India must focus on AI and its environmental impact

UPSC Prelims + Mains Study Note | GS-III | Science & Technology / Environment

1. At a Glance

• AI's environmental footprint — energy, water, carbon, and land — is an emerging governance challenge that intersects climate action, technology policy, and sustainable development. [1][2]
• The global ICT sector (including AI) accounts for 1.8%–3.9% of global GHG emissions, comparable to the aviation industry. [4]
• India faces a 35.6× higher carbon emission multiplier per computational task than Norway due to its coal-heavy grid — making green AI a strategic imperative. [2]
• The IndiaAI Mission (March 2024) is India's primary policy vehicle for AI; its sustainability angle is critical for Mains essay and GS-III answers. [5]

2. Why in the News

• January 14, 2026 — The Hindu published an article by former Rajya Sabha MP and CAG bureaucrat Amar Patnaik arguing that India must urgently address AI's environmental impact alongside its economic promise. [4]
• June 2026 — UN News reported AI's environmental costs threatening water, land, and climate at scale. [3]
• September 2024 — UNEP issue note highlighted that AI data centres may consume 4.2–6.6 billion cubic metres (bcm) of water by 2027, risking water scarcity. [4]
• August 2025 — A Google report claimed a single text AI prompt consumes only 0.24 watt-hours, but was widely criticised for incomplete methodology. [4]
• OECD Working Paper — "Measuring the Environmental Impacts of AI Compute and Applications" quantified AI's lifecycle carbon costs. [4]

3. Background & Evolution

4. Core Static Facts

Key Definitions

• AI Compute: Computing resources (GPUs/TPUs) used to train and run AI models; primary driver of energy demand.
• Water Usage Effectiveness (WUE): Metric for data centre water efficiency; lower = better.
• Carbon Footprint of AI: Operational (inference/training energy) + embodied (hardware manufacture, disposal).
• Green AI: AI systems designed to minimise energy, water, and material resource use.

Key Numbers

Implementing Bodies (India)

• Nodal Ministry: Ministry of Electronics and Information Technology (MeitY)
• Implementation arm: IndiaAI (under Digital India Corporation)
• CERT-In: Cybersecurity and AI advisory roles
• NITI Aayog: National Strategy for AI (2018); policy advisory

Enabling Frameworks

• IndiaAI Mission (Cabinet approval, March 2024) — 7 pillars: Compute, Foundational Models, Datasets Platform, Application Development, Skilling, Startup Financing, Safe & Trusted AI
• Digital Personal Data Protection Act, 2023 — adjacent data governance framework
• National Action Plan on Climate Change (NAPCC) — broader climate policy context

5. Multi-Dimensional Analysis

Environmental

• AI training and inference consume massive electricity; data centres globally could triple electricity demand by 2030. [1]
• Water stress: AI server cooling requires freshwater; 4.2–6.6 bcm projected by 2027 — threatening countries already facing water scarcity, including India. [4]
• Land footprint: AI infrastructure's land use may exceed 14,500 km² by 2030 — larger than several Indian districts. [1]
• E-waste: Rapid GPU refresh cycles generate toxic hardware waste; underdiscussed externality.

Economic

• AI projected to add $15.7 trillion to global GDP by 2030 (PwC) — but environmental externalities are unpriced.
• India's ₹10,372 crore IndiaAI Mission aims to democratise compute access for start-ups and academia at affordable rates. [5]
• Shared compute model under IndiaAI can reduce per-unit environmental cost versus fragmented private data centres. [6]

Scientific / Technological

• Training costs for frontier AI models growing ~2.4× per year; without efficiency gains, energy demand is unsustainable. [1]
• Region-aware scheduling (running AI workloads when/where the grid is green) can cut emissions significantly. [2]
• Smaller, efficient models (e.g., distillation, quantisation) and neuromorphic chips are emerging mitigation technologies.
• India's grid being coal-heavy (>50% thermal) makes carbon-efficient AI deployment critical; every GPU cluster placed on renewable energy matters. [2]

Geopolitical / Strategic

• US, EU, and China are racing to control AI compute infrastructure; India's IndiaAI Mission positions it as a "third-way" sovereign AI power.
• EU AI Act (2024) mandates energy reporting for high-impact AI systems — sets a precedent India may need to follow for trade compatibility.
• Access to critical minerals (for GPUs) links AI strategy to mining diplomacy.

Ethical / Governance

• Carbon footprint data for AI models is often incomplete or misleading (Google report critique) — raising transparency concerns. [4]
• Water consumption data is routinely omitted from AI model cards; model developers self-report selectively. [4]
• Regulatory gap: India has no mandatory AI environmental disclosure requirement yet.
• Differential impact: energy and water costs of AI disproportionately burden Global South nations with weaker grids and greater water stress.

