As artificial intelligence becomes central to digital transformation, the demand for computing power is rising at an unprecedented pace. From generative AI models to cloud services, the technology’s growth hinges not just on sophisticated algorithms and cutting-edge chips but also on the physical backbone of data centres. These digital warehouses consume vast resources — electricity, water, and land — to process and store the world’s ever-expanding volumes of data.
Global tech giants have responded with massive capital outlays. Microsoft plans to invest $80 billion in AI data centres by 2025. According to a 2024 McKinsey report, global demand for data centre capacity could triple by 2030. The momentum is no longer confined to the West. India, too, has seen large-scale announcements from Reliance Industries and the Adani Group to build and expand data infrastructure. Southeast Asia and Africa are witnessing similar activity.
READ | YouTube’s pay-per-view model challenges OTT giants
The proliferation of data centres is not merely a technological necessity — it is also a strategic imperative. Geopolitical tensions have reinforced concerns about data sovereignty and national security. Governments and corporations alike are moving to localise critical infrastructure, further fuelling demand for hyperscale facilities across regions.
AI data centres: The energy equation
Data centres already consume an estimated 1–2% of global electricity, a share that could rise to 3–4% by 2030, according to Goldman Sachs. The same report projects a 160% increase in data centre power demand, driven largely by AI. In smaller or densely populated economies, the figures are even starker. In Singapore, data centres already account for 7% of total electricity consumption—a share expected to grow to 12% within five years.
While technological advances have made data centres more energy-efficient over time, these gains are being overwhelmed by the explosive growth in demand. The sheer scale of computational workloads required for training and running large AI models necessitates ever-larger centres, with energy requirements rising accordingly.
The latest Synergy Research Group findings underscore this trajectory: the total capacity of operational hyperscale data centres is set to nearly triple by the end of the decade. The implication is clear—without a significant pivot toward cleaner energy sources, the AI boom could come at the cost of climate goals.
Will AI growth derail decarbonisation
The global conversation on AI has so far focused on ethics, bias, and data privacy. But lurking beneath these concerns is a structural risk: AI’s growing energy appetite may outpace the expansion of clean energy capacity. The result could be a net increase in global emissions, unless systemic interventions are made.
Some of the world’s biggest technology firms are now exploring nuclear energy as a potential solution. Microsoft, for instance, struck a deal to revive the long-defunct Three Mile Island nuclear plant. Google has partnered with Kairos Power to build seven small modular reactors to supply its data centres. These moves signal a shift from symbolic green commitments to hard-nosed investments in alternative baseload sources.
Nuclear energy, while less volatile than solar and wind, is not without its challenges. Construction timelines are long, costs are high, and safety concerns linger. Accidents—though rare—can be catastrophic, as the world saw with Chernobyl and Fukushima. While modern reactor designs promise improved safety, the technology still faces regulatory, financial, and public trust hurdles.
Even if these constraints are overcome, nuclear is unlikely to carry the full load. Goldman Sachs estimates that nuclear power could account for only 10% of global data centre energy needs by 2030. The rest must come from a combination of renewables and energy optimisation strategies.
Making data centres energy efficient
Beyond fuel source diversification, improving operational efficiency offers a more immediate path to reducing emissions. Data centre workloads vary throughout the day. Dynamic monitoring and scheduling of tasks during periods of low energy cost or high renewable availability could substantially improve energy utilisation.
Smart building technologies—particularly advanced chiller and cooling systems—can also reduce energy draw. Investing in energy-efficient infrastructure pays dual dividends: lower operating costs and a reduced carbon footprint.
Governments have a role to play in nudging the sector toward sustainability. Targeted subsidies, regulatory frameworks, and clean energy mandates can guide investment decisions. Singapore, for example, has announced plans to quadruple its solar energy infrastructure by 2025, with a goal to generate 3% of its electricity from solar. While modest, it reflects the kind of state-led initiative required to align national infrastructure with climate commitments.
A policy imperative, not an afterthought
As the world races to deploy AI at scale, energy use must be seen as a core concern—not a footnote. The technology sector, once celebrated for its intangibility and low footprint, now finds itself at the heart of the climate debate. Whether AI serves as an accelerant or a hindrance to global decarbonisation will depend on decisions made today—by corporations, regulators, and consumers alike.
Emission reduction at data centres must be hardwired into policy frameworks as a necessary safeguard. The push to “decarbonise the grid” will have to outpace the exponential energy demands of the digital economy. Otherwise, the pursuit of intelligence—artificial or otherwise—may come at an environmental cost the world cannot afford.
Prakash Gupta is a Delhi-based public policy professional. Bhumika Pant is a public policy scholar.