Four Questions India Must Answer Before Its AI Data Centre Boom Outpaces Its Grid

The Lift Line

Growth without published data is a claim; growth with published data is a licence.

Why This Editorial Matters for Your Exam

AI infrastructure, energy transition and industrial resource use now cross GS3 syllabus, and the data-centre sector is the current-affairs peg for a compact answer. The specific numbers, 1.8-1.9 litres per kWh, 1.45 PUE improving to under 1.3, 1.1 million litres per MW nationwide, are the difference between a middle-band and top-band Mains answer. The comparison of a 50 MW data centre against a 4,000-person commercial building is a good illustrative device.

GS Paper 3: Indian Economy; issues relating to planning, mobilization of resources; effects of liberalization on the economy, changes in industrial policy; conservation, environmental pollution and degradation; environmental impact assessment.

Background and Context

This is a signed commentary by Sharad Agarwal, Chief Executive Officer of Sify Infinit Spaces, a data-centre operator. Read the efficiency figures below with that interest in mind: they are one operator’s internal numbers, published by the operator’s chief executive, and they are not independently verifiable. The argument is his, not the newspaper’s editorial line.

For decades, data centres worked quietly, largely out of sight. They powered digital banking, e-commerce, streaming and enterprise technology. During Covid, they kept the country’s digital spine intact. That quiet phase is over. Large investment announcements have brought data centres firmly into public view, and AI has brought even greater scrutiny. This commentary raises four questions India must answer.

The Analysis

1. The first question is about water intensity. Do data centres consume more water than comparable industrial or commercial facilities? Water intensity is 1.8-1.9 litres per kWh of electricity consumption, annualised to roughly 15-16 million litres per MW. A closed-loop air-cooled system, rejecting heat using fans and refrigerant loops rather than evaporating water, consumes less than 0.2 litre per kWh. India’s nationwide operation reportedly consumed 1.1 million litres per MW, about 7 per cent of the concern figure. A 50 MW data centre on 4 lakh sq ft uses roughly 55 million litres a year against a commercial building of similar area housing 4,000 people at a 45 litre-per-person-per-day norm, which comes to 65.7 million litres.

2. The second question is about the water-electricity trade-off. Saving water while raising energy consumption is not a sustainability victory. If heat transfer is done using mechanical refrigeration, more electricity is consumed. Running compressors harder to save water can defeat the purpose. This is where grey and recycled water use matters as an efficiency lever.

3. Hybrid cooling is the operational answer. A hybrid cooling system, closed-loop air-cooled combined with evaporative grey water, reduces electricity consumption by 10 per cent. An air-cooled data centre running at 1.45 PUE is expected to improve to under 1.3 PUE with grey water. Whether that is sufficient in a climate as hot and humid as many parts of India is the operational question the industry must answer transparently.

4. The third question is about the electricity intensity gap. How is a data centre in India compensated for using more electricity than facilities in the Nordic countries, where cold climates make cooling almost free? The answer the industry has adopted is to decarbonise the electricity used through renewable capacity procured under long-term contracts.

5. The fourth question is about community grid impact. A large data centre typically draws tens of megawatts continuously. Does it crowd out capacity meant for homes and local businesses? Most large facilities in India connect through dedicated high-voltage substations rather than the local distribution network, and a growing share of their power comes from renewable generators under long-term contracts, adding new generation capacity rather than drawing down what exists. Predictable industrial loads can also improve utilities’ load-factor economics. Well-planned clusters can catalyse grid strengthening that benefits a region beyond the facility’s own walls.

6. The additionality problem cannot be handwaved. Renewable PPAs in India often depend on open access rules that are subject to state resistance, and real additionality, the extent to which the PPA represents new generation rather than diverted supply, is a policy question, not just a corporate one. A shift from industry self-reporting to an audited public register is the way to convert claim into licence.

7. The task is dual, not sequential. AI will require significant physical infrastructure. Growth cannot be an excuse for careless resource use. The task is to build AI infrastructure and manage its footprint at the same time, with the same rigour already applied to uptime and reliability, publishing data, acknowledging gaps and investing continuously in improvement.

8. The counter-argument the piece anticipates. India can stay on the sidelines and later ask why the next generation of technology was built elsewhere. Or it can build, with renewable energy as the backbone, efficiency embedded in design from the outset, and transparency to prove it is working. The four questions this piece raises are answerable, provided they are answered in public.