Administrative

• IndiaAI's shared national compute infrastructure model is a sustainability lever — concentrating GPUs reduces redundant cooling systems. [6]
• CERT-In advisory (March 2025) on responsible Generative AI use is a step; no binding environmental standard yet.
• Coordination gap between MeitY (AI) and MoEFCC (environment) in AI sustainability policy.

6. Recent Developments (Last 12–18 months)

• March 2024: Cabinet approves IndiaAI Mission — ₹10,372 crore; includes compute sustainability provisions. [5]
• September 2024: CERT-In & SISA launch CSPAI (Certified Security Professional in Artificial Intelligence) programme. [6]
• September 2024: UNEP Issue Note warns of 4.2–6.6 bcm water use by AI data centres by 2027. [4]
• March 2025: CERT-In publishes advisory on best practices for responsible Generative AI use. [6]
• August 2025: Google report claiming 0.24 Wh per text prompt draws criticism for underreporting full lifecycle costs. [4]
• Within 24 months of launch (by early 2026): IndiaAI Mission onboards >38,000 GPUs for shared compute facility for start-ups and academia. [6]
• June 2026: UN News flags AI's environmental costs — water, land, climate — as an urgent multilateral concern. [3]
• January 14, 2026: Amar Patnaik op-ed in The Hindu calls for India to adopt specific AI sustainable practices. [4]

7. Prelims Hooks

1. The global ICT sector (including AI hardware) accounts for 1.8%–2.8% of global GHG emissions per OECD estimates; alternative calculations go as high as 3.9%. [4]
2. The IndiaAI Mission was approved by the Union Cabinet in March 2024 with an outlay of ₹10,372 crore. [5]
3. IndiaAI Mission has 7 pillars; nodal ministry is MeitY (Ministry of Electronics and Information Technology). [5]
4. UNEP's September 2024 Issue Note estimated AI data centres will use 4.2–6.6 billion cubic metres of water by 2027. [4]
5. A Google report (August 2025) claimed a single text AI prompt uses 0.24 watt-hours — criticised for incomplete methodology. [4]
6. Training costs for frontier AI models have grown approximately 2.4× per year since 2016. [1]
7. Data centres could consume ~945 TWh of electricity annually by 2030 — nearly triple the combined annual use of Pakistan, Bangladesh, and Nigeria. [1]
8. Identical AI tasks produce 35.6× higher emissions on India's grid compared to Norway's grid, due to coal dependence. [2]
9. IndiaAI Mission set up 3 Centres of Excellence (CoEs) in Healthcare, Agriculture, and Sustainable Cities. [6]
10. CERT-In (under MeitY) published an advisory on responsible Generative AI use in March 2025. [6]
11. IndiaAI Mission onboarded >38,000 GPUs for shared compute within 24 months of launch, accessible to start-ups and academia at affordable rates. [6]
12. The OECD Working Paper "Measuring the Environmental Impacts of AI Compute and Applications" is a key reference document for AI's carbon footprint globally. [4]
13. GPT-4o's annual water consumption is projected at 1.33–1.58 million kiloliters. [1]
14. AI infrastructure's land footprint may exceed 14,500 km² globally by end of decade. [1]

8. Mains Relevance

Plausible Mains Question Stems

1. "Artificial Intelligence is simultaneously a tool for climate solutions and a driver of climate problems. Critically examine the environmental costs of AI and suggest a sustainable AI policy framework for India." (GS-III, 15 marks)
2. "India's coal-heavy electricity grid makes AI deployment carbon-intensive. Evaluate the environmental sustainability provisions of the IndiaAI Mission and suggest improvements." (GS-III, 10 marks)
3. "The global race for AI supremacy risks exacerbating water scarcity and carbon emissions. How should multilateral institutions respond? What role can India play?" (GS-II/III, 15 marks)

9. Related Topics to Study Next

10. Common Errors / Trap Areas

1. Wrong ministry: AI policy = MeitY (not NITI Aayog, which only advises). NITI Aayog published the National Strategy for AI 2018 but does not implement IndiaAI Mission.
2. Confusing IndiaAI pillars: There are 7 pillars, not 5 or 4. "Safe & Trusted AI" is a distinct pillar — don't conflate with CERT-In's separate advisory role.
3. UNEP water figure: The 4.2–6.6 bcm figure is for 2027 projection, not current usage — framing matters in answers.
4. Carbon figure range: ICT sector emissions cited as 1.8%–2.8% or 2.1%–3.9% depending on methodology — both appear in OECD documents; state "estimates vary" rather than asserting one figure.
5. Approval year: IndiaAI Mission approved March 2024 — not 2023 (when NITI Aayog's AI strategy was being revised) and not 2025.