How much water does it actually cost to ask an AI model a single question - and who is paying that bill, the person typing or the town next to the server farm?

The Case for Guardrails Now, Not After the Damage

India needs a binding, responsible data-centre regulatory framework before hyperscaler expansion outpaces the state's ability to manage its consequences - not after the country's power grids, water tables and local communities have already absorbed the cost.

The reasoning starts with what has just happened elsewhere. On July 14, New York became the first US state to impose a moratorium on new hyperscale data centres, pausing permits while regulators build standards for energy demand, water use and grid impact.

Massachusetts, Amsterdam, Singapore and Ireland have each taken similar steps after watching data centres strain local infrastructure without generating proportional local benefit - these are not labour-intensive facilities; once operational, they run on a small, highly skilled workforce, with most other roles limited to housekeeping and security.

India is moving in exactly the opposite direction. State policies in Uttar Pradesh, Tamil Nadu and Telangana offer capital subsidies, stamp-duty exemptions and fast-track single-window clearances to attract hyperscaler investment, with essentially no equivalent framework governing how much water and power these facilities can draw from communities that are, in many cases, already water-stressed.

A reported case of one UP data centre coinciding with a steep local groundwater decline illustrates the kind of cost that shows up only after the incentive has already been granted. The strongest objection to this position is that India cannot afford caution right now.

The country hosts a disproportionately large share of the world's data relative to its installed data-centre capacity and closing that gap is treated as a matter of digital sovereignty and economic opportunity - every month spent building an assessment framework is a month hyperscaler investment could go instead to Southeast Asia or the Gulf, taking jobs, tax revenue and strategic compute capacity with it.

Given that AI's return on investment for business processes is itself still unproven, some would argue this is precisely the wrong moment to add friction to infrastructure India is racing to build. That objection assumes regulation and growth are opposites, but the international examples cited above don't actually support that assumption.

New York's moratorium is not a ban - it is a pause to build standards, after which permitting resumes under clearer rules. Singapore didn't exit the data-centre business after its 2019-2022 freeze; it emerged with a quota system precise enough to keep growing sustainably.

The choice India actually faces is not "regulate and lose investment" versus "don't regulate and keep it"

• it is "build the framework now, on India's own terms" versus "build it later, in reaction to a crisis, on terms dictated by whichever community or court forces the issue first." Investment capital that is genuinely long-term - the kind India wants - tends to prefer predictable rules over an unregulated market that could be shut down overnight by public backlash, exactly as happened in New York after $130 billion in projects had already been blocked or delayed by local opposition. Guardrails now are what make sustained, rather than reversible, growth possible.