Data and Institutions Vault

Prelims-grade facts:

The efficiency numbers:

• Evaporative cooling water intensity: 1.8-1.9 litres per kWh, roughly 15-16 million litres per MW annualised.
• Closed-loop air-cooled water intensity: under 0.2 litre per kWh.
• India’s nationwide data-centre operation consumed 1.1 million litres per MW, roughly 7 per cent of the concern figure.
• A 50 MW data centre on 4 lakh sq ft uses roughly 55 million litres of water a year.
• A comparable 4 lakh sq ft commercial building housing 4,000 people at 45 litres per person per day uses 65.7 million litres a year.
• Hybrid cooling (closed-loop air-cooled plus evaporative grey water) reduces electricity consumption by 10 per cent.
• PUE improvement expected: from 1.45 air-cooled to under 1.3 with grey water.

The technical concepts:

• PUE (Power Usage Effectiveness) is the ratio of total facility energy to IT equipment energy; the industry standard efficiency metric; a PUE of 1.0 is a theoretical minimum.
• Grey water refers to gently used water from bathroom sinks, showers and washing machines, not fresh potable water.
• Nordic countries are the benchmark for low-PUE data centres because cold ambient temperatures make cooling almost free.

The India power context:

• Peak power demand touched 270.8 GW on 21 May 2026 against roughly 180 GW in 2019.
• Most large Indian data centres connect through dedicated high-voltage substations rather than local distribution networks.
• Open access under the Electricity Act, 2003, allows large consumers to purchase power from any generator across state boundaries, subject to state cross-subsidy surcharges and additional surcharges.

The AI infrastructure context:

• Major Indian data-centre players include Sify Infinit Spaces, Yotta Data Services, CtrlS, ESDS and NTT-Netmagic.
• The Ministry of Electronics and Information Technology (MEITy) has announced Data Centre Policy consultations; the IndiaAI Mission approved in March 2024 has a Rs 10,371.92 crore outlay over five years.

Watch the trap: PUE is a facility-level ratio of total to IT energy; it does not directly measure water use. Water Usage Effectiveness (WUE) is the parallel metric for water. Both matter, and improving one can worsen the other.

A second trap: Open access is a right granted under the Electricity Act, 2003; it is not an automatic exemption from state cross-subsidy surcharges and additional surcharges levied under the Act.

The Debate

FOR the industry’s architectural answer: Modern hybrid cooling, dedicated high-voltage substations and long-term renewable PPAs together mean that the water and electricity footprint per unit of computing is falling, not rising. Predictable industrial loads improve utility load factors; well-planned clusters catalyse grid strengthening. India cannot afford to sit out the AI infrastructure build.

AGAINST accepting industry self-report: Self-reported PUE and water numbers cannot substitute for independent audit. Renewable PPAs often depend on open access rules that face state resistance, so additionality is uncertain. The local social licence of a 50 MW facility on land-scarce metros is a stronger constraint than the aggregate national footprint.

Balanced verdict: Both are true. The industry’s architectural answer is real; the accountability layer is missing. A mandatory published PUE-and-water-intensity register audited annually, minimum renewable-energy sourcing benchmarks with certified additionality, enforceable open access across state boundaries, siting rules away from stressed groundwater zones and a community benefit fund tied to grid drawdown are the correct combination.

How to Think About This

For any industrial infrastructure with dual resource intensity, ask four questions. What does the industry itself claim, and can that claim be independently audited? What is the trade-off between the two resources, and does the technology permit optimisation on both? What is the marginal community impact, and does the connection architecture avoid crowding out? What accountability mechanism converts the claim into a licence? AI data centres sit on all four. The industry has credible answers to the first three; the fourth is a policy job.

Diagram-in-Words

PYQ Linkage

• UPSC CSE Mains 2021, GS3: “How is S-400 air defence system technically superior to any other system presently available in the world?” Different subject; same specifications-versus-benchmark discipline.
• UPSC CSE Mains 2019, GS3: “How is the Government of India protecting traditional knowledge of medicine from patenting by pharmaceutical companies?” Different frame; same regulatory-audit logic.
• UPSC CSE Mains 2016, GS3: “Give an account of the current status and the targets to be achieved pertaining to renewable energy sources in the country. Discuss in brief the importance of National Programme on Light Emitting Diodes (LEDs).” Direct on the renewable-sourcing dimension.

Sources: Economic Times, Ministry of Electronics and IT, MNRE

Source: Four Questions India Must Answer Before Its AI Data Centre Boom Outpaces Its Grid — Ujiyari.com | Free UPSC & State PCS Editorial Analysis