Context:

• While Artificial Intelligence is widely discussed for its applications across sectors, its environmental impact has received limited attention.
• An Organisation for Economic Co-operation and Development working paper highlights that developing and deploying AI algorithms entails environmental costs, notably an increased carbon footprint that worsens climate change challenges.
• The report estimates that the global ICT industry contributes about 1.8%–2.8% of global greenhouse gas emissions, with some estimates placing this figure even higher.
• At the same time, reliable data on AI’s carbon footprint remains contested.
• A 2025 report by Google claims that a single AI text prompt consumes minimal electricity, but this has been criticised for offering incomplete and potentially misleading conclusions about AI’s true environmental impact.
• This article highlights the growing but under-recognised environmental impact of Artificial Intelligence, examining its carbon, energy, and water costs, global regulatory responses, and the urgent need for India to integrate sustainability into AI governance.

Environmental Impact of AI Across Its Life Cycle

• An issue note by the United Nations Environment Programme (UNEP) warns that the full AI life cycle could significantly strain natural resources.
• It estimates that AI servers may consume 4.2–6.6 billion cubic metres of water by 2027, intensifying water scarcity.
• Studies cited by UNEP indicate that training a single large language model can generate nearly 300,000 kg of carbon emissions.
• Similarly, earlier research found that training one large AI model can emit over 626,000 pounds of carbon dioxide—comparable to the lifetime emissions of five cars.
• Further, AI usage also raises energy demands: a UNEP study notes that a single ChatGPT query consumes about ten times more energy than a standard Google search, underscoring AI’s growing contribution to climate change.

Global Efforts to Address AI’s Environmental Costs

• In 2021, UNESCO released its Recommendation on the Ethics of Artificial Intelligence, urging recognition of AI’s negative impacts on society and the environment.
• Though non-binding, it was adopted by around 190 countries.
• Among major jurisdictions, the United States and the European Union have taken the lead by proposing legislation specifically addressing AI’s environmental footprint, including the Artificial Intelligence Environmental Impacts Act, 2024, and the EU’s harmonised AI rules.

The Indian Gap in AI–Environment Discourse

• While global debates increasingly focus on AI’s carbon costs, discussions in India largely emphasise how AI can help address climate change, overlooking the environmental downsides of developing large AI models.
• There is a growing need for India to formally recognise and address these hidden environmental costs.
• Measuring Environmental Impact of AI
  • A crucial first step is to systematically measure the environmental impact of AI development and deployment.
  • In India, the Environmental Impact Assessment Notification, 2006 mandates EIAs for major infrastructure and development projects.
  • Its scope could be expanded to include AI systems and algorithms, given their increasing resource intensity.
• Standards, Stakeholders, and Data Collection
  • The government could also establish common standards to assess AI’s environmental impact by involving technology companies, think tanks, and environmental NGOs.
  • This would help build consensus on definitions, indicators, and reporting requirements.
  • Alongside this, systematic data collection using sustainability metrics—such as greenhouse gas emissions, energy use, water consumption, and land and resource impacts—would enable evidence-based and environmentally informed AI policy-making.

Integrating AI into ESG Disclosure and Sustainability Frameworks

• The government can consider including the environmental impact of developing and deploying AI models within environmental, social and governance (ESG) disclosure standards overseen by the Ministry of Corporate Affairs and the Securities and Exchange Board of India.
• India could draw lessons from the European Union, where the Corporate Sustainability Reporting Directive mandates disclosure of emissions from data centres and high-compute activities, including the training of large language models.
• At the same time, the emphasis should move toward positioning AI as part of the solution for global sustainability goals.
• This includes adopting sustainable AI practices such as using pre-trained models to reduce compute intensity, powering data centres with renewable energy, and reporting AI-specific environmental impact estimates to minimise AI’s ecological footprint.