Why in News: A NITI Aayog report in October 2026 on AI data centres recommended grid-readiness checks, 20–30-year water-availability assessments and milestone-based State incentives, as India’s data-centre capacity of about 1.6 GW is projected to reach about 6 GW by 2029. MeitY highlighted the growth of domestic cloud services (about 1.57 GW of data-centre capacity in August 2026 and 30 empanelled cloud service providers), and the National Supercomputing Mission reported 40 indigenous systems with about 68 petaflops of combined capacity.
Why Compute Is the New Infrastructure
- Training and running AI models needs vast numbers of GPUs, huge power and cooling — a single large AI data centre can draw as much power as a mid-sized city.
- Data localisation, the DPDP Act and government cloud demand keep data in India; global firms have announced multi-billion-dollar Indian data-centre investments.
- The IndiaAI Mission provides subsidised GPU compute (tens of thousands of GPUs) to start-ups and researchers.
NITI Aayog’s Concerns and Recommendations
- Power: data centres need reliable, 24×7, increasingly green power — grid connections and renewable tie-ups must be planned in advance.
- Water: evaporative cooling uses large volumes; siting in water-stressed cities (Chennai, Bengaluru, Hyderabad, Mumbai) risks conflict — hence long-horizon water assessments and air or liquid cooling.
- Incentives tied to milestones rather than upfront land and tax giveaways.
The Bigger Ecosystem
- Cloud: MeghRaj (GI Cloud, 2014) and empanelment of providers; sovereign cloud for sensitive data.
- Supercomputing: National Supercomputing Mission (2015, C-DAC and IISc) — PARAM series (PARAM Rudra systems, 2024); AIRAWAT AI supercomputer at C-DAC Pune.
- AI “neolabs”: independent research labs founded by leading AI researchers (about 29 globally, having raised about US$ 23 billion) focus on safe, human-aligned advanced AI — India is exploring similar public-interest labs.
Way Forward
- Green data-centre policy: renewable sourcing, waste-heat reuse, water-use efficiency standards.
- Domestic chip and server manufacturing under the India Semiconductor Mission.
- Distribute capacity beyond the big metros to cooler, water-secure locations with grid capacity.
Static Background — Computing Terms
- FLOPS: floating-point operations per second; 1 petaflop = 10^15 FLOPS; exaflop = 10^18.
- GPU (graphics processing unit): many small cores working in parallel — ideal for AI training.
- Data-centre efficiency: Power Usage Effectiveness (PUE) = total facility energy ÷ IT equipment energy (1.0 is ideal); Water Usage Effectiveness (WUE).
- Cloud models: IaaS (infrastructure), PaaS (platform), SaaS (software).
- Fastest Indian supercomputers rank in the global TOP500; the world’s fastest are exascale machines.
| Prefix | FLOPS |
|---|---|
| Teraflop | 10^12 |
| Petaflop | 10^15 |
| Exaflop | 10^18 |
Exam Angle — Prelims & Mains
- UPSC PYQ Link: Prelims has asked about cloud computing models and supercomputers (e.g., the National Supercomputing Mission).
- Practice Mains Question (GS Paper 3, 15 marks, 250 words): The AI boom will be constrained less by algorithms than by electricity and water. Discuss with reference to India’s data-centre expansion.
- Prelims Angle: FLOPS prefixes, PUE and the missions.
- Practice Prelims MCQ 1: Consider the following statements:1. One petaflop equals 10^15 floating-point operations per second.2. A Power Usage Effectiveness (PUE) value closer to 1 indicates a more energy-efficient data centre.3. The National Supercomputing Mission is implemented by C-DAC and IISc.Which of the statements given above are correct?(A) 1 and 2 only(B) 2 and 3 only(C) 1 and 3 only(D) 1, 2 and 3
Show answer
Answer: (D) 1, 2 and 3 — Statement 1 is CORRECT, Statement 2 is CORRECT and Statement 3 is CORRECT. - Practice Prelims MCQ 2: In cloud computing, a service that provides ready-to-use applications over the internet (e.g., email) is:(A) Infrastructure as a Service(B) Platform as a Service(C) Software as a Service(D) Network as a Service
Show answer
Answer: (C) — SaaS delivers complete applications; IaaS provides servers and storage; PaaS provides development platforms.
- Practice Prelims MCQ 1: Consider the following statements: