AI and carbon: the business case that cuts both ways

The companies racing to deploy artificial intelligence are, in many cases, also the ones with the most ambitious net zero commitments. That is not a contradiction. It is, if managed carefully, the starting point for a credible business case.

The International Energy Agency's 2025 report 'Energy and AI' estimates that the widespread adoption of existing AI applications across industry, buildings, transport, and the power sector could reduce global energy-related emissions by around 5% by 2035.

These are not speculative applications awaiting commercialisation. Grid management, demand forecasting, real-time optimisation of renewable generation assets, and industrial process control are in operational deployment today. The caveat, which the IEA is clear about, is that realising this potential depends on overcoming real barriers: data access, digital infrastructure, regulatory constraints, and the risk that efficiency gains are offset by new sources of energy demand. The business case must be built on those conditions, not around them.

For industrial and commercial organisations, the strongest near-term case concentrates in three areas, each with a distinct financial logic. Energy efficiency in buildings and operations is the most immediate. AI-driven building management systems reduce energy consumption by continuously adjusting heating, cooling, lighting, and ventilation based on occupancy patterns, weather data, and tariff signals. Where energy costs are significant, the financial return is direct and measurable, and the emissions reduction follows from the cost saving rather than requiring separate justification.

Supply chain visibility compounds that case at the strategic level. Understanding the emissions embedded in purchased goods and services, the challenge at the heart of Scope 3 reporting under the GHG Protocol (which covers indirect emissions across an organisation's value chain), requires processing large volumes of supplier data, logistics records, and product lifecycle inputs. AI-assisted analysis makes this tractable at scale. This matters in particular for organisations subject to mandatory sustainability disclosure under the Corporate Sustainability Reporting Directive (CSRD), Directive (EU) 2022/2464.

The European Commission's Omnibus simplification package, published in 2025, proposes to limit CSRD reporting obligations to companies with more than 1,000 employees, but the direction of regulatory travel is not in doubt and many Irish organisations will face supply chain disclosure requirements either directly or through the value chains of larger customers. The commercial case for AI-enabled emissions mapping is no longer purely environmental.

Predictive maintenance completes the picture in energy-intensive industries precisely because it makes the financial argument independently of any sustainability framing. Reducing unplanned downtime and keeping assets operating within designed parameters cuts energy waste and extends equipment life. In manufacturing, logistics, and infrastructure, the combined financial and sustainability case for this class of AI application is well-established and does not require a green premium to justify.

The organisations we work with increasingly recognise that these applications align financial and sustainability incentives rather than trading one against the other. That alignment matters considerably. Sustainability investments justified solely on environmental grounds face a different internal approval process, and a higher risk of deferral, than those that also improve margins, reduce energy costs, or strengthen supply chain resilience.

There is, however, a counterpoint that any credible business case must address. AI infrastructure is itself energy-intensive. Training large models consumes substantial electricity. Inference at scale adds steadily to that demand. The IEA projects that electricity consumption from data centres will more than double by 2030 to around 945 terawatt-hours, with AI as the primary driver. For organisations whose AI strategy runs ahead of their energy strategy, the net emissions picture can deteriorate even as efficiency gains accumulate elsewhere.

This is not an argument against AI adoption. It is an argument for clarity about where the emissions sit, how they are measured, and whether the energy powering AI workloads is clean. Organisations that own and operate their own compute infrastructure carry the associated electricity consumption as Scope 2 emissions under the GHG Protocol. Those purchasing compute as a cloud service, which is the more common arrangement, carry it as Scope 3, Category 1 (purchased goods and services). Either way, these emissions are material GHG accounting items that auditors, regulators, and informed stakeholders are beginning to examine in sustainability disclosures.

For boards and executives, the governance implication is straightforward. AI-enabled emissions reduction is a credible lever, and the financial case for many applications is strong enough to stand independently of any sustainability argument. But realising that potential requires treating AI deployment and energy strategy as connected decisions, not parallel workstreams. The companies that will extract the most value from AI's role in decarbonisation are not those that deploy the most sophisticated models. They are those that apply the same analytical discipline to their own energy and emissions footprint as they expect AI to apply everywhere else.