Gaurav Arya | Energy Decoded Newsletter | April 2026
India just crossed 250 GW of non-fossil power capacity. In the same month, it quietly became home to over ₹5 lakh crore in committed data center investments from Microsoft, Amazon, Google, and Adani. The connection between these two facts is the most important energy story of the decade — and almost nobody is telling it the right way.
What Even Is a Data Center?
If you have ever used Google, watched Netflix, sent a WhatsApp message, or asked an AI chatbot a question, you used a data center. A data center is a warehouse full of servers — computers that run all day, every day, without stopping. Because servers run constantly and generate enormous heat, a data center has two giant costs: electricity to run the machines, and more electricity to cool them. Together, these account for roughly 40% of a data center’s operating expenses. Power is not a supporting cost — it is the business.
A single mid-sized data center of 100 MW consumes as much electricity as a city of 100,000 homes. India, as of early 2025, had 1,280 MW of operational data center capacity across 130 facilities. That number is about to explode.
The Scale That Should Shock You
India’s data center sector is growing faster than almost any infrastructure category in the country’s history:
- Installed Capacity: 1.2–1.5 GW (2025) → 8–10 GW (2030) → 13 GW (2032)
- Electricity Consumed: 10–15 TWh/yr (2025) → 40–57 TWh/yr (2030) → ~100 TWh/yr (2032)
- Share of India’s Power: ~0.84% → 2.5–3% → ~5%
- Renewables Required: ~30 GW (2030) → ~40 GW (2032)
S&P Global projects India will become the second-largest data center electricity consumer in Asia-Pacific by 2030, overtaking Japan and Australia. Microsoft, Amazon, and Google together have committed over $67.5 billion to Indian data center infrastructure since December 2025 alone.
The Renewable Paradox — Why 10 GW of Demand Needs 40 GW of Supply
Here is where most coverage gets it wrong. When journalists report that India needs “more renewables for data centers,” they imply a simple equation: 10 GW of data centers needs 10 GW of solar. That equation is dangerously incomplete.
The core problem is intermittency. Solar panels generate electricity only when the sun shines — about 5–6 hours of peak generation per day. A data center, by contrast, needs full power 24 hours a day, 365 days a year.
The math: A hybrid wind-solar project in India has an average CUF of 40–45%. To deliver 100 MW annually, you need to install 100 ÷ 0.45 = 222 MW of generation capacity. But annual energy equivalence is not enough — you also need battery storage. To ensure hourly reliability for 100 MW, you need 400 MWh of usable battery storage per day. Net result: each 100 MW of data center load requires 250+ MW of renewable generation and 400–500 MWh of battery storage.
Scale that to 13 GW of national data center demand by 2032, and you arrive at the headline number: 30–40 GW of dedicated renewable capacity — before accounting for any other power demand in the country. This is the 3:1 ratio that no mainstream coverage is explaining.
The Geography Problem — Why Clustering Creates Grid Risk
India’s data centers cluster in five cities: Mumbai, Delhi NCR, Chennai, Hyderabad, and Bengaluru. As of H1 2025, Mumbai alone accounts for 594 MW of operational capacity — nearly half of India’s total. By 2030, data centers in each major hub state could add 2–3 GW of peak electricity demand, equivalent to 5–20% of their current total peak load.
There is a second hidden problem: water. A 100 MW data center consumes roughly 2 million litres of water per day for cooling. India’s hottest data center cities already face summer water stress — and AI-linked data center expansion is colliding with water security in the same geography.
The AI Multiplier
Standard data centers are already power-hungry. AI data centers are in a different category entirely. A traditional server rack consumes 10–15 kilowatts. An AI server rack running dense GPU clusters consumes 50–150 kilowatts — up to 10 times more. As India’s data centers shift to AI-specific workloads, the power density of the average facility will rise sharply.
The Indian Express reported that AI workloads alone could push India’s national peak power demand up by 30 GW — an addition larger than the entire installed capacity of states like Rajasthan.
The $100 Billion Bet — Adani’s Answer
On February 17, 2026, Adani Group announced $100 billion by 2035 to build renewable-powered, hyperscale AI data centers. The plan integrates: Renewable generation (anchored by Adani Green Energy’s 30 GW Khavda project in Gujarat), Battery storage (large-scale BESS co-located with generation), Transmission infrastructure, and Compute (AdaniConnex’s 2 GW data center footprint expanding to 5 GW). All four layers under one architecture.
My Take — What India Must Get Right
What is working: India’s renewable buildout is genuinely world-class. The country crossed 250 GW of non-fossil capacity in 2025. Solar tariffs are among the world’s lowest. India has structural cost advantages in electricity, land, and talent that no other major data center market can replicate simultaneously.
What is not working:
- The grid is not data-center-ready. Data centers need 99.999% uptime. India’s grid currently delivers nowhere near that reliability at the distribution level.
- Storage policy is moving too slowly. India’s Ministry of Power mandated co-located BESS for new solar projects in February 2025 — but the mandate covers only 10% of installed solar capacity at 2-hour duration.
- Water is the invisible constraint. There is no national policy framework for data center water use.
- Grid planning is still backward-looking. Grid reinforcement takes 5–7 years. Data centers are being built in 2–3 years.
The single most important policy action: Establish a Data Center Energy Compact — a joint MoPNG-MoP-MNRE framework linking data center operating licenses to demonstrated round-the-clock renewable procurement, co-located storage, and water use reporting.
Sources: All data cited from government sources, IEA, CEA, Deloitte, S&P Global, and company announcements.