India’s AI Moment Needn’t Be a Climate Reckoning
In New Delhi, the first India Artificial Intelligence (AI) Impact Summit has brought together governments, technology firms, multilateral agencies, and innovators to examine how AI can benefit people, the planet, and inclusive progress. Anchored in the Global South, the gathering signals a shift from receiving norms to shaping them. This matters because the next billion people will come of age in Asia and Africa, whose economies, public services, and climate resilience will be built alongside, or locked into, AI systems being designed and deployed today.
India has an opportunity to model its own path. But there is a tension that this summit has made plain: AI requires vast resources. A single hyperscale facility can draw as much electricity as an aluminium smelter. In 2024, India’s data centres consumed an estimated 150 billion litres of water. By 2030, that figure will more than double. According to recent analysis by the Council on Energy, Environment and Water (CEEW) and SYSTEMIQ, only five of 15 current state-level data centre policies embed sustainability provisions.
Yet resource intensiveness is only half the story. The deeper risk is that infrastructure built without foresight erodes the very thing expansion requires—social licence. The land-water-energy nexus compounds these pressures in ways that isolated metrics cannot capture. India’s blueprint for AI must, therefore, be not only efficient but co-owned. By states, by utilities, and by the communities that host it.
The False Binary
The summit also makes clear that AI is one of our most powerful tools for climate action. It optimises grid storage to absorb more renewable energy, forecasts floods with hyperlocal precision, monitors methane leaks, and strengthens agricultural resilience. An expert group convened for the summit examined this duality across four layers—models, infrastructure, governance, and ecosystem. The findings are unequivocal: AI ambition and climate responsibility form a false binary. It only appears real when we treat the two as separate portfolios, managed by different ministries, funded from separate budgets and investment vehicles, and thought about by segregated leadership clusters.
Therein also lies the opportunity. A country with 18% of the world’s population and 4% of its water should not have to afford the model of AI now scaling in less-stressed geographies. We must build something else. Call it frugal, even Gandhian, AI—fit-for-purpose, resource-efficient, and aligned with public purpose from the start.
The Scale of the Challenge
Consider what is already underway. India’s installed data centre capacity has nearly tripled since 2020, with committed investments exceeding $95 billion. The India AI Mission is advancing application-led, socially responsive deployment. Digital public infrastructure demonstrates that population-scale systems can be open, interoperable, and oriented toward delivering public goods.
But this rapid expansion comes with environmental costs that cannot be ignored. Data centres are voracious consumers of electricity and water. In a country where water stress is already a reality for millions, the projected doubling of data centre water consumption by 2030 is not just an environmental concern; it is a social and political one. Communities that host these facilities have a right to know what they are getting, and what they are giving up.
Four Shifts for a Sustainable AI Future
Scaling this momentum requires four fundamental shifts in how India approaches AI infrastructure and governance.
First, embed environmental accountability into the core governance of AI systems. India should bring in site-level disclosure of power usage effectiveness, water-use effectiveness, and carbon intensity—verified, not just self-reported. The EU AI Act now requires foundation model providers to document training data and capabilities. India should go further: require similar disclosure for environmental footprints, and create an AI Energy and Water Star rating system that allows developers, enterprises, and governments to choose efficient models. If a consumer can know the energy cost of a refrigerator, a policymaker should know the water cost of a large language model.
Second, align infrastructure with resource realities. Analysis of 15 state policies shows that incentives currently reward investment volume rather than performance. Rajasthan’s 2025 policy is the exception—mandating zero liquid discharge, rainwater harvesting, and green building standards. This should become the template everywhere. We need integrated spatial planning that evaluates grid capacity, water stress, and climate risk before land is allocated. Not every location is suitable for a hyperscale data centre; pretending otherwise is a recipe for conflict and inefficiency.
Third, treat climate-relevant data as digital public infrastructure. India’s Energy Stack and AgriStack demonstrate what is possible when data is combined with flexible approaches to enable smart grids. Satellite launches have increased by roughly 45% year-on-year over the past five years, much faster than our capacity to convert pixels into policy. AI can bridge that gap, but only if the underlying data is treated as a public good rather than a commercial asset. It is also how we assert data sovereignty. The Global South cannot remain a source of raw data and a market for finished models; we must capture value at home.
