The European Union aims to be carbon neutral by 2050, combining sustainability and economic competitiveness. In this race against time, theartificial intelligence (Ia) could become a strategic ally: according to the document “The AI OPPORTITY FOR EUROPE’S CLIMATE GOALS – A Policy Roadmap“, Published on April 8, 2025, IA technologies could contribute to Reduce between 5 and 10% of global greenhouse gas emissions By 2030, the equivalent of the entire annual EU production. But this same technology, if not managed carefully, risks becoming an environmental boomerang.
To believe it, among others, is Googlethe protagonist of a transition that promises sustainable innovation. The company, which declares itself Carbon Neutral since 2007 and which has compensated 100% of its energy consumption since 2017 with renewable sourcesaims to fully operate with zero emissions energy by 2030. Yet, between promises and reality, growing criticalities emerge.
In only 2023, the emissions of Google data centers increased by 48% compared to 2019. The culprit? THE’Generative artificial intelligencelike Geminiwhich feeds numerous Google products: from Earth to Gmailfrom DOCS to Research. These models require enormous calculation power, with consequent intensive workloads, both from an energy and water point of view.
In 2021, Google data centers used 5.6 billion water gallons for cooling systems. This figure is likely to be even higher. In regions such as The Daes (Oregon), where Google has a large data center, the local communities have reported the pressure on water resourcesespecially during periods of drought.
The paradox is evident: the IA can help solve the climatic crisis, but in its current form it risks aggravating it. A concept also reiterated by the researcher Fieke Jansen project Date of the University of Amsterdam: «Google’s ambitions regarding artificial intelligence are incompatible with its climatic objectives. Emissions cannot be increased and claiming to be sustainable ».
The problem is not only of consumption. It is also – and perhaps above all – of transparency. Google’s environmental reports, as well as those of other tech giants, do not provide disaggregated data on the energy and water consumption of IA -based applications. This makes it almost impossible to evaluate the real impact of Gemini and similar technologies.
In the meantime, the European Commission suggests a strategy on three fronts to face this challenge: to enable, implement and guide. The idea is to create the infrastructures necessary for sustainable Irs (access to data, development of skills, technological investment), favor its adoption in the key (energy, industry, transport, agriculture) sectors and regulate the entire ecosystem to mitigate their risks.
The potential is immense: the effective adoption of the AI could reduce the energy consumption of European industrial buildings up to 40% and generate an economic increase of 1,200 billion euros in ten years. But the direction is everything.
Google, for its part, has started some initiatives with Deepmind To reduce the environmental football of models, such as the optimization of architecture and the planning of workloads. But these are still small steps compared to the rapid expansion of artificial intelligence.
The context is wider: also Half, Microsoft And Amazon They are integrating generative in their platforms, multiplying the energy demand. The global race to artificial intelligence is creating an increasingly evident gap between digital innovation and environmental sustainability.
To reduce this waste, you need to act now. Serve transparency in data, regulatory obligations on specific environmental reporting for the IA And Investments in zero emissions infrastructure.
As Jansen points out, “technological companies cannot count on compensation objectives or on promises for 2030. They must immediately demonstrate that their models are truly sustainable”. IA can be a powerful tool in the fight against climate change. But without responsibility, he risks becoming yet another solution that feeds the problem.
