The servers never sleep. Right now, somewhere in the world, a data centre is being built to power the next wave of artificial intelligence. Then another. Then thousands more. Behind the weightlessness of the digital world lies an expanding physical architecture of chips, cables, cooling systems, and power lines. In 2024, data centres consumed about 1.5% of total electricity use. In 2025, that demand grew by 17%, far faster than overall global electricity demand.
AI is becoming part of the operating system of modern growth, embedded in how goods are made, how services are delivered, and how nations build competitiveness. In India, this shift is already visible in steel-and-concrete terms. The country’s data-centre market has reached about 1.6 GW of operational capacity, with another 3.1 GW under construction or planned.
And yet, in many sustainability conversations, AI still barely appears.
What We Measure, and What We Are Beginning to See
Corporate sustainability accounting was built for a physical world. Scope 1 covers what we burn on-site. Scope 2 covers the electricity we buy. Scope 3 captures the value chain; the emissions embedded in the goods and services that flow in and out of our operations. It is a rigorous framework that has served us well for a world of fuel, materials, and supply chains.
However, AI is a different input that thinks, predicts, and generates. But it is not weightless as every query draws on physical infrastructure of servers, cooling systems, electricity grids. The International Energy Agency estimates that water consumption for cooling could reach 1.2 trillion litres annually by 2030.
A question that is only now beginning to be asked is: when a company uses AI to run its operations, how should its environmental cost be understood and accounted for?
A Framework with an Emerging Gap
Technically, the answer already exists. Under current Scope 3 rules, AI consumed as a purchased service sits within Category 1: Purchased Goods and Services. The framework has a place for this footprint.
But in practice, it is not yet being systematically reflected there.
The reason is structural. The data required to do this consistently is still emerging. As of early 2026, Google was one of the few that published per-query environmental information for its AI models. Such comparable disclosures are still limited. As a result, companies adopting AI at scale have limited ability to account for its environmental footprint in a standardised way.
At the same time, the GHG Protocol, the global standard used by over 92% of Fortune 500 companies, is undergoing its first major revision in 15 years. That revision does not yet explicitly address AI as a distinct consumption category.
The result is what might be described as a “ghost room” in the framework—an emerging gap in how we currently measure. Companies are procuring AI as an IT or productivity input. It sits outside traditional energy accounting boundaries and is not consistently visible in sustainability disclosures.
When a Tool Becomes Infrastructure
Until recently, AI was seen as a productivity tool, a competitive advantage for early adopters. That distinction is starting to shift.
In India, AI is increasingly embedded in development priorities such as manufacturing, agriculture, governance, financial inclusion. Similar patterns are emerging globally. AI is moving closer to infrastructure.
Infrastructure has always carried an environmental cost. We account for the carbon in steel used to build a factory. We track the electricity consumed by production systems. The principle underpinning sustainability reporting, that dependencies carry shared responsibility, has evolved over time to reflect changing economic realities.
As digital systems become more central to operations, that principle may need to extend further into how we understand computational inputs.
A Conversation Whose Time Has Come
The productivity gains are real, as is the development opportunity. AI will also play a critical role in advancing climate solutions, from grid optimisation to agricultural efficiency to accelerating the clean energy transition.
But as with any foundational technology, its integration raises new questions for sustainability frameworks.
The corporate sustainability community has an opportunity here: to help evolve how measurement keeps pace with innovation. Scope 3 itself emerged from recognising that value chains extend beyond direct control. A similar shift may now be underway in how we think about digital infrastructure and its environmental footprint.
The opportunity for sustainability leaders today is to help shape that evolution, alongside regulators, technology providers, and investors, so that the frameworks we rely on continue to reflect the realities they are meant to measure.
