India's 100 MW+ Operator Cohort and Why It Needs a Different Insurance Programme
By the end of FY2026 India will host at least 19 operational or under-commissioning hyperscale campuses sized at 100 MW IT load or above, against three such campuses at the start of 2023. The operator cohort breaks into three segments. The first is the global hyperscalers running their own real estate: AWS at Hyderabad and Mumbai, Microsoft Azure at Pune and Hyderabad, Google Cloud at Mumbai and Delhi NCR, Oracle Cloud at Mumbai and Hyderabad. The second is the Indian colocation hyperscalers selling wholesale and retail capacity: CtrlS at Mumbai and Hyderabad, Yotta at Navi Mumbai (NM1 with 50 MW operational and NM2 scaling to 250 MW), Adani Connex at Chennai and Noida, NTT Global Data Centers at Mumbai and Chennai, Sify at Rabale and Siruseri, Nxtra by Airtel at Manesar and Pune. The third is the IndiaAI Mission anchor compute facilities being built under the MeitY common compute programme with operators bidding for the 18,693 GPU initial tranche.
A 100 MW IT-load campus carries a critical-equipment replacement value (servers, GPUs, networking, storage, UPS, switchgear, transformers, generators, cooling plant) of INR 22,000 to 38,000 crore depending on GPU density and tier. A single building inside a campus typically holds INR 4,500 to 8,000 crore of equipment, which exceeds the gross underwriting line of any single Indian non-life insurer. The placements that worked at 20 to 40 MW (one or two Indian leaders fronting, with treaty and selective facultative support) break at 100 MW. The 2026 programme architecture for these accounts now routinely involves a domestic lead (ICICI Lombard, Tata AIG, HDFC Ergo, Bajaj Allianz, or SBI General), a domestic coinsurance panel of three to five insurers, a substantial facultative reinsurance tower placed through GIC Re, Lloyd's, Munich Re, Swiss Re, and Singapore-based reinsurance hubs, and a parallel placement at IFSC Gift City for the international layers.
IRDAI (Reinsurance) Regulations, 2018 read with the 2024 amendments allow GIC Re to cede beyond its retention to category 1 cross-border reinsurers without additional approvals where domestic capacity is exhausted, which materially helps these placements. The IFSC route through IFSCA-registered IIOs has become the preferred channel for the upper layers because of premium remittance ease and capacity from Lloyd's syndicates writing through Lloyd's India. Operators new to the Indian programme are routinely surprised that an account with USD 4 to 5 billion of replaceable equipment requires a 14 to 22 insurer/reinsurer signing panel and a placement cycle of 12 to 16 weeks from broker submission to bound cover.
Liquid Cooling, Direct-to-Chip, and the MEP Loss Exposure That Drives 2026 Premiums
GPU rack densities at 80 to 130 kW have forced wholesale adoption of rear-door heat exchangers (RDHx), direct-to-chip cold-plate cooling (DLC), and selective immersion deployment across India's 2026 hyperscale builds. Mechanical, electrical and plumbing (MEP) plant now represents 32 to 38 percent of total capital cost per megawatt at hyperscale tier, up from 24 to 28 percent in the air-cooled era. The MEP layer is also where most loss events originate.
The four loss patterns underwriters are pricing for in 2026 are coolant distribution unit (CDU) failure, manifold and quick-disconnect coupling weeps, water-side chiller plant breakdown, and electrical fault propagation through busways and switchgear. A single CDU loss in a pod feeding 32 to 72 racks can idle INR 280 to 420 crore of GPU equipment within hours unless emergency cooling restoration is achieved. CDU rebuild lead times from Vertiv, Schneider, Stulz, and Motivair currently run 14 to 26 weeks for production units, which means BI exposure dominates property exposure on these claims. Coupling weeps (CPC Everis, Staubli SPT, Parker FEM) typically cost INR 8 to 25 lakh per event for board replacement on directly affected servers, but the litigation tail when an SLA-protected tenant workload is interrupted runs into far larger numbers.
