The Spanish government’s “Algoritmos Verdes” initiative explains that artificial intelligence models have a measurable carbon footprint that depends on the energy consumed in data centres and the carbon intensity of the electricity mix. It presents tools such as CarbonTracker, CodeCarbon and Green Algorithms, which allow developers to estimate and monitor the greenhouse gas emissions associated with training and running AI models. The article frames this measurement as essential to reduce the environmental impact and move AI towards more sustainable practices.

Artificial intelligence has a significant environmental impact, arising from the consumption of energy, water, materials, and emissions. The life cycle of AI generates emissions at multiple stages, including equipment manufacturing, electricity used to train and operate models, and the energy needed to keep data centers running.

The report states that electricity consumption from the use of information and communication technologies is between 5% and 9% of the global total, with a projection of reaching 20% in 2030. It estimates that greenhouse gas emissions produced by the use of these technologies were between 1.1 and 1.3 gigatons of CO₂ in 2020. It notes that so‑called ‘Red AI’ or high‑energy‑consuming artificial intelligence has a significant environmental impact and that the computing resources required to train AI models have been doubling every 3.4 months since 2012. At the same time, the report highlights that AI can help reduce energy and resource consumption, promote decarbonization and boost the circular economy.

UNESCO describes a ‘fundamental role’ for AI in addressing environmental sustainability while warning about the challenges posed by its energy consumption. It calls for a conscious approach to AI deployment, evaluating its direct and indirect environmental impacts, including energy use, consumption of raw materials, and water used to cool data centers. The article stresses the importance of developing smaller, domain‑specific models, improving efficiency, and deploying these models in data centers powered by renewable energy to reduce the carbon footprint.

The article explains that artificial intelligence has a significant environmental impact across its entire life cycle: manufacturing of hardware, data center operation, and end‑of‑life of equipment. It states that mitigating AI’s life‑cycle emissions "requires a fundamental change in the way we approach its development and deployment," highlighting the need to optimize models, use renewable energy, and improve hardware efficiency to reduce its carbon footprint.

This analysis by the Spanish Ministry of Defence’s CESEDEN cites a PitchBook report estimating that by 2025, about 3.2% of all global carbon emissions will come from AI server farms and their associated infrastructure. It notes that the rapid expansion of AI accelerates electricity consumption and hardware demand, which in turn increases emissions unless energy systems are decarbonised. The text stresses that the environmental footprint of AI is significant but can be mitigated by more efficient algorithms, better cooling systems and the use of renewable energy in data centres.

UNIA explains that artificial intelligence generates environmental impact because it relies on physical infrastructures that consume energy, water and materials. It states that training an advanced AI model can use more than 1,200 MWh of electricity, and that a single query to a generative system uses up to ten times more energy than a traditional web search. The article reports that training GPT‑3 emitted about 552 tonnes of CO₂ and that a commercial chatbot can emit around 1.5 grams of CO₂ per query, warning that data‑centre emissions linked to AI could triple by 2030 if electricity systems are not rapidly decarbonised.

Data centers consume energy and water at a growing rate with every click. Each new AI model implies an increase in power demand, and therefore a greater consumption of energy and water. The text also says the environmental footprint of the digital sector is approaching that of the aviation industry.

The article notes that the AI industry has opted for development with a high economic and ecological cost and states that, despite promising applications, ‘AI is far from being an ecological technology.’ It explains that behind AI there is a physical infrastructure of huge data centers and countless components such as microchips and GPUs that depend on long and complex supply chains whose environmental impact is very difficult to audit. Citing estimates from the U.S. Department of Energy, it reports that the electricity consumption of data centers has tripled in the last decade and is expected to triple again by 2028, reaching 325 TWh per year, a demand greater than that of countries such as Spain, Italy or the United Kingdom. It also highlights increased carbon footprints for major tech firms and estimates that by 2027 AI could consume between 4.2 and 6.6 billion cubic meters of water annually, equivalent to four to six times Denmark’s consumption or half that of the United Kingdom.

