with the universities, the startup pitching competitions, because with AI, it's all about developing skills, especially in most of the world. There are two big giant countries have AI, but the rest of the world is using AI. And so it's important to develop the talent. So I want to close by really inviting you all to pencil in your calendar, 21 to 25 June, 2027. There will be the Swiss AI summit following from the Indian summit, from the French summit, from the UK, from the Korean. The next one will be back-to-back with AI for Good. The Swiss are hosting a global political summit on AI and the AI for Good will be 21 to 25 June. So hopefully you can join us and there will be a lot more progress on our topic of AI and the environment and AI and also quantum computing. Thank you.

Yes, thank you, Doctor. I think it's a very open invite, so we are really fortunate. So, anyway, we are moving with the closing remarks with the things actually. Now, the point is the growth of AI is not only a technical transformation alone. Now we can wind up like this. Our distinguished speakers have brought together different and but deeply connected perspectives from evidence and measurement to land use pressure, renewable energy, international standards and environmental governance. A central finding from our discussion is that what we cannot adequately measure, we cannot responsibly govern actually. So greater transparency, reliable assessment and stronger standards must therefore be translated into thoughtful land use planning and sound public policy aspect. So we are addressing these questions at a time when choices are still possible. We can plan more carefully, locate infrastructure more responsibly, use resources more efficiently and create governance system that are both environmentally sound and socially fair as well. So may today's conversation move beyond this room and into a flourish into the research, policy, international standards, and responsible decisions on the ground I am really expecting from as an output. So thank you. Thank you to our distinguished speakers and to all of you for making this such a meaningful and timely discussion in this panel. Thank you so much.