Water-side chiller failures introduce a property exposure of a different scale. A campus typically operates four to six 2,500 to 4,000 ton-of-refrigeration chillers (York, Carrier, Trane, LG, Daikin) in N+1 or N+2 redundancy, with chilled water headers, plate heat exchangers, and dry coolers in the heat rejection loop. A header burst or a chemical injection error that causes ferrous fouling can take the entire mechanical system offline for 7 to 14 days. The standard Indian Machinery Breakdown wording responds to sudden and unforeseen damage to the chiller plant but does not respond to consequent damage to IT equipment, which falls to Electronic Equipment Insurance. Brokers are increasingly placing manuscript endorsements that bridge the EEI/M&BD interface for cooling-induced IT loss, with the bridge endorsement priced as a separate sub-line.
Fire suppression interactions remain unresolved. Clean agent systems (Novec 1230, FM-200 legacy, IG-541 inert gas) inside DLC halls face elevated false discharge risk from coolant vapour plumes and condensation mist. Refill costs for a 600 kg Novec 1230 bank at current Indian agent prices of INR 1,400 to 1,900 per kilogram run to INR 8 to 11 lakh before recommissioning, and tenants whose workloads are interrupted by a false discharge invoke SLA penalties regardless of insurance coverage. False discharge endorsements are now standard on hyperscale placements at premium rates 0.04 to 0.08 percent of agent system value, but insurers are tightening sub-limits as 2025 loss experience has shown more false discharges than originally modelled.
Underwriting for the MEP layer in 2026 demands engineering survey detail that domestic insurers historically did not produce in-house. Most Indian insurers now contract specialist surveys from Indian engineering consultancies (Cundall India, M Moser, AECOM India), supplemented by international risk engineering reports from FM Global, Liberty Mutual, and Chubb's risk engineering teams routed through their international networks. The cost of a full hyperscale MEP survey runs INR 18 to 35 lakh per building and is now factored into the placement budget.
Lithium UPS and BESS: Thermal Runaway, NFPA 855, and the Underwriting Cliff
Hyperscale UPS has shifted decisively from valve-regulated lead-acid (VRLA) to lithium iron phosphate (LFP) chemistry through 2023 to 2026. The drivers are footprint reduction (LFP UPS modules are roughly 40 percent of the physical volume of equivalent VRLA), cycle life (LFP modules deliver 5,000 to 8,000 deep cycles against 800 to 1,200 for VRLA), and operating temperature tolerance. A 100 MW IT-load campus typically deploys 110 to 130 MW of UPS capacity in LFP modules from Schneider Galaxy VL, Vertiv Trinergy, ABB MegaFlex, Eaton 93PM, or domestic builds from Numeric and Delta. Battery rooms run to INR 280 to 450 crore in equipment value per building.
The property exposure is no longer dominated by the cell chemistry alone. It is dominated by the consequences of thermal runaway propagation across modules and racks in a battery room. LFP chemistry is significantly more thermally stable than NMC (nickel manganese cobalt), with thermal runaway onset temperatures of 230 to 260 degrees Celsius against 150 to 180 degrees for NMC, but an LFP module can still enter thermal runaway through internal short circuit, external heat exposure, or BMS failure. Once runaway begins, propagation across adjacent modules is governed by spacing, cooling, and the presence or absence of barrier construction.
NFPA 855 (the US standard for battery energy storage systems) has become the de facto design reference in India because no equivalent BIS standard exists with the same level of detail. Insurers are conditioning capacity on compliance with NFPA 855 spacing, barrier, and ventilation requirements, supplemented by CEA Measures of Safety and Electricity Supply Regulations 2023 for the electrical side. A battery room that does not segregate UPS modules with 1-hour fire-rated barriers, that exceeds 600 kWh per zone, or that lacks deflagration venting to a safe outdoor location is now routinely uninsurable or quoted at rates 2 to 3 times the conforming benchmark.
Fire suppression design for lithium UPS rooms is contested. Water-based suppression (sprinklers, water mist) is the only suppression chemistry with demonstrated effectiveness against propagating LFP runaway, because the cooling action of water absorbs the runaway heat. Clean agent suppression alone is ineffective because the runaway reaction is self-oxygenating and continues despite oxygen displacement. Insurers are insisting on hybrid systems: pre-action water sprinklers plus deflagration venting plus a clean agent layer for early-stage non-thermal-runaway fires. Operators resist the water systems because of the secondary damage risk to UPS modules during accidental discharge, but the underwriting position has hardened through 2025 to 2026.
