It’s cool, calm, collecting: Why data centres won’t cripple consumer availability of water and power

GoI's thrilled. The world's biggest tech companies are pouring humungous sums into Indian data centres. Amazon, the biggest, plans to invest $35 bn over 5 yrs. Google has committed $15 bn, plus partnership with Indian conglomerates. Meta is building a major facility near Google's proposed sites. Microsoft has announced $17.5 bn in AI-related investments.

These investments represent more than warehouses full of computing capacity. India has made data sovereignty a central objective of digital policy. It says data generated in India must be stored in India and not in distant overseas facilities. That is impossible without enormous domestic computing and storage capacity. Massive data centres are, therefore, essential for India's national security.

They are also the indispensable foundation for AI. India cannot afford to be left behind in the AI race. AI systems require immense computing power that only hyperscale data centres can provide. That is why the entry of hyperscalers - top AI companies in the world with enormous scale - is both necessary and mutually beneficial.


Every massive new technology invites a backlash. Environmentalists warn that data centres are voracious consumers of electricity and water. They fear the proliferation of data centres, whether owned by Indians or hyperscalers, could create critical shortages of power and drinking water.

Last week, similar anxieties surfaced in the US, where demonstrations against new data centres were held in several places. Protesters complained that these facilities consume too much electricity, guzzle scarce water, and cover vast stretches of land with unattractive industrial buildings.

These concerns are not irrelevant, but are greatly exaggerated. They represent the 'Not in my backyard' (Nimby) syndrome that has greeted every major infrastructure project in the US, from airports and highways to power plants and wind farms. Every locality wants the benefits of new technology, but wants it located somewhere else. The relevant question is not whether data centres consume resources - they do - but whether those costs are manageable with suitable technology and adaptations. The answer is yes.

The choice is not between data centres and the environment. We can have both. Hyperscalers are steadily adopting new technologies to overcome existing problems. Google aims for net-zero water consumption in data centres by 2030. Others are moving in the same direction.

The strategic stakes are enormous. AI is a revolutionary technology set to match or exceed other great turning points of history like the steam engine, spinning and weaving machine, automobile, electricity, computer, internet, and smartphone. Each of these in its time displaced other technologies and caused bankruptcies and unemployment. But, in time, they revolutionised production, income and living standards. We owe today's prosperity to those revolutions.

Many of the earlier turning points caused immense environmental damage, too. The Industrial Revolution's 'great satanic mills' converted London to a city of fog and respiratory disease. Automobile and power plants are major polluters even today.

So, it should be no surprise that AI and data centres, too, have environmental implications. The answer is not to avoid or slow down adoption of AI, but to have regulations that make its environmental side effects, as well as its security and privacy challenges, manageable.

Early data centres used evaporative technologies to cool the immense heat generated by data centres. This was a simple technology, but as the scale increased, concerns arose about their impact on water availability. Today, new technologies with airflow management and highly efficient heat exchangers have dramatically reduced consumption of water.

AI itself is helping optimise cooling by continuously adjusting temperatures, fan speeds and workloads to minimise energy and water consumption. Google's DeepMind demonstrated years ago that AI could cut cooling energy by around 40% in some facilities. Since then, cooling technologies have cut water consumption by up to 80%, and some companies even talk of technologies that become net producers of drinking water, by making seawater and sewage potable.

One breakthrough is in liquid cooling. Instead of cooling entire rooms with chilled air, specially designed liquids circulate directly around processors, or even immerse servers completely in non-conductive fluids. Because liquids transfer heat far more efficiently than air, they require much less overall cooling. This simultaneously reduces electricity and water use.

Fears that data centres will use scarce electricity at the expense of consumers are wrong. Data centres need electricity 24x7 and so cannot rely on grid electricity, as consumers do. Data centres will have almost universally dedicated captive sources of power, plus storage back-ups. Consumers need not worry that their grid supply will be reduced.

In the US, some neighbourhoods resist data centres saying they are large eyesores. Really? Data centres are simply buildings filled with computing power. They don't handle dirty materials like iron ore or coal as do steel plants, power plants and other industries that take just as much space.

In sum, let's welcome data centres. They represent our future. Their side effects are real. But new technology can manage that.