India, June 25 -- "Should we choose Artificial Intelligence (AI) growth or climatestability" is the wrong question to ask. Both are imperatives for India's future. However, the realtrilemma is this - AI needs large amounts of electricity and water; meanwhile, grids are under stressfrom energy insecurities as well as extreme heat; and the climate crisis is making both sets of problems worse. We cannot solve any one withoutsolving the other two. Each hyperscale data centre draws as much electricity as an aluminium smelter and consumes millions ofgallons of water daily for cooling - even as heatwaves and the worseningclimate crisis turn those resources scarce. Global electricity demand hit a record 31,779 TWh in 2025, after crossing 30,000 TWh in 2024. Last month, India's peak power demand hit an unprecedented 270.82 GW asheatwaves pushed cooling loads tonever-before-seen levels. Can we rethink where we set up data-centres, design new power markets, and leverage digital infrastructure to enable rather than undermine sustainable economic development? The answer is yes, but only if we abandon 20th-century power grid thinking. Consider India's geography of exposure: 70% of its operational data centres are concentrated in Mumbai, Delhi-NCR, Chennai, Bengaluru, Hyderabad and Kolkata. More than half of these data centres already endure temperatures above 35degC for over 90 days a year, according to the Climate Resilience Analytics and Visualisation Intelligence System (CRAVIS.ai), developed by the Council on Energy, Environment and Water (CEEW). By 2040, nearly 90% will face similar heat stress. Globally, over 600 data centres sit in locations where average temperatures exceed the ideal operating bracket of 18-27degC. India could see 15-40 additional unusually hot days each year in the next two decades, and 20-40 more unusually warm nights annually. That means air-conditioning loads won't ease even after sunset. For every data-centre investor, location determines exposure to heat and water stress, while market design sets power prices - all three directly affect operating costs and long-term viability. Meanwhile, the world's major economies are scrambling to fix their grids. But most are treating symptoms. The US will be spending a trillion dollars to replace ageing grid infrastructure. China is using coal plants as frequency stabilisers after a 13% wind speed drop last March, even as it races to build ultra-high voltage transmission lines. These are necessary measures, but are also defensive or reactive. The proactive move is to redesign power markets for the age of flexible, concentrated, weather-sensitive loads. That redesign rests on five shifts. First, move from bundled electricity procurement to rewarding flexibility. Short-term markets must pay for ramping, storage, and demand response. Data centres - large, controllable loads - could become grid assets by shifting computation to times of renewable energy abundance. Utilities should open participation to aggregators and virtual power plants. Second, site data centres with foresight. While electricity prices vary across time and location, developers must also account for renewable energy availability, transmission readiness, future heat exposure, and water security. These factors progressively determine both operating costs and infrastructure resilience over the asset's lifetime. Today, only five of 15 states in the country have data-centre policies that consider climate risks. Platforms such as CRAVIS can be used to map temperature, rainfall, and water availability over the next several decades, so that hyperscalers are designed with appropriate cooling and water-efficient technologies. Third, mandate transparent disclosures to drive resource efficiency and cost-effectiveness. Every new data centre must report verified power and water use and carbon intensity - just as we require energy labels for refrigerators. Also, replace rigid decadal power purchase agreements with a layered contract stack. Developers need long-term certainty to finance clean assets. Buyers want shorter terms because renewable costs keep falling and AI demand is uncertain. The International Energy Agency suggests that the tenor gap could be bridged with intermediaries or two-sided contracts-for-difference - arrangements where both the clean energy developer and the buyer share price-risk around an agreed benchmark - while preserving exposure to real-time price signals. Fourth, leverage digital infrastructure to enable power markets of the future. Smart metering is the nervous system of the future grid. India must complete its smart-metering rollout by 2028, integrating them with other operational systems of the utilities, with due attention to consumer safeguards. We can build a Unified Energy Interface on that platform. With granular real-time information, energy data can feed into a digital public infrastructure (DPI) for electrons, just as India did with payments. Proprietary platforms will not deliver flexibility, common data protocols will. Utilities have been leveraging an estimated 2-4% of the data they collect. That is a design mismatch we can resolve. Fifth, make energy and waterefficiency core to AI viability. The revenue calculus for AI is unforgiving:Globally, to sustain the $500 billion annual capex that AI infrastructure will require by 2030, cloud and AIproviders need $2 trillion in annual revenue. This gap can be reduced by locating data centres where power is clean, cheap, and cool - and where markets send the right price signals while optimising for efficiency. By 2050, space cooling could account for 44% of India's peak load, up from 11% today. Indonesia, the US, West Asia, and Brazil face similar trajectories.A data centre built without accounting for rising temperatures, humidity,and water stress is a stranded asset waiting to happen. This trilemma - involving AI demand, grid fragility, climate extremes - is one of the most important design challenges for AI infrastructure that will drive future growth. India has shown how DPI can scale. Now we must show how climate-intelligent power markets can scale alongside AI. Our children deserve advanced technology and a sustainable planet, both of which must be responsibly nurtured....