India's data center sector has become the most heavily capitalised digital-infrastructure play in Asia outside China. CBRE places cumulative investment commitments at USD 126 billion by end-2025, projected to cross USD 180 billion in 2026, with operational capacity reaching 1,700+ MW in 2025 and forecast to add another ~500 MW in 2026 — a 30% year-on-year jump — on a trajectory toward 8–10 GW by 2030.
The 2024–2026 window has been defined by a hyperscaler/AI-infrastructure shockwave: Google's USD 15B Vizag AI hub, AWS's USD 35B India plan, Microsoft's USD 17.5B commitment, Reliance's gigawatt-scale campuses, Adani's USD 100B 10-year plan, and TCS's USD 2B HyperVault JV with TPG — an unprecedented convergence of global and domestic capital.
The structural case is clear: India generates approximately 20% of the world's data but hosts only 3% of global data centers. This deficit, combined with the DPDP Act 2023 mandating on-shore data residency, creates a demand tailwind unmatched outside the US and China.
A data center is a physical facility that houses computing infrastructure — servers, storage arrays, networking equipment — along with the power, cooling, security, and connectivity systems required to keep that infrastructure running continuously. Every digital service — from a UPI payment to an AI model training run — depends on data center capacity somewhere in the chain.
The Uptime Institute Tier Classification System defines four levels of data center resilience. Every lease negotiation, insurance policy, and SLA references these tiers. RBI mandates Tier IV for payment system data.
PUE (Power Usage Effectiveness) = Total Facility Energy / IT Equipment Energy. A PUE of 1.0 means every watt goes to compute; a PUE of 2.0 means half is lost to cooling and overhead. India's colo average is 1.4–1.6; hyperscale target is 1.1–1.3. PUE is referenced throughout this report as the primary efficiency benchmark.
Traditional data centers are designed for general-purpose compute at 5–8 kW per rack. AI-ready data centers house GPU clusters (Nvidia H100, B200, B300) requiring 40–130+ kW per rack — a 10–20× increase in power density. This demands fundamentally different infrastructure: liquid cooling (direct-to-chip or immersion), reinforced floor loading, higher-capacity power distribution, and specialised fire suppression. AI-ready facilities command a 15–25% lease premium over traditional colo, and are the only facilities hyperscalers will sign long-term MSAs with from 2027 onward.
Typical colocation build costs in India are approximately INR 400–430 million per MW (USD 5–5.5M/MW) — among the lowest globally. This has risen from ₹40–45 crore/MW in 2020 to ₹60–70 crore/MW in 2026, driven by land inflation, construction cost escalation, and higher power density requirements for AI workloads.
AI-ready builds with liquid cooling run USD 8–12M/MW globally. In India, given lower labour and construction costs, AI-ready builds are estimated at ₹80–100 crore/MW (USD 9.5–12M/MW), still ~30% below US equivalents.
| Metric | India Benchmark | Global Comp. |
|---|---|---|
| EBITDA margin (stabilised colo) | 40–55% | 45–60% |
| Revenue CAGR (FY17–FY23) | ~25% | 12–18% |
| CareEdge projected CAGR (FY24–26) | 32% | — |
| Absorption rate (2023) | 93% | 85–92% |
| Absorption rate (H1 2025) | >100% | — |
| CtrlS operating margin (FY25) | ~50% | — |
| CtrlS DSCR (FY25) | 1.6–1.9× | — |
Data center REITs globally trade at 25–40× P/FFO — a dramatic outlier versus the broader REIT sector at 24×. M&A transactions in the DC sector averaged 20–30× EBITDA during 2023–2025. The KKR/Singtel acquisition of STT GDC at ~USD 10.9B EV for ~400 MW India capacity implies roughly USD 27M/MW — a premium reflecting contracted revenue and platform scale.
