India Datacentre Theme

Layer 0

The Signal

Why now

AI Compute Wave 900M+ India Internet Users Data Localisation Mandate $100B Hyperscaler Commits 12/30 top performers = DC theme

↓

Layer 1

The Mechanics

What is actually being built

Power Infrastructure Cooling Systems Compute & Storage Fibre Connectivity Physical Shell & EPC AI GPU Racks: 40–120 kW vs 5–10 kW standard

↓

Layer 2

Cause Tree

5 root causes

① Digital Consumption Scale ② Data Localisation Policy (DPDP Act 2023) ③ Infrastructure Status + Tax Incentives ④ Hyperscaler India Build (MSFT $3B, Google $15B, AWS) ⑤ AI Inference Density — 10–30x power per rack

↓

Layer 3

Solution Space

What each cause demands

Colocation & Hyperscale Capacity Sovereign Cloud Infrastructure Utility-Scale Power + Grid Bypass High-Density Liquid Cooling Fibre Optic Backbone Fuel Cell / Gas-Based Backup Power

↓

Layer 4

Build Requirements

Bottlenecks & inputs

⚡ Transformers [18–24 mo lead time] 🧊 Liquid Cooling Systems [AI-grade] DG Sets & UPS Backup High-Current Busbars & Wiring 🔌 Fibre Preforms [global shortage] Precision Hose & Piping Structural Steel & PEB Shells

↓

Layer 5

Value Chain Actors

India-listed companies at each node

⚡ Power, Transformers & Electrical

🇮🇳 POWERINDIA 🇮🇳 GVT&D 🇮🇳 SCHNEIDER 🇮🇳 ABB 🇮🇳 CUMMINSIND 🇮🇳 TDPOWERSYS 🇮🇳 CGPOWER 🇮🇳 GVPIL 🇮🇳 MTARTECH 🇮🇳 PRECWIRE

🧊 Cooling & Thermal Management

🇮🇳 KRN 🇮🇳 DEEDEV 🇮🇳 AEROFLEX 🇮🇳 VOLTAS 🇮🇳 AMBER 🇮🇳 BLUESTARCO

🔌 Fibre, Cables & Connectivity

🇮🇳 STLTECH 🇮🇳 FINCABLES 🇮🇳 HFCL 🇮🇳 APARINDS 🇮🇳 PRECWIRE 🇮🇳 BHARTIHEXA 🇮🇳 VINDHYATEL ⚠ consider adding

🏗 EPC, Construction & Operators

🇮🇳 LT 🇮🇳 KPIL 🇮🇳 TECHNOE 🇮🇳 ANANTRAJ 🇮🇳 INTERARCH 🇮🇳 BHARTIARTL 🇮🇳 ADANIENT 🇮🇳 RELIANCE 🇮🇳 LODHA

💻 Compute, Hardware & Distribution

🇮🇳 NETWEB 🇮🇳 E2E 🇮🇳 REDINGTON 🇮🇳 SYRMA 🇮🇳 DSSL 🇮🇳 BBOX

↓

Layer 6

Research Landscape

Stage, access, milestones

Power/Transformers — Mid Acceleration | 2yr revenue visible Cooling — Early Acceleration | least re-rated layer Fibre/OFC — Active Supercycle | Finolex preform Mar 2026 EPC/Build — Growing | large-cap well-tracked Compute — Early | higher volatility, thin coverage

📍 Watch: Finolex preform commissioning 📍 Watch: Transformer order inflows Q-Q 📍 Watch: China preform utilisation vs 90% 📍 Watch: ANANTRAJ first MW commissioned 📍 Watch: India-US trade deal (STL tariff)

⚡ Key Research Insight

Power is the binding constraint, not fibre or land. An AI-grade datacentre draws 40–60 MW vs 25–30 MW for a standard facility. India's grid cannot deliver this with hyperscaler-grade uptime SLAs — making every POWERINDIA, GVT&D, CUMMINSIND, and TDPOWERSYS order book a direct proxy for the build cycle. Transformer lead times of 18–24 months make this the most duration-visible layer through FY27–28.

In fibre, Finolex is the cleaner 2-year risk-reward — zero debt, ₹2,713 Cr cash, preform plant commissioning March 2026, margin inflection from 2.5% → 8–9% ahead. STLTECH carries superior technology moat but US tariff (-760bps/quarter) and Prysmian lawsuit create binary risk that requires a trade deal resolution or legal settlement to fully unlock.

The cooling layer (KRN, DEEDEV, AEROFLEX) is the least re-rated sub-theme with the highest structural growth rate — liquid cooling CAGR of 27.7% through 2030, driven entirely by AI rack density exceeding air cooling thresholds.

India Datacentre Theme — Value Chain Map