1,575 MW and 610 Million Sq Ft: Property Underwriting When the Building Is the Cheapest Part of the Claim

India's data centre capacity has grown from about 375 MW in 2020 to 1,575 MW, with projections above 4 GW by 2030, according to Asia Insurance Post reporting of 29 August 2026 on India's new economy assets. Warehousing has scaled on a similar curve: over 549 million sq ft across eight major markets and more than 610 million sq ft nationwide.

Those two numbers describe a property portfolio that did not exist in its current form when most Indian corporate property programmes were first structured. A 2015 warehouse placement covered a shed, some racking and stock. A 2026 logistics park placement covers automated racking, sortation lines, a fire protection system engineered around high-bay storage, and a tenant base whose fulfilment commitments depend on the site staying open. A data centre placement covers a shell whose replacement cost is a minority of total insured value, sitting under plant that carries the rest.

The property market is repricing both asset classes, and not because these buildings burn more often. When something goes wrong, the physical repair bill is the part everyone can estimate, and it is usually the smallest part of the claim. Operators who explain that loss shape get better terms than operators who send a valuation certificate and a fire NOC and leave the underwriter to assume the worst version of it.

Amit Agarwal, CEO of Howden India, put the problem directly in the same 29 August 2026 coverage: the property insurance market is changing rapidly because these assets form the crux of digital services, e-commerce, modern manufacturing, healthcare and global supply chains, and underwriters must assess operational dependencies, asset concentrations, cyber-physical convergence and business interruption, because repairing a damaged physical structure is often the easiest part of a claim.

Take one event at a colocation facility. A switchgear fault takes out a power train. The material damage element is the switchgear, the cabling, some smoke and heat damage next door, and the making good. That is a bounded, surveyable number. What follows is not:

• Racks in the affected hall go dark, and every tenant in them triggers a service level claim against the operator.
• Tenants whose workloads cannot fail over migrate to another provider, and some do not come back after the site is restored.
• Replacement switchgear and transformers are ordered against a global order book, so the restoration timeline is set by an OEM production slot, not the contractor's site programme.
• The operator's customers, running payment, healthcare or e-commerce workloads, suffer losses that come back as liability notices.

The warehouse version is a high-bay fire. The steel portal frame can be re-erected and the stock is insured. What takes longer is rebuilding the automated storage and retrieval equipment, requalifying the fire protection design for the new racking, and holding the customer contracts through the gap. Our note on underwriting automated storage and retrieval warehouses covers how high-bay fire behaviour drives that timeline.

The underwriting consequence is straightforward. On these assets, the business interruption section is the policy.

Most Indian property declarations are still assembled the way a builder would assemble them: civil works at replacement cost, plant and machinery at gross book value or a valuer's estimate, stock at average holding. That suits a factory where the building and the machines inside it are comparable in value. It breaks on a data centre and strains on an automated warehouse.

On a data centre, the shell and core is a modest share of total insured value. The rest sits in transformers, switchgear, UPS systems, battery rooms, generators, chillers, computer room air handling units, busways, fire detection and suppression, building management systems and structured cabling. Much of that plant is import-dependent, and a declared value fixed at the invoice cost of equipment bought three years ago will not buy the same equipment today once currency movement, freight, installation, commissioning and professional fees are added.

On a warehouse the same gap appears in the handling equipment, the racking and the fire protection system. In-rack sprinklers, ESFR heads and the pump house are engineered around a specific storage configuration, so reinstating them is an engineering project, not a purchase.

How to build the declaration instead

1. Value the shell and core separately from the mechanical and electrical fitout and state the split. Underwriters price the two differently and will not guess in your favour.
2. Use reinstatement value for plant rather than book value, refreshed annually against current OEM quotations rather than by indexing last year's figure.
3. Include the soft costs a total loss incurs: debris removal, professional fees, expediting expenses, and re-commissioning and re-certifying the facility.
4. State who insures tenant-owned servers, and the currency in which imported plant would be replaced. A rupee-denominated declared value against a euro-denominated chiller is a currency exposure sitting inside a property insurance programme.

Operational dependency is why the loss travels further than the damage. The underwriter needs a map of what stops when a given asset stops, and only the operator can produce it. Build it at the level of the single point of failure, not the site. For a data centre the useful units are the power train, the cooling loop, the meet-me room and the fibre entry path. For a logistics park they are the sortation line, the dock bank, the racking aisle, the refrigeration plant and yard access. For each, state what share of revenue-generating capacity depends on it and whether the redundancy is concurrently maintainable or only failover.

Redundancy is not independence, and this is where most maps are optimistic. A facility with N+1 chillers still has one condenser water loop and often one plant room. A site with two utility feeds may take both from the same substation. Two fibre entries leaving the plot on the same side of the road are one fibre entry with extra cost. An operator who has done the walk-down and disclosed the shared elements prices better than one who claims full redundancy and is found out at survey.

Asset concentration inside the portfolio

Concentration works on two axes. The physical axis is how much capacity sits inside one fire compartment, one building or one campus: a 60 MW campus built as three 20 MW buildings on one plot is a different accumulation from three 20 MW buildings in three cities, and should be presented with separation distances and compartmentation stated. The contractual axis is how much of the revenue base sits with a few tenants.

