The AI Boom Is Challenging Net-Zero Goals: Can Technology And Sustainability Grow Together?

The AI boom has an inconvenient physical reality. Every model trained, every chatbot query processed and every AI-enabled workflow deployed ultimately depends on data centres ( buildings filled with servers, cooling equipment and power infrastructure). The technology may feel weightless to the user; its environmental footprint is not.

The scale is beginning to show in the numbers. The International Energy Agency (IEA) reported that global electricity demand from data centres rose by 17% in 2025, while electricity consumption at AI-focused data centres grew by around 50%. The IEA expects overall data-centre electricity consumption to roughly double by 2030, with AI the most important driver of that growth.

For companies pursuing net-zero targets, that creates a tension that is becoming difficult to ignore. Microsoft offered a striking example in July 2026: it reported that its total Scope 1, 2 and 3 emissions increased 25% year on year, driven primarily by the expansion of its data-centre infrastructure. This is a company that remains committed to becoming carbon negative by 2030.

The lesson is not that AI and sustainability are incompatible. It is that AI has become a sustainability management issue in its own right.

AI is digital. Its infrastructure is not.

AI's footprint extends beyond the electricity used to run a model. Data centres require continuous power, sophisticated cooling and significant physical infrastructure. Advanced chips bring their own upstream footprint through semiconductor manufacturing and demand for critical minerals. Depending on location and cooling technology, water can also become a material consideration.

There is an important nuance here. AI systems are becoming more efficient. The IEA notes that electricity consumption per AI task is declining rapidly. But efficiency per task does not automatically mean lower total consumption. When the number and complexity of AI tasks grow even faster, aggregate electricity demand can still rise sharply.

That distinction matters for business leaders. The useful question is not simply, 'Does AI consume energy?' Of course it does. The better question is: do we understand the incremental footprint well enough to manage it?

AI can also be one of sustainability's most useful tools

The other half of the story matters just as much. Used well, AI can help companies detect energy losses, optimise industrial processes, improve logistics, anticipate equipment failures, analyse climate risks and process large volumes of environmental data far faster than conventional methods.

For a manufacturer, predictive maintenance can reduce downtime and avoid wasted material. For a logistics business, better routing can reduce fuel consumption. For an ESG team, AI can help identify data anomalies and make large, fragmented datasets easier to interrogate. In energy systems, AI can support forecasting, grid optimisation and the integration of variable renewable power.

But there is a trap in automatically labelling every efficiency gain as a sustainability gain. If an AI application saves energy in one part of the business while its computing requirements create a larger footprint elsewhere, the net environmental benefit may be smaller than it first appears. The benefit needs to be measured, not assumed.

This is where the conversation should move beyond 'AI for sustainability' to 'sustainable AI': using the technology where it creates genuine value while understanding the resources required to deliver that value.

The ESG accounting blind spot

For sustainability teams, AI creates an awkward accounting question: where does its footprint actually sit?

If a company owns and operates its computing infrastructure, the impact may be visible more directly through purchased electricity and the emissions associated with that infrastructure. If the same computing is purchased through cloud, software or AI service providers, much of the attributable impact can sit upstream in the value chain instead. The precise Scope 2 and Scope 3 treatment depends on the delivery model, contractual arrangements and the organisation's GHG inventory boundary.

In practical terms, an organisation can significantly increase its use of AI without seeing an obvious jump in the electricity meter at its own offices or factories. That does not mean the environmental impact has disappeared; it may simply have moved into a supplier's data centre.

This makes supplier information increasingly important. If a business cannot obtain credible energy and emissions data from its cloud and AI providers, estimating the incremental footprint of digital growth becomes difficult. For companies with science-based targets, net-zero commitments or demanding global customers, that is not a trivial data gap.

Why Indian businesses should pay attention now

Indian businesses do not need to wait for an 'AI emissions regulation' before addressing this issue. Many listed companies already disclose entity-level energy consumption and Scope 1 and Scope 2 greenhouse-gas emissions under SEBI's BRSR framework. At the same time, exporters and multinational suppliers increasingly operate within customer-driven carbon and ESG requirements that extend beyond domestic disclosure rules.

AI will not normally appear as a neat standalone line item in an ESG report. That is precisely why its footprint can be easy to miss. As organisations scale cloud computing, generative AI and automated analytics, sustainability teams will need a clearer view of which impacts are occurring inside the organisational boundary, and which are being transferred into the value chain.

Regulation is also beginning to recognise the issue. In the European Union, providers of general-purpose AI models are required under the AI Act framework to include known or estimated energy consumption in their technical documentation. In 2026, the European Commission also consulted on approaches for measuring the energy consumption and emissions of AI models and systems, including work that could support a future AI energy and emissions label.

That does not create an equivalent disclosure obligation for every Indian company using AI. But it is a useful signal of direction: energy transparency is starting to become part of the governance conversation around AI itself.

Five questions boards and sustainability leaders should ask

The immediate requirement is not another elaborate framework. It is better visibility. Before AI adoption moves from experimentation to enterprise-scale infrastructure, management teams should be able to answer five basic questions:

1. Where does our AI actually run - on infrastructure we control, in the cloud, or through third-party applications?
2. Can our major cloud and AI providers give us credible information on energy use, renewable electricity, water and associated emissions?
3. Have the implications of increased AI and digital infrastructure been reflected in our emissions forecasts and net-zero pathway?
4. Are we evaluating AI workloads for efficiency as well as performance and cost?
5. Does our AI governance process consider environmental impact alongside privacy, cybersecurity, ethics and business risk?

These questions will not produce perfect numbers overnight. They will, however, expose whether a company's sustainability strategy is keeping pace with its technology strategy.

The debate is often framed as a choice between AI and sustainability. That is the wrong framing. AI is neither sustainable nor unsustainable in the abstract; its impact depends on how it is designed, powered, procured and used.

The organisations that manage this well will not be the ones that slow innovation. They will be the ones that make the environmental cost of innovation visible early enough to manage it. As AI becomes part of everyday business infrastructure, that visibility will be essential to keeping digital ambition and net-zero ambition on the same path.

The views expressed in this article are solely those of the author in her personal capacity and are based on her understanding of the applicable legal and regulatory framework. They do not necessarily represent the views, policies or practices of organisation with which the authors is associated.