We Cannot Drink DATA: The AI Boom Hits the Water Table

For two years the AI race has been fought in the abstract — models, chips, capital, and now containment failures. This week it hit something solid: a hillside in southern India, stripped to red earth, and the people downstream of it who say they were never asked. Google's largest AI hub outside the United States has become the clearest sign yet that the next limit on artificial intelligence may not be technological at all.

What Google Is Building

Google is spending 15 billion dollars over five years to build an AI data-centre hub in Visakhapatnam, in the southern Indian state of Andhra Pradesh. Google Cloud's chief executive Thomas Kurian has called it the company's largest AI hub anywhere outside the United States. The campus will open with one gigawatt of capacity and is designed to scale to several. It is being built in partnership with the group of Indian billionaire Gautam Adani, is projected to create up to 188,000 jobs, and sits inside Google's roughly 85-billion-dollar data-centre spending commitment for this year alone. In the language of the state's IT minister, in an age where data is the new oil, this is a strategic advantage.

That framing — data as oil, the campus as national triumph — is exactly the story the project's backers want told. This week a different story pushed its way into view.

Where the Resistance Comes From

When Reuters visited the site, earthmovers were still working the exposed ground, and opposition was mounting around it. The project faces several legal challenges in the Andhra Pradesh High Court, centred on two concrete fears: what a gigawatt-scale, water-hungry facility will do to local water supplies, and its proximity to the Kambalakonda Wildlife Sanctuary — home to leopards and pangolins — which sits just 860 metres away. Residents worry about heavy construction and noise pressing against protected habitat.

The protest has a face, and it is a pointed one. In recent weeks, activists and children have marched through Visakhapatnam holding banners reading "We cannot drink DATA," and painting handcuffs onto the Google logo. The state government, led by an ally of Prime Minister Modi, has hailed the project as historic and transformational, and denies it was fast-tracked without weighing the risks; it says the site is farther from the sanctuary than the law requires. Google says the campus will comply with all applicable law and will use advanced air cooling to protect local water resources, and promises to be a quiet, unobtrusive neighbour.

The Constraint Nobody Priced In

Step back from the specific dispute and the larger point comes into focus. The entire conversation about AI's limits has been about supply of the digital kind — enough chips, enough memory, enough model capability. This is a different kind of ceiling. A one-gigawatt data centre is a small city's worth of electricity and, depending on its cooling, an enormous draw on water. Multiply that by the hundreds of billions being poured into data-centre construction worldwide, and the boom stops being an abstraction floating in the cloud. It becomes land taken, rivers tapped, and grids strained, in specific places, next to specific people who can vote, sue, and march.

That is the shift worth naming. The scarce resource gating AI's expansion is quietly moving from silicon to something older and more physical: water, power, land, and the consent of the communities that hold them. Chips can be manufactured. A watershed cannot. Google's choice of advanced air cooling over water cooling in a water-stressed region is itself a tell — the physical constraints are now shaping the engineering, not just the PR. The cloud, it turns out, is a thirsty thing, and it has to touch the ground somewhere.

Claim and Counter-Claim

The case for the project is substantial and should not be waved away. Fifteen billion dollars and up to 188,000 jobs is a genuine development opportunity for the region, and Andhra Pradesh's government was democratically elected to make exactly these bets. Google's commitment to air cooling and legal compliance is a real, if partial, answer to the water concern. Demand for AI compute is soaring, that compute has to live somewhere physical, and India offers scale, ambition, and political will that more constrained regions cannot. From this angle, the opposition is the friction any large infrastructure project generates, and the benefits outweigh it.

The counter-case is equally grounded. A gigawatt-scale facility beside a wildlife sanctuary and atop a contested water supply is not a neutral act, and "farther than the law requires" is a low bar when the law predates data centres of this size. The promise of jobs is real but often front-loaded into construction and thin thereafter; the water draw is permanent. And there is a deeper asymmetry: a global company captures the strategic asset, while the local community carries the ecological risk, adjudicated by a state government that has staked its prestige on the project going ahead. The honest synthesis is that both the opportunity and the cost are real, they fall on different people, and the mechanism for weighing one against the other — a rushed permitting process and after-the-fact litigation — is not built for a technology moving this fast.