Will Demand Keep Pace with India’s 7.1 GW Data Centre Pipeline?

by Pratik Mundhada, Director – Corporate Ratings, India Ratings and Research (Ind-Ra)

India’s data centre (DC) sector is in the midst of huge capacity buildout with announced non-captive capacity pipeline of around 7.1 GW, majorly sponsored by domestic and global institutional capital. As per Ind-Ra’s project-wise analysis, total installed DC capacity in India currently stood at 1.3 GW of IT load. The pipeline therefore represents roughly a 6.3x step-up from current levels. Of the total 7.1GW announced capacity, near about 4.3GW is still in early stage of discussion and hence, the supply phasing becomes an important element of the supply-demand equation. For business economics to succeed, the huge supply should be timely matched with adequate demand visibility, resulting in consistent absorption rate while maintaining the returns and credit metrics.

Ind-Ra assesses the demand story for Indian DCs is pillared on two distinct segments, viz, consumer-facing (B2C) and enterprise (B2B), each with independent growth vectors but create a structurally resilient and diversified demand base. Moreover, the strong policy tailwinds including India AI mission and government-led digitalization initiatives, as well as the possible relocation of DC to India due to ongoing West Asia conflict, further supports the overall demand dynamics.

The Structural Demand Drivers
Ind-Ra’s analysis of B2C increased data demand is anchored on two metrics. Firstly, as per industry estimates, mobile data consumption estimated at around 29 exabytes (EB) per month in CY25, which is projected to grow at a 14% CAGR to reach 58 EB per month by CY30. This growth is expected to be driven by increased 5G adoption and higher data processing requirement per user for video, real-time applications and AI-native applications, compared to legacy applications. Secondly, India’s non-mobility demand is supported by a broadband subscribers’ base of around 64.4 million as of April 2026. Even with a conservative estimate of 50% of developed nations’ broadband penetration level, the subscriber base is estimated to increases to 100+ million subscribers by CY30 which will be supported by continued investment in 5G fixed wireless access (FWA) by India telcos.
On the B2B side, the India’s public cloud market is estimated to grow by an estimated CAGR of 24% over the CY24-27 (Source: ICRIER) with majority of demand expected to be catered by large hyperscalers. Large government demand through NIC, MeghRaj, Aadhaar, UPI and the Common Services Centre network, aimed towards government record digitalization and achieving data sovereignty, is expected to generate incremental enterprise and hyperscaler demand for DCs.

The West Asia Conflict: A New Dimension
While long-term demand drivers were already in place, including fibre connectivity, power availability and regulatory tailwinds; recent geopolitical developments in West Asia have introduced a new dimension of physical security consideration for the international DC developers and hyperscalers. This may have prompted a few hyperscalers to explore workload rerouting out of Middle East and to evaluate India as geographic diversifier given the established sub-sea cable connectivity with Middle East countries. However, we believe that emergency workload rerouting does not guarantee a permanent migration and conversion into long-term offtake agreement. Nevertheless, West Asia led disruption has substantiated India’s position as disaster recovery hub for Middle East’s enterprises and hyperscalers.

AI Workloads: Real, But Non-Linear
Global DC capacity is projected to grow to around 230 GW by CY30 as per industry estimates, with generative AI workloads growing at approximately 40% CAGR compared to 15% for the non-AI workload. In India, the AI adoption for most enterprises is still in early stage of short-term and functional deployment, with transformative and integrated cohorts been largely led by only few large entities. The ability of Indian enterprises to scale AI use cases and keep pace with global adoption remains critical for India to fully capitalise on the expanding AI ecosystem.
The non-linearity in AI computing requirements also stems from the transition of language models from training phase to inference phase. Inference phase typically entails moderate computing intensity along with the increased low-latency processing requirements, which could lead to more distributed and localised deployment of AI workloads. However, as per industry experts, despite the low computation intensity, the overall AI computing demand in India is unlikely to decline. This is driven by significant scaling of inference cases in India and policy tailwinds including India AI Mission, which had allocated a INR100 billion corpus to provide graphic processing units (GPU) compute capacity to Indian language model developers.

Calibrated Execution Critical to keep Vacancy in Check
Ind-Ra estimates the pace of DC supply addition in India to increase to 300–350MW per year over FY26–28 (around 20% CAGR), compared to 150–250MW per year over FY22–25. Given the early stage of India’s DC demand cycle, Ind-Ra believes that the market remains susceptible to interim volatility in absorption rates. The volatility is further accentuated by emerging AI computing demand layer, which introduces non-linearity in capacity absorption through large hyperscaler-led deployments and potential spikes in offtake, similar to what was witnessed in the USA during cloud adoption phase over CY18-23. As a result, the demand trajectory could exhibit both sharper upside during periods of strong commissioning and downside risks in the interim. Hence, a cautious development approach such as aligning major MEP capex closer to leasing visibility or securing soft pre-commitments from hyperscalers, large enterprises, or government entities, becomes critically important for project economics.

Return Sustainability and AI Hardware Obsolescence Risk
While the demand for India DC remains structurally intact, the unit economics and consequently, return on capital employed (ROCE) remain critical to assess the long-term viability of the DC projects in India. As per Ind-Ra estimate, a pure-play colocation DC requires a capex of INR500-700 million per MW (excluding land cost) and generates an annual offtake revenue of INR75-85 million from hyperscalers, implying an offtake yield of around 14-15%. With EBIT margin of around 40%, the key challenge lies in sustaining the ROCE beyond 5-6%. Hyperscalers, who remained a primary offtakers for India colocation DC, negotiates for lower offtake rates in exchange for large and long-term capacity commitments and thereby, offer good revenue visibility. On the other hand, retail and enterprise customers (<1 MW capacity offtake) offer higher rental yield but slows the absorption rate of the project. Hence, achieving the optimum tenant mix between hyperscaler, and retail or enterprise customer remains critical for achieving better ROCE. The ROCE also remains hinged on the hyperscaler offtake rates hike, in line with other developed nation. The AI layer further complicates the return profile due to rapid advancement in GPU technology, increasing the risk emanating from the technology obsolescence of existing GPUs.

Ind-Ra Outlook
In Ind-Ra’s base case, Indian DC demand is expected to grow in line with the structural drivers including increasing mobile data volume, rising broadband penetration, enterprise cloud migration, government record digitalisation and AI workload emergence among others, and thereby supporting absorption of the supply pipeline over the upcoming years. However, a calibrated supply addition in the 2026-2029 transition window aligned with offtake visibility remains key to sustaining healthy absorption rates. While rental stability is not a key concern, the ability of the Indian DC market to achieve rental escalations in line with developed markets remains critical to improving the overall ROCE profile.