The non-physical exposure that matters most is environmental. Thermal runaway events release hydrogen fluoride, phosphorus oxides, and lithium compounds. Cleanup runs through CPCB-authorised hazardous waste handlers under the Hazardous and Other Wastes (Management and Transboundary Movement) Rules, 2016, and on-site decontamination of building structure after a significant runaway can take 8 to 14 weeks. Environmental Impairment Liability cover has become standard on hyperscale placements at premium of INR 35 lakh to INR 1.2 crore per annum depending on capacity and chemistry. Pure-property and BI cover without an EIL layer leaves operators exposed to a meaningful regulatory and third-party liability tail.
Battery energy storage outside the UPS room (grid-scale BESS for demand response or renewable firming, where campuses participate in state DSM markets) introduces a separate underwriting conversation. These installations sized at 20 to 80 MWh are quoted as standalone risks against IRDAI standard fire and special perils, supplemented by EAR cover during commissioning. The crossover between the IT-side UPS and the grid-side BESS is one of the more complex placement questions on India's 2026 hyperscale programmes.
Tenant SLA Liability and the Business Interruption Architecture
Hyperscale operators sell capacity under master service agreements with availability commitments typically structured as 99.99 percent monthly uptime (52 minutes of allowed downtime per month) for Tier IV-design facilities and 99.95 percent (21 minutes per month) for Tier III. Below the SLA threshold, liquidated damages typically run as service credits of 5 to 30 percent of the monthly recurring charge for the affected capacity, escalating with cumulative downtime hours. For the largest enterprise and AI workload customers, contracts increasingly include uncapped consequential loss carve-outs that expose operators to direct damage claims for failed AI training runs, missed financial closing windows, and lost customer revenue.
An AI training run on 1,024 H100 or H200 GPUs over a 14-day window costs INR 28 to 42 crore in raw compute and represents months of model development opportunity cost. A power or cooling event that interrupts the run at hour 240 of 336 can produce a customer claim quantum that dwarfs the operator's monthly recurring revenue from that tenant. The 2024 to 2025 period saw at least three publicly reported hyperscale outage events in India where tenant claims exceeded INR 100 crore per event, although final settlement values were below the claimed amounts after liability caps in master agreements were applied.
The insurance architecture covering this exposure has three layers. The first is the operator's own business interruption cover, structured as gross profit BI tied to the SFSP material damage trigger with indemnity periods of 18 to 24 months and time deductibles of 7 to 14 days. The second is contingent business interruption and dependent business interruption extensions covering tenant-side losses where the trigger is at the operator's premises but the loss accrues to the tenant. The third is third-party liability cover, structured through Commercial General Liability and a specialty Tech Errors and Omissions layer, addressing tenant claims that survive contractual liability caps.
The tenant-liability layer is the most contested part of the 2026 placement. CGL wordings in the Indian market do not contemplate pure economic loss arising from service interruption without underlying bodily injury or property damage. The market response has been to write specialty Tech E&O cover with manuscript wordings, placed predominantly with Lloyd's syndicates and Bermuda-based insurers writing through Indian fronts. Premium rates for the Tech E&O layer run 0.18 to 0.32 percent of declared annual recurring revenue with limits of INR 250 to 750 crore per loss and per annum, with self-insured retentions of INR 5 to 15 crore.
Underwriters are insisting on detailed disclosure of master service agreement liability caps before binding the Tech E&O layer. A standard liability cap of 12 months of recurring charges paid by the tenant is acceptable. Uncapped consequential loss exposure is generally not acceptable and is excluded by endorsement, which means operators offering uncapped liability terms to win large enterprise deals must self-insure that residual layer or buy a more expensive specialty wrap.
A particular issue for the 2026 programmes is the interaction between tenant SLA penalties and BI indemnity. The operator's BI policy compensates for lost gross profit during the indemnity period. SLA service credits paid to tenants represent revenue reduction, which is captured in the BI gross profit formula. However, lump-sum liquidated damages above standard service credits are typically classified as third-party liability rather than first-party BI, and the policy coordination requires explicit endorsement to avoid double payment or double exclusion. Brokers experienced in this segment are now drafting joint loss adjuster appointment clauses across the BI and liability policies to ensure consistent treatment of single-event multi-claim scenarios.