Fourth, reskill at scale and with precision. CEEW analysis has identified 36 emerging green value chains—circular economy, bio-economy, energy transition—that can employ 48 million people by 2047 and unlock $1.1 trillion in market value. Those jobs will not materialise automatically. They require skills taxonomies that connect worker competencies to employer demand, personalised learning pathways, and institutional ownership within public systems. AI can enable this matching. It cannot substitute for the political choice to invest in it.
Learning from What Works
India has done this before. Both the International Solar Alliance and the Coalition for Disaster Resilient Infrastructure began here as ideas. The Green Development Pact was negotiated in New Delhi. Each time, we refused the false choice between growth and sustainability. Each time, we built coalitions of the doing.
The same approach must now be applied to AI. The technology is new, but the principles are not. Transparency, accountability, public purpose, and inclusive governance are not obstacles to innovation; they are its foundations. When communities trust that AI infrastructure will not harm their water supply or strain their electricity grid, they are more likely to welcome it. When developers know that environmental performance will be measured and rewarded, they will invest in efficiency.
The Hard Work Ahead
When the summit closes, the harder work begins—translating principles into procurement guidelines, disclosures into statutes, pilots into systems that outlast political cycles, and delivering for those who will inherit both the technology and the planet.
This means embedding sustainability into every stage of the AI lifecycle. It means designing models that are not just powerful but efficient. It means siting infrastructure where it can be supported without straining local resources. It means governing data as a public good. And it means ensuring that the jobs of the future are accessible to those who need them.
Conclusion: A Choice, Not a Destiny
India now has an opportunity to show the world that the most powerful technology of our time need not come at the cost of the planet. The choice between AI ambition and climate responsibility is false. We can have both, if we build deliberately, govern wisely, and invest in the human agency to exercise our choice.
The next billion people will come of age in Asia and Africa. The systems we build today will shape their lives for decades. India can lead by example, demonstrating that population-scale technology can also be planet-scale responsibility. We must build not just AI agents but invest in the human agency to choose a better path.
Q&A: Unpacking India’s Sustainable AI Challenge
Q1: What is the environmental cost of AI infrastructure that the article highlights?
AI systems require vast resources. A single hyperscale data centre can draw as much electricity as an aluminium smelter. In 2024, India’s data centres consumed an estimated 150 billion litres of water, a figure projected to more than double by 2030. Only five of 15 current state-level data centre policies embed sustainability provisions. Beyond direct resource consumption, infrastructure built without foresight can erode social licence and compound pressures on the land-water-energy nexus.
Q2: Why is the choice between AI ambition and climate responsibility described as a “false binary”?
The binary appears real only when AI and climate are treated as separate portfolios, managed by different ministries, funded from separate budgets, and thought about by segregated leadership. In reality, AI can be both a driver of resource consumption and a powerful tool for climate action—optimising grids, forecasting floods, monitoring emissions, and strengthening agricultural resilience. The challenge is to design AI systems with environmental accountability from the start, not as an afterthought.
Q3: What is “frugal” or “Gandhian” AI, and why is it relevant to India?
India has 18% of the world’s population but only 4% of its water. It cannot afford to replicate the resource-intensive AI models being scaled in less-stressed geographies. Frugal AI means building fit-for-purpose, resource-efficient systems aligned with public purpose from the outset—drawing on India’s tradition of doing more with less while ensuring that technology serves human needs rather than the other way around.
Q4: What four shifts does the article recommend for sustainable AI development?
First, embed environmental accountability through mandatory, verified disclosure of power usage, water usage, and carbon intensity, plus an AI Energy and Water Star rating system. Second, align infrastructure with resource realities through spatial planning that evaluates grid capacity, water stress, and climate risk before land allocation. Third, treat climate-relevant data as digital public infrastructure to enable AI applications while asserting data sovereignty. Fourth, reskill at scale with precision to connect workers to emerging green value chains that can employ millions.
Q5: How does India’s experience with digital public infrastructure inform its AI strategy?
India’s Energy Stack and AgriStack demonstrate that population-scale systems can be open, interoperable, and oriented toward delivering public goods. These platforms show what is possible when data is combined with flexible approaches to enable smart systems. The same principles—transparency, accountability, public purpose, inclusive governance—must now be applied to AI. They are not obstacles to innovation but foundations for sustainable, trusted technology.
