What happened with artificial intelligence energy demand?
The extraordinary growth of artificial intelligence in 2024-26 has come with an extraordinary and increasingly visible cost: electricity. Training large AI models and serving billions of AI queries every day consumes amounts of electrical power that are beginning to strain grids, alarm climate scientists, and create new strategic calculations for both technology companies and governments. As India simultaneously pursues its own AI ambitions and deals with record-breaking summer electricity demand that is already causing power cuts in Delhi-NCR and other cities, the AI-energy challenge has arrived with particular urgency.
Key Points
- Training one large AI model can consume as much electricity as thousands of Indian homes use in a year
- Global data centre electricity demand projected to double by 2030 — largely due to AI
- Microsoft, Google, Amazon are investing in nuclear power specifically to run AI data centres
- India's peak summer electricity demand is already at record highs in May 2026
- India's AI data centre ambitions will require massive new power infrastructure investment
- AI-powered by renewable energy can be relatively climate-friendly; AI-powered by coal is not
Background
The energy cost of computing has always existed but has become dramatically more significant as AI scales. A simple Google search consumes a tiny amount of electricity. An AI query to a large language model consumes roughly 10 times as much. Training the model that powers those queries requires thousands of times more energy than running it. As hundreds of millions of people interact with AI assistants daily and companies train new, larger models continuously, the cumulative electricity consumption is becoming a major planning challenge for power utilities worldwide.
Main Details
The International Energy Agency has projected that global electricity demand from data centres could double by 2030 — moving from approximately 200 terawatt-hours annually in 2022 to over 400 terawatt-hours — driven primarily by AI workloads. This is equivalent to adding the entire electricity consumption of Japan to global power demand within eight years.
In response, major technology companies are making unprecedented investments in power generation. Microsoft has announced it will restart the Three Mile Island nuclear plant in Pennsylvania specifically to power its AI data centres. Google has signed agreements with nuclear operators to supply clean electricity. Amazon has invested in small modular nuclear reactors.
For India, which is simultaneously expanding its AI ambitions through the government-backed IndiaAI mission and dealing with record-breaking summer electricity demand that is causing power cuts across Delhi-NCR in May 2026, the AI-energy challenge has direct policy implications. Building AI data centres in India requires not just land and connectivity, but reliable, affordable power — which is currently under strain.
India's electricity generation mix includes significant coal capacity alongside rapidly growing renewables. AI data centres powered by coal-based electricity would significantly increase India's carbon emissions — complicating its climate commitments. Renewable energy-powered data centres would avoid this problem but require co-location with renewable energy capacity.
Reactions
Environmental organisations have raised concerns that the AI boom could slow global progress on climate goals. Technology companies have responded with renewable energy commitments, though critics note a gap between stated commitments and actual data centre energy sourcing in practice.
India's power ministry and the Ministry of Electronics and Information Technology are jointly evaluating data centre policy to ensure that AI infrastructure expansion is paired with appropriate power infrastructure investment.
Impact Analysis
India faces a strategic choice: it can pursue AI data centre development aggressively, attracting global technology companies and building domestic AI capacity, but this will require major power infrastructure investment and could strain an already-pressured electricity grid. Or it can proceed more cautiously, ensuring power supply keeps pace with data centre demand. The optimal path involves accelerating renewable energy deployment specifically tied to data centre demand — a model several Indian states are exploring.
What Happens Next
India's data centre policy framework is expected to be updated in 2026-27 to address the AI power demand question. The IndiaAI mission is evaluating computing infrastructure partnerships with global technology companies. Citizens will notice the AI-energy tension most directly through summer power demand — AI data centres add to the same electricity grid that their homes, factories, and offices depend on.
FAQ
Q: How much electricity does AI use?
A: Training a large AI model can consume tens of thousands of megawatt-hours. Running AI at scale across billions of daily queries adds enormous continuous demand.
Q: Is AI bad for the environment?
A: It depends on the electricity source. AI powered by renewable energy can be relatively clean. AI powered by coal contributes significantly to carbon emissions.
Q: What are tech companies doing about AI's energy use?
A: Microsoft, Google, and Amazon are investing in nuclear and renewable energy specifically to power AI data centres.
Q: How does India's electricity shortage relate to AI?
A: AI data centres add to electricity demand at the same time India is struggling with summer peak demand. Without power infrastructure investment, AI growth will be constrained.
Q: Can India become a global AI data centre hub?
A: Yes, but it requires parallel investment in renewable energy, grid infrastructure, and data connectivity. The potential is significant given India's large market, talent pool, and cost competitiveness.