Power, fibre, flood, title, zoning and water: the six criteria that kill data-centre sites independently, the records that verify each, and how to run the funnel.
Data-centre site selection is not a weighted-average exercise: a parcel that scores brilliantly on five criteria and fails on power is not a strong candidate, it is a failed one. The efficient method tests the binary killers first, in order of how expensive they are to discover late: power feasibility, fibre diversity, flood and seismic posture, title and litigation, zoning and use permissions, and water for the cooling design.
Only parcels surviving all six earn the detailed scoring round, where cost, expansion headroom, construction logistics and incentives differentiate the survivors.
Power: substation capacity, diverse-source feasibility and connection timeline, from TANTRANSCO and TANGEDCO records. Fibre: genuinely diverse duct routes from at least two carriers, walked and mapped, not asserted. Flood: elevation against recorded high-water, drainage catchments and water-body buffers; Chennai sits in seismic Zone III, which design absorbs but insurers price. Title: the full 30-point read, 50-plus-year encumbrance, chain integrity, litigation. Zoning: industrial or appropriate use under the master plan, with data-centre-specific clarity where the state policy provides it. Water: a sustainable source for the cooling design, with TNPCB-compliant discharge.
Each line is a record or a measurement; a week of structured verification typically clears or kills a parcel.
Before parcels are sourced at all, the workload defines geography: interconnection-heavy and content workloads want minimal fibre distance to landing stations and exchange density; enterprise and cloud-region capacity tolerates more distance for better land and power; edge capacity follows population. In Chennai practice, that translates into distinct search areas, the OMR-Siruseri axis, the western industrial belts, and the northern-coastal vector, each with characteristic trade-offs.
Defining the radius first keeps the funnel honest: a spectacular parcel outside the latency budget is a distraction, not an option.
Mature operators run selection as a staged funnel: a long list from the search radius, desktop disqualification on the six criteria, field verification on the survivors, then commercial structuring, price, incentives under the state's data-centre policy, SIPCOT options where relevant, on the final two or three. Documentation discipline matters because the findings become the basis for board approval, lender diligence and the EPC contract's site assumptions.
We run this funnel as a service for operators and investors: the six-criteria verification, graded findings with sources, and the land execution that follows.
Power feasibility, fibre diversity, flood and seismic posture, clean title, compatible zoning and cooling water. Each is independently disqualifying, so test them in that order before detailed scoring.
Indicatively, campus builds run from a few acres for 10-20 MW to 20+ acres for hyperscale phases, driven by hall count, electrical yards, cooling plant and expansion headroom. The envelope should be computed against zoning rules, not assumed.
Chennai lies in seismic Zone III; standard data-centre structural design accommodates it. Flood posture is the more differentiating local risk and is parcel-specific.