Cyber Physical Risk, OT Network Segmentation, and the Coverage Bridge
A hyperscale facility runs three roughly distinct networks: the IT production network carrying customer workloads, the corporate IT network for operator staff and management, and the operational technology network controlling Building Management Systems, Data Centre Infrastructure Management, power switchgear, UPS, chillers, cooling distribution units, fire alarm panels, and physical access control. The OT network in most 2026 Indian facilities is segmented from the corporate IT network through a demilitarised zone, but the segmentation depth varies significantly by operator, and audit evidence of effective segmentation is the most common underwriting condition imposed before binding cover.
The loss scenarios insurers model follow patterns established by Stuxnet, the 2015 and 2016 Ukraine power grid attacks, the 2021 Oldsmar water treatment incident, and more recent OT incursions at Asian semiconductor and petrochemical facilities. A threat actor with BMS access can manipulate cooling setpoints, force CDU bypass modes, disable pump arrays, or mask alarms. The physical consequence is thermal runaway in GPU pods, with capital exposure as set out earlier. The operator's loss can extend across tenant SLA liability, environmental clean-up, equipment replacement, and reputational damage.
Indian cyber insurance policies sold under the IRDAI (Cyber Insurance) framework and the IRDAI (Information and Cyber Security) Guidelines 2023 respond reliably to data breach, privacy liability, ransomware extortion, and business interruption from IT network events. Physical property damage arising from an OT cyber event sits in a coverage grey zone. The London market's CL380 cyber exclusion pattern, as adopted by most Indian property policies in cyber war and cyber operations clauses, excludes physical damage from cyber events on property policies. The cyber policy excludes physical property damage. The result is a coverage gap that operators usually do not discover until a loss occurs.
The market is bridging this gap through three mechanisms. First, a specific OT write-back endorsement on the property policy that gives back physical damage cover where the trigger is a cyber event on the OT network, subject to network segmentation evidence and BMS audit conditions. Second, a difference-in-conditions placement at IFSC Gift City or Lloyd's that picks up cyber property damage above the operator's primary cyber tower. Third, parametric structures that trigger on defined OT events (BMS access logs showing unauthorised changes, sustained out-of-spec cooling readings) and pay a fixed amount within hours to fund emergency response.
The CERT-In April 2022 directions on cyber incident reporting (six-hour reporting window for prescribed incidents) apply to data centres, and DPDPA 2023 obligations cover any personal data processed within the facility. The IRDAI (Information and Cyber Security) Guidelines 2023 set baseline security requirements for insurers and intermediaries, and similar baselines are now contractually flowed down to data centre operators by financial services tenants. Insurance underwriting now routinely demands evidence of CERT-In and DPDPA compliance posture, MITRE ATT&CK ICS framework mapping, and at minimum a SOC 2 Type II report on the OT environment.
Construction Phase Crossover and the Operational Phase Transition
Hyperscale campuses are built in phases. A 100 MW campus typically commits to a first phase of 12 to 24 MW with subsequent phases of 24 to 36 MW each, layered over a 36 to 60 month delivery window. At any point in the campus life cycle, an operator is running a permanent operations programme on completed buildings while a project programme covers buildings under construction and commissioning. The insurance treatment of these two phases is distinct, and the transition between them is one of the more error-prone moments in the placement cycle.
The construction phase runs under Erection All Risks (EAR) for the MEP, IT, and integration scope, supplemented by Contractors All Risks (CAR) for the civil and architectural scope. Delay in Start Up (DSU) or Advance Loss of Profits (ALOP) covers BI losses arising from delayed handover caused by an EAR or CAR loss event. DSU sums insured are typically set at 12 to 18 months of contracted recurring revenue once the building is operational, calibrated to the contracted go-live date with the anchor tenant. Premium rates for EAR with DSU on hyperscale builds run 0.35 to 0.55 percent of contract value for the 24 to 36 month build window.