This academic article in Spanish sets out to analyse the environmental impact associated with intensive use of generative artificial intelligence. The authors describe how large‑scale language and image models require substantial computational resources, which translates into high electricity consumption and associated greenhouse gas emissions during both training and inference. The study discusses the often ‘hidden’ nature of these impacts for end users and calls for greater transparency and measurement of AI’s life‑cycle emissions to inform sustainability assessments.

The report defines "green AI algorithms" as being designed "to minimize the environmental impact of computing, in particular by reducing the energy consumption and carbon emissions associated with AI models and their training processes." It notes that these algorithms aim for efficient computation through strategies such as algorithmic efficiency optimization, model compression, and the use of less computationally intensive models, "without compromising performance." This framing explicitly treats AI sustainability as a design and engineering challenge rather than an inherent impossibility.

The Observatorio of Tecnológico de Monterrey reports that the creation of the GPT‑3 model emitted 552 tonnes of CO₂ and produces around 8.4 tonnes of CO₂ annually, illustrating the substantial carbon footprint of large AI systems. It notes that Google’s Environmental Report 2024 states the company’s greenhouse‑gas emissions rose 13% in 2023, largely due to increased electricity consumption by its data centres and emissions from its supply chain. At the same time, the article cites studies suggesting that AI could help mitigate 5–10% of global greenhouse‑gas emissions by 2030, showing both environmental costs and potential benefits.

The paper states that "AI itself is not only a beast in terms of energy demand; the scale of its growth has the potential to quickly surpass the energy consumption of many current digital technologies." It argues that, without measures, the environmental impact of AI could grow rapidly, but then analyses methods for designing energetically efficient algorithms to reduce energy use in training and inference.

The blog explains that training an advanced model such as ‘GPT‑5’ requires thousands of graphics processing units (GPUs) running in parallel for weeks, translating into gigawatts of energy consumption and the use of millions of liters of water to cool the data centers that support these workloads. It argues that, although training large models requires a considerable energy investment, AI should be seen as an enabling technology whose capacity to improve efficiency in many sectors can generate energy and emissions savings that are much greater than what AI itself consumes. The article also highlights mitigation strategies, including energy optimization via AI, algorithmic efficiency improvements, low‑power hardware such as TPUs, innovative cooling (like liquid or underwater data centers), and greater use of renewable energy by cloud providers.

The article notes that training modern AI models "requires large amounts of data and computing power, which translates into high energy consumption and CO₂ emissions." It then explains the concept of **Green AI**, distinguishing "Green-by AI" (models designed to help fight climate change and improve energy efficiency) from "Green-in AI" (strategies to make AI models themselves more efficient and sustainable in training and execution). It lists strategies such as algorithm optimization (pruning, distillation, low‑precision computation), more efficient hardware (TPUs and specialized accelerators), and data‑center optimization with renewable energy, all aimed at reducing AI’s carbon footprint.

Telefónica’s BlogThinkBig explains that a startup, Hugging Face, has developed an AI system specifically to calculate the carbon footprint of AI models more accurately across their whole life cycle. The article notes that training and running AI requires large amounts of energy, and that the total carbon footprint is estimated using direct measurements of electricity consumed during training and inference, combined with factors reflecting the emissions intensity of the power grid. It also mentions including the environmental impact of electronic components and final disposal of hardware, underlining that AI’s emissions can be systematically quantified.

The piece highlights that training deep‑learning models requires massive amounts of energy and cites a 2019 study finding that training a large natural language processing model can emit up to 284 tons of CO₂, "equivalent to the lifetime emissions of five average cars in the United States." At the same time, it describes AI as having a "transformative impact on energy efficiency," noting that AI‑based building energy management systems can achieve 20–40% energy savings and that AI‑enabled traffic management in cities like Singapore and Stockholm has reduced CO₂ emissions by an estimated 15–25%.

BBVA explains that AI can help mitigate climate change by optimizing the use of energy, monitoring emissions, and improving resource efficiency in key industries. It notes that AI‑based solutions improve energy efficiency, optimize use of the power grid, and reduce associated emissions. The article acknowledges that the adoption of AI may increase energy consumption but stresses that "its potential to reduce CO₂ emissions is considerable" when applied to sectors such as energy, consumer goods, and automotive.