| Tier | Monthly Cost | Notes |
|---|---|---|
| Full rack (42U) | ~₹50,000/mo | Incl. power, bandwidth, basic services |
| Half rack | ~₹30,000/mo | SME / startup segment |
| Quarter rack | ~₹25,000/mo | Entry-level |
| Wholesale (per kW) | $80–140/kW/mo | 250 kW – 4 MW commitments |
| Hyperscale (per kW) | $60–100/kW/mo | >4 MW, 10–15yr leases |
| Investor | Target / Vehicle | Amount | Category | Location | Capacity | Strategic Rationale |
|---|---|---|---|---|---|---|
| Microsoft | Azure India (self-build) | $17.5B | Hyperscaler | Mumbai, Pune, Hyderabad | 150+ MW | First-mover AI cloud; GPU-rich Azure regions; Azure OpenAI training clusters; 10M AI skill target |
| Amazon / AWS | AWS India (self-build + leases) | $35B | Hyperscaler | Mumbai, Hyderabad | 2–3 GW target | Localize AI/cloud; capture $30B+ cloud mkt by 2029; Navi Mumbai self-build ($430M, Apr-26) |
| AdaniConneX + Airtel JV | $15B | Hyperscaler | Visakhapatnam, AP | 1 GW AI hub | Largest Google India investment ever; AI training infra for India + APAC; subsea cable landing station | |
| Reliance Ind. | Digital Connexion JV (Brookfield + Digital Realty) | $110B (10yr plan) | Indian Cong. | Jamnagar, Visakhapatnam | 3–4 GW | USD 11B MoU Andhra Pradesh (1 GW Vizag); 6 GWp dedicated solar; Jamnagar 3 GW flagship campus |
| Adani Group | AdaniConneX (50:50 EdgeConneX) | $100B (10yr plan) | Indian Cong. | Chennai, Noida, Hyderabad, Vizag, Pune | 5 GW target | World's largest integrated DC platform; powered by 30 GW Khavda renewable project; Google AI hub anchor tenant |
| TCS + TPG | HyperVault JV | $2B | Indian Cong. | Pan-India | 1.2 GW (Phase I) | Liquid-cooled AI-ready; OpenAI signed 100 MW (Nov-25); TPG Rise Climate / ALTÉRRA funding; 51% TCS control |
| KKR + Singtel | STT GDC (82% buyout) | S$6.6B (~$5.1B) | Private Equity | 10 Indian cities | 400+ MW | EV ~$10.9B; largest APAC DC M&A; STT GDC India ~28% revenue share; 550 MW expansion committed |
| Blackstone | Lumina CloudInfra | $5B+ (MoU) | Private Equity | Mumbai, Chennai, Telangana | 600 MW pipeline | Lumina: Chandivali 60 MW (₹5,000 cr); Chennai USD 1.1B Ambattur 216 MW; twin USD 3B MoUs MIDC & CIDCO |
| Alpha Wave, Carlyle, Anchorage | Nxtra by Airtel | $1B | Private Equity | Pan-India | 300 MW → 1 GW | Post-money: $3.1B; tripling capacity; AI-ready campuses Chennai, Mumbai, Kolkata; Google Vizag hub partner |
| Meta | Sify Technologies (lease) | ~$1.7B (est.) | Hyperscaler | Visakhapatnam, AP | 500 MW | Waterworth subsea cable (50,000 km) CLS via Sify; first large-scale Indian leased DC for Meta; Vizag subsea diversification |
| Investor | Target / Vehicle | Amount | Category | Location | Capacity | Strategic Rationale |
|---|---|---|---|---|---|---|
| Princeton Digital Group | PDG India (greenfield) | $2.5B+ | Private Equity | Mumbai, Chennai, Hyderabad | ~1 GW pipeline | $160M green loan MU1; India's first 24/7 CFE PPA (Tata Power RE + Flexidao, Sep-25); IGBC Platinum |
| Yotta (Hiranandani) | Yotta AI Hub | $2B | Indian Cong. | Greater Noida, Mumbai | 250 MW (Phase I) | Asia's largest Nvidia Blackwell B300 supercluster (20,000+ GPUs); controls ~70% India GPU capacity; IPO planned |
| Sify Technologies | Sify Infinit Spaces | $5B | Indian Cong. | Multi-city (8+ sites) | 350 MW+ build | IPO DRHP filed Oct-25; Tier-2 expansion (Lucknow, Chandigarh, Nagpur); NVIDIA-certified; Meta Vizag build-out |
India hosts ~19 international submarine cables with landing stations concentrated in Mumbai, Chennai, Kochi, Tuticorin, and Thiruvananthapuram. Combined lit and activated capacity stands at approximately 193 Tbps. Two-thirds of India's international subsea capacity lands in Mumbai (primarily Versova), with Chennai as the secondary hub. A new wave of cables — driven by hyperscaler demand — is transforming Visakhapatnam into a third major gateway, breaking the Mumbai-Chennai duopoly.
