Quick glance
India is discovering a conflict that has been raging in America and Europe over the last year.
A country without serious compute infrastructure risks becoming a consumer rather than a producer of the AI revolution.
The question is not whether to build them. It is how to build them without making host communities feel that everyone else gets the benefit while they get the bill.
Good regulation should not obstruct data centres; it should make solutions possible by encouraging better technology, smarter siting and a fairer social contract.
Perhaps we should think of major data centres less as private warehouses full of computers and more as quasi-public utilities.
The AI Revolution Has a Neighbourhood Problem
For years, ‘the cloud’ was one of technology’s smartest metaphors. It made the infrastructure behind our digital lives seem gentle, fluffy, and weightless.
But the cloud has always had an address. Increasingly, people living at that address are asking questions.
In Visakhapatnam, Google’s $15 billion AI hub is being built across multiple sites with Adani-linked partners, with gigawatt-scale compute at its heart. Around Tarluvada, villagers have alleged that land they had cultivated for years was taken back without adequate consultation or promised benefits. At Adavivaram and Mudasarlova, activists have challenged tree-clearing and the project’s proximity to the Kambalakonda Wildlife Sanctuary.
In Thane, residents around Amazon’s proposed 53-acre hyperscale data centre are protesting its location amid homes, schools and hospitals, citing concerns about electricity, water, heat and noise.
India is discovering a conflict that has been raging in America and Europe over the last year. In Edgerton, Kansas, a town of about 2,000 people has been divided by a proposed data centre, with residents seeking a ballot initiative and the city going to court against them.
Finland offers an even starker warning. Google has apologised after clearing about 330 hectares of forest in Muhos and just under 200 hectares in Kajaani while environmental assessments were incomplete. This has forced regulators to halt the work.
It would be easy to make this another story about Big Tech versus helpless communities. While we need that, we need something more: a way out, a balance between the two sides.
We need data centres. They run far more than ChatGPT. Search, banking, streaming, social networks, cloud software, government services, hospitals and almost every company’s digital operations depend on them. AI simply raises the requirement dramatically.
Read more from Dharma of AI
Google says its Vizag hub will provide gigawatt-scale compute and new subsea connectivity for India. The IEA says electricity use by AI-focused data centres jumped 50% in 2025. A country without serious compute infrastructure risks becoming a consumer rather than a producer of the AI revolution.
So, the question is not whether to build them. It is how to build them without making host communities feel that everyone else gets the benefit while they get the bill.
There are four parts to the answer.
The first is technology. Data centres consume enormous power, and some cooling designs consume substantial water. But those numbers are not immutable laws of physics.
Microsoft’s newest designs use closed-loop, direct-to-chip cooling that requires no water evaporation during normal operation. Recycled wastewater, better chips, smaller efficient models, smarter workload scheduling and renewable power can all reduce the footprint. The objective should be simple: squeeze more intelligence out of every litre, watt and acre.
The second is geography. Do all these facilities need to sit beside dense urban populations?
China’s ‘East Data, West Computing’ programme deliberately places much of its capacity in western regions rich in wind and solar energy while serving demand concentrated in the east. Five of its eight national computing hubs are in western China.
India will need low-latency facilities near cities for some workloads, but model training, backup capacity and less time-sensitive computing can often be located where land, clean energy and water are more abundant. Data centres should follow resources where possible, rather than forcing resources to follow data centres.
Read more from Dharma of AI
The third, and perhaps hardest, is the social contract. Technology companies learnt from platforms that enormous value can be captured centrally while users receive only a slice. That model will not work when the ‘platform’ occupies hundreds of acres in someone’s neighbourhood.
Communities need to share visibly in the upside through jobs, local infrastructure, reliable power, water restoration, taxes and perhaps direct community funds.
The fourth is regulation, and it needs to arrive before the damage rather than after it.
We have seen the opposite movie with social media: governments are now scrambling to regulate privacy, addictive design, misinformation, child safety and platform accountability only after these systems have reshaped society. We should not repeat that sequence with AI infrastructure.
Governments, technologists, companies and civil society need to sit together now and establish clear rules around land use, water, power, environmental impact, community benefit and transparency. Good regulation should not obstruct data centres; it should make the other three solutions possible by encouraging better technology, smarter siting and a fairer social contract.
The objective is not regulation versus innovation, but regulation that allows innovation to retain public consent. India has the opportunity to take a lead in this and make it an example for the rest of the world.
There are useful examples. Google’s older data centre in Hamina, Finland, now feeds waste heat into the local district-heating network, enough to meet a large share of the town’s heating needs. In America, some developers are proposing direct payments and community benefit packages.
These cannot become bribes for bad projects, but they point towards the right principle: if a community hosts critical digital infrastructure, it should participate materially in the value created by it.
The Industrial Revolution learnt this painfully. Factories initially privatised profits while socialising smoke, congestion and polluted rivers. Eventually, societies demanded zoning, regulation and environmental standards.
Economists call the underlying problem the tragedy of the commons: when everyone draws from shared resources, nobody has sufficient incentive to protect them.
AI infrastructure needs to avoid repeating that history.
Perhaps we should think of major data centres less as private warehouses full of computers and more as quasi-public utilities. They can be privately owned and commercially operated, but their relationship with electricity, water, land and surrounding communities makes them part of the civic fabric.
The next great constraint on AI may therefore not be chips, models or money. It may be consent.
And consent is not won by telling communities to get out of the way of progress. It is earned by designing technology, locations and economics so that progress belongs to them too.
The cloud may be virtual. Its neighbourhood is not.
— Jaspreet Bindra
See you next week!