This academic article (in Spanish, abstract available via Dialnet) analyzes the relationship between artificial intelligence and environmental sustainability, identifying both opportunities and risks. The authors note that the training and operation of AI systems entail high energy consumption and a growing demand for data center infrastructure, contributing to greenhouse gas emissions. At the same time, they describe use cases in which AI contributes to sustainable development, such as optimizing energy systems, improving resource efficiency and monitoring environmental change, and call for regulatory frameworks and technical standards to ensure that AI development is aligned with sustainability goals.

This article describes how AI can improve the energy efficiency of data centers, which are a major source of digital‑sector emissions. It explains that AI‑driven systems can dynamically manage cooling, workloads, and power usage to reduce electricity consumption and carbon footprint while maintaining service levels. The piece frames AI not only as a load on data centers but also as a tool to achieve "more sustainable data‑center operations" through continuous optimization.

Every search on the internet, every click, and every interaction with AI-based tools generates a digital footprint stored in physical databases and processors that consume energy and water, generate waste, and emit greenhouse gases. The document states that all actions involving technology, including the manufacture, installation, training, maintenance, and use of AI, have an environmental impact.

This article focuses on using artificial intelligence as a tool to reduce carbon footprints in industry and services. It explains that AI can analyse large volumes of data to optimise processes, improve energy efficiency and reduce waste, which in turn can lower greenhouse‑gas emissions. The text highlights that AI systems can accurately measure the emissions of processes, products and supply chains, identifying critical points where improvements can significantly cut the carbon footprint.

AI can also help analyze large ecological datasets, allowing scientists to identify patterns, but the article is relevant here because it situates AI within sustainability and environmental impact discussions rather than presenting it as impact-free. The search result excerpt indicates the paper addresses AI, sustainability, and environmental impact as linked topics.

In this Spanish‑language special report, Green Continuity explains that sustainable AI must meet environmental, social and economic criteria. The video states that the energy consumption of AI development and operation represents a direct challenge to global decarbonization commitments and cites an ‘AI sustainable for a greener future’ report estimating that by 2028 AI workloads could represent more than half of total data center energy consumption, equivalent to 22% of the electricity consumption of all households in the European Union. It identifies four key indicators to evaluate AI’s ecological footprint: energy demand, global warming potential, water consumption and depletion of abiotic resources, and stresses that it is possible to achieve sustainable AI if an integrated approach combining energy efficiency, ethics, responsible governance and alignment with the Sustainable Development Goals is adopted.

Gmp reports that artificial intelligence represents "a decisive step towards a more efficient and responsible energy model" in buildings. It describes how real‑time data analysis, predictive algorithms, and intelligent automation allow optimization of resources and reduction of energy consumption and CO₂ emissions without sacrificing comfort. As a concrete example, in its Oxxeo building, applying AI tools has reduced air‑conditioning energy consumption by more than 25% and avoided the emission of 49 tons of CO₂ per year.

The article explains that AI applied to energy management in companies can monitor electricity, heating, and lighting consumption in real time and automate the operation of systems such as air conditioning and electrical equipment. It states that this leads to measurable economic and environmental benefits: direct savings on electricity bills and "reduction of CO₂ emissions" by improving corporate environmental footprints. Citing the European Commission, it notes that companies adopting intelligent energy management can cut their annual consumption by 10–20%, depending on their activity.

Background context: authoritative energy and climate assessments have repeatedly stated that data centers and information and communication technologies contribute a non-trivial share of global electricity use and emissions, with estimates varying by methodology and year. This is relevant because AI is built on data centers and compute infrastructure.

In this Spanish‑language video report, the presenters explain that training and running AI models leads to a sharp increase in energy consumption for technology companies. They state that training AI tools in high‑power data centres can take months and consumes large amounts of electricity, much of which is still generated from fossil fuels. The report suggests that the resulting carbon footprint of AI is substantial and often under‑reported compared with other digital activities.

The video states that generative AI has a very high environmental cost because the data centers that power it consume large amounts of energy and water. It says training a large language model can generate between 200 and 300 tons of CO2, and that servers also produce large amounts of electronic waste.