AI’s insatiable appetite for computing power is rapidly reshaping global priorities in energy and infrastructure. A 2024 McKinsey report projects that demand for AI-ready data centre capacity will expand at an annual average rate of 33 per cent between 2023 and 2030, with nearly 70 per cent of this growth geared towards infrastructure capable of handling advanced AI workloads.
This transformation is not confined to the developed world; India too is experiencing a surge in data centre investments. As of April 2025, India’s total capacity stood at 1,263 MW, with projections of more than $100 billion flowing into the sector by 2027, growing at a CAGR of over 24 per cent between 2023 and 2029. Mumbai, Chennai, and Delhi NCR have emerged as the country’s main AI hubs. But hidden beneath this momentum lies an irony: the very infrastructure that underpins large language models (LLMs) and other AI systems is beginning to complicate global energy transition goals. According to the International Energy Agency (IEA), data centres consumed about 415 terawatt hours (TWh) of electricity in 2023, roughly 1.5 per cent of global consumption.
That figure has grown at nearly 12 per cent annually over the past five years. The IEA now projects demand to more than double to 945 TWh by 2030, exceeding Japan’s entire electricity use today. This escalating demand is already straining national grids. Ireland provides an early warning sign In 2023, data centres accounted for 21 per cent of the country’s electricity consumption, a factor contributing to
rising blackouts.
In response, authorities froze approvals for new centres around Dublin until 2028. Elsewhere, the pressure is prompting governments and companies to turn back to fossil fuels as a stopgap. In the US, federal agencies were asked this year to identify retired coal plants that could be repurposed to supply electricity to AI data centres.
Pennsylvania’s former Homer City Generating Station, once the state’s largest coal-fired plant, is being redeveloped with a $10 billion investment to house seven gas-fired turbines dedicated to powering on-site centres. Similarly, utilities in Georgia have delayed shutting coal-fired plants that were scheduled for closure under earlier decarbonisation commitments, citing the electricity needs of expanding data hubs. The pattern is visible globally. In the UK, demand for AI and cloud facilities is outpacing the grid’s ability to provide stable electricity, pushing developers to explore direct connections to main gas pipelines. Singapore, after imposing a moratorium on new data centre projects in 2019 due to energy concerns, lifted the freeze in 2022. It now hosts over 70 facilities and plans to raise capacity by nearly a third in the coming years. China, with the largest concentration of data centres in Asia, still relies on coal for more than half of its power supply. Most of its facilities remain concentrated in coal-heavy eastern provinces.India’s story is more complex. While the country is racing to establish itself as a hub for AI infrastructure, its coal-heavy energy mix — where fossil fuels still account for over 70 per cent of electricity generation — raises tough questions. The national grid is ageing and already stressed by peak demand surges and transmission losses. Unless carefully managed, the rapid buildout of AI-ready data centres could amplify these vulnerabilities.
At the same time, India has an opportunity to chart a different path. Retired coal plants are already being earmarked for nuclear energy projects, though not yet explicitly linked to AI-related demand. With careful foresight, India could ensure that new AI data centres are built in tandem with renewable energy integration, smart grid upgrades, and strong efficiency standards.
This would prevent the country from locking into carbon-intensive growth patterns and instead position it as a model for balancing digital and energy transitions. The rise of AI does not have to stall global decarbonisation. With proactive policies and open public debate, India and other fast-growing economies can turn a potential energy liability into an asset. By aligning data centre expansion with renewable power and grid modernisation, governments can not only meet the needs of the AI revolution but also accelerate the shift towards resilient, sustainable energy systems. In this sense, the challenge of powering AI could become a catalyst for strengthening the very foundations of the clean energy transition.
The writer is a Research Assistant at the Center for Policy Research and Governance