The Hindu's available RSS headline records a round-table discussion about damage associated with data centres. That headline provides a starting point for examining the wider energy evidence, but cannot establish the participants' statements or the meeting's conclusions. The International Energy Agency's published assessments explain why these facilities have become a significant electricity-planning issue.
Why a global percentage can conceal a local challenge
In its 2025 assessment, the IEA estimated that data centres used 415 terawatt-hours of electricity during 2024, representing approximately 1.5% of worldwide consumption. The agency also described an unusually concentrated industry: almost half of American data-centre capacity was located within five regional clusters. Its global percentage therefore needs to be read alongside its account of where the infrastructure is located.
The same assessment described electricity networks already facing connection queues and equipment constraints. It estimated that roughly a fifth of planned data-centre projects faced a risk of delay unless those obstacles were addressed. For advanced economies, it put construction times for new transmission lines at four to eight years. These are findings about infrastructure and planning conditions in the agency's assessment; they are not measurements of environmental damage at the meeting's location or findings about a particular proposed development.
The equipment behind a facility's electricity bill
The IEA's 2025 demand chapter separates a data centre into computing equipment and the systems that support it. Servers perform processing and storage tasks, while dedicated storage equipment provides centralised capacity and backup. Switches, routers and related networking components connect the equipment and manage traffic. Cooling equipment regulates temperature and humidity so the computing systems can operate under suitable conditions.
In the chapter's account of modern facilities, servers take approximately 60% of electricity demand on average, while storage accounts for about 5% and networking for up to 5%. Cooling varies much more substantially: its share ranges from roughly 7% at efficient hyperscale facilities to more than 30% at less efficient enterprise sites. The chapter also identifies batteries, standby generators and supporting infrastructure. These distinctions explain why a single generic percentage cannot describe every facility's electricity allocation. The agency explicitly notes that the equipment mix, its efficiency and the type of data centre affect the result.
Electricity sources change the environmental picture
The IEA's 2025 supply chapter examines the electricity physically used by facilities, including power supplied through their local grids. It distinguishes that electricity mix from the contractual arrangements an operator signs. That distinction matters when reading the agency's estimates: an account of electricity purchases and an assessment of physical generation are different descriptions of supply.
In that assessment, coal supplied approximately 30% of data-centre electricity worldwide, renewables about 27%, gas 26% and nuclear 15%. The chapter describes renewables as "the fastest-growing source of electricity" for data centres. Its base-case projection nevertheless assigns gas and coal together more than 40% of additional demand through 2030. It also describes substantial regional differences in generation. These figures belong to the 2025 analysis and its scenarios. They should not be presented as the current fuel mix of an unnamed facility, or used to attribute a specific quantity of emissions to a local project without project-level evidence.
What the agency's newer assessment adds
The IEA's 2026 executive summary updates the demand picture. It reports that overall data-centre electricity consumption increased by 17% during 2025, while consumption at AI-focused facilities rose by 50%. Its central outlook puts total data-centre electricity demand at approximately 950 terawatt-hours in 2030, compared with 485 terawatt-hours in 2025. The document identifies those future quantities as projections, rather than observed consumption.
The updated assessment also describes a relationship between efficiency and demand that cannot be reduced to one trend. Energy required for an individual AI task has fallen rapidly, it says, while adoption and more energy-intensive applications have expanded. It identifies constraints involving chips, electricity equipment, capital and grid connections. The executive summary also discusses community opposition and concerns about affordability and environmental effects. These observations provide a broader context for public debate; they do not independently verify the allegations, speakers or outcomes of the round-table event named in the RSS headline.
Affordability, flexibility and the response from industry
In its accompanying announcement, the IEA explains that rapidly expanding, concentrated electricity loads can require additional generating capacity and network investment. It also reports that higher demand does not automatically mean higher prices when appropriate infrastructure investment and policies are in place. Its account therefore presents affordability as a question involving how facilities are integrated into power systems, rather than an inevitable single outcome.
The announcement describes technology companies signing renewable-power agreements and supporting nuclear and geothermal developments. It also discusses onsite gas generation and batteries as responses to grid constraints and changing loads. Executive Director Fatih Birol states that "there is no AI without energy", while calling for cooperation between public authorities and the technology and energy industries. Those are the agency's documented positions. They offer material for a substantive explanation of the energy debate while leaving the local meeting's proceedings, project-specific effects and participants' responses to be established through further reporting.