Rule of thumb: 1,000 km of fibre adds ~10 ms of round-trip delay. Chennai provides the lowest-latency route from India to Singapore and APAC — critical for trading, fintech, and SaaS. Bengaluru, despite being India's IT capital, faces a 350 km terrestrial backhaul to Chennai's cable landing station, adding 4–6 ms of latency. Mumbai's westward cables to the Middle East and Europe traverse the Red Sea — a demonstrated vulnerability zone. Vizag's new direct cables to Singapore and South Africa bypass both chokepoints.
The Telecommunications (Authorisation for Captive Telecommunication Services) Rules 2025, finalised in early 2026, enable large enterprises and hyperscalers to build dedicated captive networks — including subsea capacity — without requiring a full telecom licence. TRAI has also updated its Cable Landing Station framework, introducing a two-tier CLS-PoP model that lowers barriers to entry for non-telco entities. This is why Google can build its own cable landing station in Vizag and why Sify is hosting Meta's Waterworth landing — the regulatory architecture has fundamentally shifted to accommodate hyperscaler-direct infrastructure.
Data centers received infrastructure status in 2022, enabling priority lending, longer-tenor debt, and lower borrowing costs — the structural financing unlock. The DPDP Act 2023 is the legal anchor for data localisation.
The Union Budget 2026–27 announced a landmark 21-year tax holiday through 2047 for foreign cloud service providers using Indian DC infrastructure (Finance Act 2026, notified March 30, 2026). A 15% cost-plus safe harbor for related-party DC services provides transfer pricing certainty. The finance ministry estimated this will attract USD 50–80 billion in additional FDI by 2035.
| State | Key Incentives | Notable Deals |
|---|---|---|
| Andhra Pradesh | 550 MW state DC framework; AP DCIM | Google, Reliance, Meta |
| Maharashtra | Stamp duty waiver, electricity-duty exemption, green DC park sub-policy | Blackstone $6B MoUs |
| Tamil Nadu | E-duty waiver 3yr, stamp duty concession, dedicated power feeders | STT, NTT, CtrlS |
| Telangana | 100% stamp/reg refund, infrastructure status | AWS $7B, STT $400M |
| Uttar Pradesh | 10yr electricity-duty exemption, capital subsidies | Sify Lucknow AI-Hub |
The Digital Personal Data Protection Act 2023 was enacted in August 2023. The DPDP Rules were notified on November 13, 2025, laying out a phased implementation framework through 2026–27. Together, they form the most consequential regulatory driver for on-shore data center demand.
"Blacklist" approach: Personal data can flow to any country except those specifically restricted by the Central Government. No obligation to provide justification for blacklisting decisions. No GDPR-style Standard Contractual Clauses or adequacy decisions.
Significant Data Fiduciaries (SDFs): Rule 12 mandates that certain categories of personal data (as determined by government) must be processed only in India. This goes beyond the general blacklist framework and creates a hard localisation mandate for sensitive data processed by large platforms.
Extraterritorial reach: Applies to any entity processing digital personal data of Indian residents, regardless of where the entity is based.
Penalties: Up to ₹250 crore per violation.
vs GDPR comparison: Unlike the EU's GDPR, which mandates formal transfer mechanisms (adequacy decisions, SCCs, Binding Corporate Rules), DPDP relies on sovereign discretion and internal accountability. This reduces procedural complexity for Indian organisations but creates regulatory uncertainty for multinational operators who cannot predict which countries will be blacklisted or when.
In 2024, India's data centers accounted for approximately 0.5% of national electricity consumption and roughly 150 billion litres of water use — both figures projected to more than double by 2030 (CEEW, Feb 2026). As India's grid remains ~69% coal-powered (Ember 2025), the carbon math is unfavourable: a 1,700 MW sector running at average PUE of 1.5 consumes roughly 22–24 TWh/year — equivalent to Sri Lanka's entire electricity consumption.