The transition to operational cover triggers at substantial completion, which is generally tied to the building's final commissioning sign-off, the certificate of occupancy issued by the local development authority, and the live customer load reaching a defined threshold. The transition involves moving cover from EAR to SFSP plus EEI plus M&BD plus liability, ending DSU and beginning BI, and rebalancing premium across the new policies. Operators have lost cover during this transition where dates were misaligned, where the EAR policy expired before SFSP attached, or where a building was partially in service for testing while still nominally under EAR.
The market practice in 2026 is to run a 30 to 90 day overlap window where both EAR and operational policies are in force on a non-duplicate basis, with the broker maintaining a written hot-cut schedule that defines per-building transition dates. The cost of the overlap is roughly 15 to 25 percent of the corresponding pro-rata premium, and brokers absorb this into the placement budget as a standard line item rather than seeking explicit operator approval per building.
Micro-commissioning loss events during the transition are a recurring problem. A new battery room failing pressure test on commissioning day after EAR has expired but before SFSP coverage on the IT load has fully attached can leave a gap. Similar issues arise with chiller plant commissioning where mechanical pressure test failure causes water damage to floors below. Brokers experienced in this segment now build a commissioning insurance bridge with a 90-day specific cover for first-fill, first-start, and customer load ramp events, placed as a separate facultative line outside the operational programme to avoid contaminating loss experience on the long-term operational policies.
Programme Architecture, Market Capacity, and Broker Placement Approach
A 2026 hyperscale Indian operator insurance programme commonly includes ten distinct policy lines. The first is SFSP for buildings and contents. The second is EEI for IT equipment. The third is M&BD for cooling plant, switchgear, and UPS. The fourth is BI tied to the SFSP material damage trigger with an extended indemnity period. The fifth is CGL with manuscript wording covering hyperscale-specific exposures. The sixth is Tech E&O for tenant SLA liability. The seventh is Cyber Insurance with first-party and third-party limits scaled to tenant data exposure. The eighth is EIL for the lithium UPS, BESS, and fluorinated fluid exposures. The ninth is D&O and Crime cover for the entity. The tenth is EAR with DSU for construction phases.
Domestic capacity on the SFSP and EEI lines is materially constrained. A single Indian non-life insurer's net retention on hyperscale property risk typically caps at INR 200 to 350 crore per location, with the balance ceded through obligatory and facultative reinsurance. The domestic coinsurance panel adds total gross signed line of INR 800 to 1,400 crore, beyond which the placement must access international capacity. GIC Re's mandatory cession (currently 4 percent across most non-life lines) provides one anchor on the reinsurance side, and the remaining capacity is built through facultative submissions to Lloyd's syndicates (notably MS Amlin, Beazley, Hiscox, AXA XL Lloyd's), Munich Re Singapore, Swiss Re Singapore, Allianz Commercial, and selective Bermuda capacity through Indian fronting arrangements.
The IFSC Gift City route has gained share through 2025 and 2026. Operators can place upper-layer cover with IFSCA-registered IIOs and access international reinsurance capacity with reduced friction on premium remittance under FEMA. The IFSC route is particularly used for the Tech E&O layer, the cyber layer, and for the difference-in-conditions wraps over the domestic primary tower. Premium for the IFSC layers is typically denominated in USD, which transfers FX risk to the operator unless hedged through a forward cover at the time of placement.
Broker selection for these programmes has narrowed to a handful of houses with hyperscale capability: Marsh India, Aon India, WTW India, Howden India, and Anand Rathi Insurance Brokers for the upper end, supplemented by domestic specialists for specific lines. The placement cycle from initial submission to bound cover runs 12 to 16 weeks for a renewal and 16 to 24 weeks for a new build at first operational year. The broker's role has moved beyond placement to risk architecture, with the broker leading the engineering survey commissioning, the policy wording harmonisation across the ten lines, and the certificate of insurance distribution to tenants under master service agreement obligations.
A practical observation for new entrants to this segment: the certificate of insurance flow is more operationally demanding than the placement itself. A 100 MW campus serving 40 to 70 enterprise tenants must issue compliant certificates to each tenant, with each tenant's procurement team imposing specific wording, named additional insured, and waiver of subrogation requirements that vary by tenant. A dedicated certificate management workflow with API issuance, version control, and audit trail is now standard at hyperscale broker shops.