Adapting Cooling Policy to India’s AI Ambitions

UPSC Syllabus: Gs Paper 3- Infrastructure

Introduction

India’s rapid AI expansion is driving large-scale growth of hyperscale data centres, but their high heat generation creates a parallel cooling challenge. Cooling now affects electricity demand, water use, grid stability and infrastructure costs. The India Cooling Action Plan (ICAP) provides an important base, but it must evolve to address AI-driven cooling requirements while supporting sustainable and efficient digital infrastructure.

India’s Rapid AI Infrastructure Expansion

1. Rapid data-centre growth: India’s data-centre capacity is projected to exceed 12 GW by 2030, rising sharply from around 1.5 GW in 2025.
2. Rising electricity requirement: Data centres could require around 26.3 GW of electricity by 2031–32, equal to nearly 6.7% of projected national consumption.
3. Higher AI rack density: Traditional racks used 5–15 kW, while modern AI racks can exceed 120 kW, sharply increasing heat removal requirements.
4. Rapid GPU power growth: GPU power has risen from about 700W in 2022 to around 1,200W, making conventional cooling increasingly inadequate.
5. Cooling as a major energy load: Cooling can consume 30–50% of data-centre electricity, making efficient thermal management essential for expanding computing capacity.
6. AI workload expansion: AI workloads are expected to drive at least 75% of future Indian data-centre growth, making cooling capacity central to digital expansion.

Emerging Challenges: Energy, Water and Infrastructure Pressures

1. Persistent water demand: Even advanced cooling systems that reduce water use cannot eliminate cooling-related water requirements, creating pressure in water-stressed regions.
2. Round-the-clock electricity demand: Data centres operate continuously, while solar generation falls after sunset, creating a need for firm power to support uninterrupted AI workloads.
3. Competing cooling demand: India is expected to add 130–150 million room air conditioners between 2025 and 2035, increasing household cooling demand alongside AI data-centre requirements.
4. Peak grid pressure: Extreme heat can sharply increase cooling demand, creating additional stress on electricity systems when both data centres and households require reliable power.
5. Unequal cooling burden: Poor households can spend up to 8% of their income on cooling electricity, compared with 0.2–2.5% for high-income households, making cooling efficiency an equity issue.
6. Regional resource pressure: Rapid AI infrastructure growth can intensify local pressure on electricity and water resources, especially in urban regions already facing scarcity.

Why India’s Existing Cooling Policy Needs to Evolve

1. ICAP’s existing framework: The India Cooling Action Plan (ICAP), introduced in 2019, provides a 20-year roadmap for sustainable cooling across buildings, cold chains, refrigeration and transport.
2. Existing policy targets: ICAP aims to reduce cooling demand by 20–25% and cooling energy requirements by 25–40% by 2037–38, providing a broad base for sustainable cooling.
3. AI-era policy gap: ICAP was formulated before hyperscale AI infrastructure became significant, so AI data centres were not recognised as a distinct cooling sector.
4. Higher cooling intensity: AI workloads use high-density GPUs that generate much more heat than conventional computing, requiring advanced cooling systems beyond ICAP’s traditional applications.
5. Need for sector-specific standards: Future cooling policy should set Power Usage Effectiveness (PUE), cooling-efficiency and water-efficiency benchmarks for AI data centres.
6. Climate and resource alignment: AI expansion needs to follow the “People, Planet and Progress” approach by balancing computing growth with energy security and environmental sustainability.

Emerging Solutions: Efficient Technologies and Integrated Planning

1. Direct-to-chip cooling: Direct-to-Chip (D2C) liquid cooling removes heat directly from high-power chips and is more efficient than conventional air-based cooling.
2. Immersion cooling for high-density AI: Immersion cooling places computing equipment in a cooling liquid, allowing high-power AI servers to manage heat more efficiently than conventional air cooling.
3. Heat recovery: Recovering waste heat from data centres can improve resource efficiency and extend ICAP’s focus on sustainable cooling technologies.
4. Water-efficient cooling: Data centres can use different grades of water, including lower-quality water where suitable, reducing dependence on high-quality freshwater resources.
5. Efficient infrastructure design: New AI campuses should be designed around 100 kW-plus rack densities, liquid cooling and renewable-powered infrastructure rather than retrofitting older facilities.
6. Geographic resource matching: Data-centre locations should consider regional availability of renewable energy, water, land, climate conditions and grid capacity.
7. City-level planning: ICAP’s implementation across 250 Indian cities provides a base for linking cooling planning with geographic decisions on future AI-centre locations.

Way Forward

1. Make data centres an ICAP sector: Future versions of ICAP should explicitly recognise hyperscale data centres and define cooling requirements for AI infrastructure.
2. Link AI with firm clean power: Long-term renewable energy combined with storage can help meet data centres’ continuous demand while reducing pressure on the wider electricity system.
3. Strengthen cooling standards: Minimum efficiency standards should cover cooling systems, PUE and water use so that digital infrastructure expands without unnecessary resource consumption.
4. Support indigenous innovation: Indian companies should develop cooling solutions suited to local conditions instead of depending entirely on imported technologies designed for different markets.
5. Improve distribution capacity: Stronger distribution companies are needed to support grid upgrades and reliable power delivery as AI and household cooling loads increase.
6. Use private investment strategically: Creditworthy data-centre operators can provide long-term demand that helps finance renewable energy and storage projects benefiting wider electricity users.
7. Coordinate public and private action: Government should set standards and de-risk infrastructure, while investors and innovators take risks on new cooling and energy technologies.

Conclusion

India’s AI ambitions require not only more computing capacity but also efficient management of the heat, electricity and water that support it. Updating ICAP to recognise AI data centres, promoting efficient cooling and planning infrastructure around regional resources can reduce future pressures. India must therefore align AI growth, sustainable cooling and reliable power to build a competitive and resource-efficient digital economy.

Question for practice:

Discuss the need to adapt India’s cooling policy to support its growing AI and data-centre infrastructure sustainably.

Source: Businessline