A 1 MW IT load with conventional evaporative cooling consumes approximately 25.5 million litres of water per year (Uptime Institute). At 6.5 GW by 2030, the sector could consume 300–350 billion litres annually — a doubling from current levels. Mumbai, Bengaluru, and Chennai face acute summer-peak water stress. AI workloads compound this: one AI data center could consume up to double the water of a non-AI facility of equivalent capacity.
| Segment | Typical PUE | Notes |
|---|---|---|
| Global hyperscale avg. | 1.1–1.2 | Google: 1.1; Microsoft: 1.12 |
| India colo avg. | 1.4–1.6 | Air-cooled, tropical climate |
| India AI-ready (target) | 1.1–1.3 | Liquid-cooled; CtrlS, Sify, Yotta |
| Edge / micro DC | 1.4+ | DOE: rarely below 1.4 w/o liquid |
| Beijing/Shanghai mandate | <1.25–1.35 | PUE-linked tariff incentives |
Union Minister Dr Jitendra Singh stated on May 22, 2026 that India's data center capacity growth to 6.5 GW by 2030 is expected to generate ~100,000 (1 lakh) engineering jobs in specialised areas including MEP, HVAC, power systems, networking, and facility management. A single 100 MW campus supports 1,000–2,000 jobs during peak construction and 150–300 permanent operational roles.
The talent gap: India has a vast IT workforce (~5.4M) but a thin pool of specialised DC professionals — estimated at fewer than 15,000 certified DC operators/engineers nationwide. Key gaps: critical facility management, MEP engineering, liquid cooling systems, high-voltage power distribution. Key salary benchmarks: DC Operations Manager ₹15–25 lakh ($18–30K) vs $90–120K in US; MEP Engineer ₹10–18 lakh ($12–22K); Facility Technician ₹5–8 lakh ($6–10K). This 5–7× labour cost arbitrage attracts build-operate-transfer models from global platforms.
| Metric | India | Malaysia (Johor) | Indonesia (Batam) | Singapore |
|---|---|---|---|---|
| Build cost / MW | $5–6M | $8–10M | $7–9M | $12–15M |
| Power tariff ($/MWh) | ~$70–90 | $133 | $60 | $178 |
| Power tariff (¢/kWh) | 7–9¢ | 10–13¢ | 6¢ | 17–19¢ |
| Land cost (tier-1 city) | $18–23/sqft | $8–15/sqft | $5–10/sqft | $200+/sqft |
| Operational capacity | ~1,700 MW | ~800 MW | ~200 MW | ~900 MW |
| Pipeline capacity | 4–6.5 GW | 4 GW | ~1.5 GW | Limited |
| Domestic demand | Very large | Moderate | Growing | Large (hub) |
| Data localisation | Strong (DPDP) | Weak | Moderate (GR71) | Moderate |
| Grid reliability | Moderate | Good | Moderate | Excellent |
| Water availability | Stressed | Good | Good | Managed |
| RE share of grid | ~31% | ~12% | ~15% | ~3% |
| Subsea cable gateways | 3 (Mum/Chen/Viz) | 1 (via SG) | 1 (via SG) | Major hub |
| FDI screening | PN3 (China) | Open | Open | Open |
This report draws on a systematic review of over 50 primary sources including company press releases, investor presentations, state government MoU filings, and regulatory submissions published between January 2024 and May 2026.
Market-size estimates were cross-referenced against multiple consultancy bases including CBRE, JLL, Colliers, Mordor Intelligence, Grand View Research, and Acumen Research. Capacity figures represent total installed IT load; power figures represent total draw including cooling and overhead.
The Adani USD 100 billion and Reliance USD 110 billion AI plans are 10-year aspirational commitments. Neither company disclosed how much of these figures represent hard-committed near-term capital as of publication date (TechCrunch, Feb 17, 2026).
The Meta–Sify 500 MW Vizag arrangement is reported by Economic Times (Nov 2025) citing anonymous sources. Neither party had officially confirmed the arrangement at time of writing.
Market-size revenue estimates diverge materially across consultancies (10.5%–15.8% CAGR range). Gravitywell uses a 13–16% consensus range for revenue and 20–23% for installed-capacity CAGR.
Several state MoUs represent signalling investments over 5–10 year timescales and are not fully committed capex at date of announcement.