Artificial Intelligence (AI) is no longer a lab curiosity — it has become part of our everyday lives. Algorithms now recommend our next series, optimise delivery routes, and respond to complex questions in seconds.
This leap is the result of three converging forces: the explosion of data, falling prices of specialised chips, and the network effect of platforms like ChatGPT, which grew from zero to hundreds of millions of users in just a few months. But this computing power comes at a significant energy cost.
According to MIT Technology Review, data centres in the United States alone consumed around 200 TWh in 2024 — enough electricity to power Thailand for a year. The International Energy Agency (IEA) estimates that global electricity demand from data centres could more than double to 975 TWh by 2030, driven mainly by the growth of AI-related tasks.
The Scientific Balance Sheet
This energy picture would be alarming if there weren’t also clear signals that AI could deliver a net positive climate impact. A recent article in npj Climate Action concludes that AI applications in sectors such as energy, food, and mobility could avoid between 3.2 and 5.4 gigatonnes of CO₂e annually by 2035, thanks to five main levers — from smart grid optimisation to accelerated discovery of clean materials.
At a macro level, a study covering 67 countries (1993–2019) shows that each additional percentage point in AI adoption increases the share of renewables by 0.14% and reduces ecological footprint, especially in more open, trade-driven economies.
Algorithms can help — but the real impact depends on context: regulation, energy mix, and data quality.
Four Areas of Measurable Impact
Carbon tracking and emissions forecasting
Machine learning models can now process emissions data — including complex Scope 3 — in minutes instead of weeks. This capability will be even more critical as the CSRD (Corporate Sustainability Reporting Directive) requires European companies to publish detailed sustainability reports. These obligations are under review, but they remain central to EU climate transparency.
Circular logistics chains
Predictive platforms anticipate returns, optimise transport routes, and scale recycling networks, turning waste into resources. According to the Ellen MacArthur Foundation, if AI, circular design, and new business models are deployed together, the annual gains could exceed USD 200 billion in food and electronics sectors alone.
Predictive maintenance of green infrastructure
AI sensors and models already predict over 90% of failures in wind farms, halving downtime and extending turbine life — delivering more megawatts without needing to install a single new blade.
Automated ESG reporting
With the arrival of EU regulations such as the ESRS, AI is being used to accelerate ESG data extraction, validation and organisation. Smart tools help structure complex data and reduce operational effort — freeing up teams to focus on real sustainability action.
The Other Side of the Coin
All these gains can be undone if the electricity powering these models still comes from fossil sources or if the hardware remains inefficient. As the IEA warns: without clean energy, emissions from data centres will rise in line with demand.
Beyond the energy footprint, there’s also the risk of algorithmic greenwashing: models trained on incomplete data can amplify errors and biases. The EU’s AI Act, in force since 2024 and applicable from 2025, aims to ensure that AI systems placed on the EU market are “safe, transparent and energy-responsible”.
From Theory to Action — What Can We Do?
To ensure AI becomes part of the solution — not the problem — society and businesses must adopt three complementary attitudes.
First, choose digital services more wisely. Just as we switch to green electricity at home, we can favour platforms that publish carbon audits and run data centres powered by renewables. This market pressure encourages suppliers to adopt clean energy contracts and invest in more efficient hardware.
Second, use AI with purpose. Every extra query or image generation consumes energy. Writing clearer prompts, avoiding trivial requests, and reusing results reduces individual footprints without sacrificing productivity. For companies, this means evaluating the “energy used vs. emissions avoided” before scaling any AI pilot.
Third, extend equipment lifespan and invest in people. Updating software, replacing batteries, or expanding memory prevents the need for extracting rare metals used in new devices. In parallel, boosting literacy in data and ethics — from bias awareness to reading energy labels — equips teams to configure, monitor, and, when necessary, halt unsustainable systems.
Strategic Intelligence for a Sustainable Future
The climate, energy, and sustainability transition relies on three pillars: decarbonising energy production, electrifying (or switching to clean molecules) high-emissions processes, and closing material loops. AI can become the nervous system of this transformation — forecasting wind and solar variations to stabilise 100% renewable grids, redesigning value chains to turn waste into resources, and freeing up human time for deeper climate action.
But this promise will only materialise if digital infrastructure moves to green sources, data is trustworthy, and ethical governance is non-negotiable.
The defining intelligence of the coming decade, therefore, lies less in algorithms — and more in our choices as citizens, companies, and governments to channel this computing power toward a regenerative, resilient, and just future.
Excerpt written by Beatriz Santos