Present both with the mitigation attached. "Largest hall 12 MW, two-hour fire wall, independent power and cooling trains, top three tenants 41 percent of contracted revenue across four halls" gives the underwriter something to price. A single site value and a revenue figure asks to be priced at the pessimistic end.

Contingent business interruption is where these programmes most often fall short, because the dependency runs in both directions and most placements cover only one.

Downstream dependency is exposure to the operator's own customers. A colocation operator loses revenue when a tenant terminates after an outage; a 3PL operator loses it when the brand it fulfils for moves volume elsewhere. This is loss arising from damage at the operator's own premises, so it belongs in the main consequential loss section, but only if the gross profit definition and the indemnity period were set with customer churn in mind rather than physical restoration alone.

Upstream dependency is exposure to damage at someone else's premises, and that list is longer than most submissions admit:

• The utility substation and transmission path feeding the site, where a supplier extension is needed because the damage is not at the insured premises.
• The fibre carrier's landing station, aggregation point or duct route, and the OEM or its Indian service partner whose spares and engineers restore the plant.
• For a warehouse, the anchor customer's factory, whose shutdown empties the shed whatever the shed's condition.

Each needs a named supplier extension or an unnamed supplier grant, with a sub-limit sized to the dependency rather than the token figure that often appears by default. The evidence for a larger sub-limit is the dependency map plus the contract terms showing what the operator owes when the dependency fails.

Tenant concentration sets the size of this exposure. A site where the top tenant is 8 percent of revenue can absorb a churn event. A site where the top tenant is 45 percent, under a service level agreement allowing termination for cause after a defined number of availability breaches, carries an exposure that no material damage valuation will reveal. Our treatment of data centre business interruption wordings sets out how service level credit pass-through is handled at renewal.

Cyber-physical convergence is the fourth item on Howden's list, and it opens the cleanest gap between two policies that both look like they respond.

The building management system, the power monitoring system, the chiller controls, access control and the warehouse execution system are all networked. A control-layer compromise can trip cooling, open a switch or halt a sortation line with no physical peril occurring. The property policy asks whether there was damage by an insured peril; the cyber policy asks whether the event was a security failure. An outage caused through the controls, with no fire and no equipment damage, can fall between the two.

Close the boundary at placement rather than at claim stage. Ask both markets to confirm in writing how a control-system outage with no physical damage responds, and align the wordings so one picks it up. Where the property programme carries a cyber exclusion, check whether it excludes only malicious acts or non-malicious system failure too, because the second version is wider than most buyers assume.

The theft exposure has moved indoors

The Munich Re Specialty Cargo Theft Tactics and Trends Report 2026, published on 22 June 2026, found that 50 percent of Asian cargo thefts targeted facilities, with warehouses and facilities overtaking roads as the primary theft location for the first time, and that 22 percent of thefts globally involved insider participation.

Access control logs, CCTV retention, dual authorisation on despatch and the screening standard for warehouse and contractor staff are underwriting evidence now, not housekeeping. Our overview of logistics and warehousing underwriting in India covers how they feed into pricing.

The most common structural error in these placements is a twelve month indemnity period chosen because that is what the last policy said. On a data centre or an automated warehouse, twelve months routinely expires before the site is back to pre-loss revenue. The period has to cover four sequential phases, and they belong in the submission as a timeline rather than an assertion:

1. Assessment and clearance. Investigation, surveyor appointment, statutory clearances, insurer agreement on scope and debris removal.
2. Procurement. The binding constraint on both asset classes. Transformers, HT switchgear, UPS modules, chillers and automated racking are made to order against an OEM queue. The evidence is a current written lead time quotation from each OEM for the model installed, not a market estimate.
3. Installation, commissioning and certification. Plant on site is not plant in service. Factory and site acceptance testing, integrated systems testing, statutory electrical inspection and fire clearance sit between delivery and load.
4. Revenue recovery. Tenants who moved out during the outage return on their own schedule, if at all. Recovery to pre-loss revenue lags recovery of physical capability, and the period must run through that lag rather than stopping at re-commissioning.

An operator who submits OEM lead time letters, a restoration programme showing the four phases with dates, and a churn assumption drawn from the contract termination clauses will get a longer period agreed with less argument than one who simply asks for twenty four months, and will negotiate a better rate for it. Set the period against the longest-lead item in the facility rather than the average: restoration finishes when the last critical component arrives, and one transformer can set the timeline.

The gap between a well-received and a poorly-received submission is documentation the operator already holds and does not send: a declared value schedule split by asset class with each valuation basis dated, a single point of failure register showing shared elements and dependent revenue, a concentration statement by fire compartment and by tenant share of contracted revenue, written OEM lead time confirmations, a dependency schedule mapped to the extensions requested, and the security position on the OT and physical sides.

None of that requires a consultant. It requires operations, commercial and the broker to assemble what each already knows into one document before the market sees the risk. These assets are being repriced because the loss shape changed and the submissions have not caught up. An operator who documents dependency, concentration, the cyber-physical boundary and the real restoration timeline is telling the underwriter which parts of the pessimistic assumption do not apply here. That is where the terms are won on an asset whose building is the cheapest part of the claim.