India's data centres are on track to drink what 1.3 billion people use at home

Manzar's most operational detail is a vocabulary problem. He told Rest of World that "data center" is not a term the affected neighbourhoods recognise, so there is little awareness of a project until construction or clearance activity starts and the machinery is visible from the doorstep ^[10].

Siting decks assume the community trades land for payroll, and factories mostly honoured that trade, employing a lot of people even where land was taken ^[12]. Manzar says the data-centre version breaks it, because the operators themselves are not sure local people can be hired at all at an automated system ^[11]. Take the payroll out and the acquisition has nothing local left in it to pay with, which is why his complaint is about standing rather than price. He says he is not a stakeholder in what his land becomes, and the answer to what he gets back is a promise of jobs ^[24].

Divide the two projections Rest of World cites and the water number covers twice the population the power number does ^[20]. Power is the figure that turns up in supply agreements and grid planning, while water is taken from wherever the building sits.

The demand side has moved fast enough to explain the government's enthusiasm. When Manzar started the Digital Empowerment Foundation in 2002, 1.2% of Indians were online, and the share is now above 70% ^[3][4], a rise of roughly 58 times ^[21]. That growth gave households a service, whereas this phase asks particular districts to supply inputs, which is the distinction behind his line about developing countries becoming "data waste centers, serving the elite of the world" and his condition that an operator arriving in a neighbourhood serve it first ^[19].

This is one interview, edited for length, and no response from the companies or from any government appears in it ^[22]. Manzar has served on Indian government and international committees dealing with access, digital inclusion and infrastructure, so the position is a long-held one ^[5]. His alternative, splitting the country into as many as 250,000 village-council data centres or district-level facilities serving only their own districts, comes with no cost figure and no capacity figure ^[18][23].

Two tests do more work here than an impact binder. First, whether the district supplying the land and water consumes any of the capacity it hosts. Second, whether a compensation mechanism pays out even when a title turns out to be contested. Both yes is the only combination that survives a village council meeting. Local output with no compensation buys goodwill until the first bad title. Compensation with exported output settles the landowners and nobody else in the catchment. Neither, which is what Manzar describes at the sites he names, gives you a protest calendar keyed to your own construction schedule ^[13].