Google's $15 Billion India Data Centre Promises 188,000 Jobs Amid Water and Wildlife Backlash

Google wants to spend $15 billion building one of India's largest data centre hubs on the outskirts of Visakhapatnam, and it wants you to know the project will create up to 188,000 jobs. What it has been considerably quieter about is where the water for tens of thousands of GPU racks is supposed to come from in a city that already can't supply its own residents, or why the site sits close enough to a wildlife sanctuary that leopards and pangolins now have a Fortune 500 neighbour. Built in partnership with Gautam Adani's conglomerate, the project has become a live test case for a question the AI industry keeps deferring: what happens when the infrastructure boom's promised prosperity collides with the physical limits of the place it's built on.

A $15 Billion Bet on a Water-Stressed City

Visakhapatnam was never an obvious choice on paper, and that's precisely why it matters. The coastal city delivers 410 million litres of water a day to 2.5 million residents who need roughly 480 million litres — a shortfall of nearly 15% that already produces routine rationing before a single server rack has gone live. Into that gap, Google and Adani are proposing to drop one of the most water- and power-intensive industrial facilities on the planet.

State officials have pushed back hard on the framing, insisting that "no water for rural or residential purposes will be used for the upcoming data centres" and that the facility will draw from dedicated sources rather than the municipal supply. Google, for its part, has committed to advanced air cooling technology specifically to reduce dependence on water-intensive cooling methods that have drawn criticism at data centres elsewhere in the world.

Why the Reassurances Aren't Landing

The trouble with "dedicated sources" as a talking point is that water tables and reservoirs in a coastal region don't respect the administrative lines drawn between "residential" and "industrial" allocation. Activists organizing under banners reading "We cannot drink DATA" argue that pressure on a nearby reservoir earmarked for the project will ripple outward regardless of how the allocation is labelled on paper. It's a dynamic that has played out at data centre sites from Arizona to Uruguay: official assurances about segregated supply chains routinely run into the reality that water systems are shared systems.

"We Cannot Drink Data": The Wildlife and Community Backlash

The water dispute is only half the resistance. Kambalakonda Wildlife Sanctuary — home to leopards and pangolins, among other species — sits roughly 860 metres from the construction site. Environmental groups, led in part by campaigners like Raja Rama Mohan Roy of Green Visakha, have filed legal challenges arguing that construction noise and long-term operational activity threaten a habitat that was never designed to coexist with a hyperscale industrial campus next door.

Authorities counter that the distance clears applicable legal thresholds and that the project complies fully with environmental regulations as written. That may be true on a technicality, and it is also beside the point that residents are making: a legal minimum distance is not the same thing as an ecologically sound one, and "we followed the rules" has never been a particularly reassuring answer to a community watching bulldozers move in next to a nature reserve.

The Protest Playbook Is Going Global

What's notable is how familiar this dispute already feels. The same combination of grievances — water stress, habitat encroachment, community groups turning to the courts because the political process moved too fast to intervene earlier — has now surfaced at AI data centre sites across multiple continents. Visakhapatnam is simply the latest node in a pattern:

• Legal recourse over political recourse: Communities are increasingly bypassing local approval processes, which tend to move in the developer's favour, and going straight to environmental courts and high courts.
• Water as the flashpoint, not power: Even in regions with adequate electricity capacity, water scarcity has become the fastest route to organized local opposition.
• Corporate mitigation pledges arriving after the backlash, not before it: Google's air-cooling commitment surfaced in response to organized pressure — not as an unprompted design choice baked in from day one.

The 188,000-Job Number Deserves a Harder Look

Here is where the story gets interesting for anyone actually trying to understand what AI infrastructure does to a local labour market. The 188,000 jobs figure is the headline number attached to nearly every report on this project, and it is doing enormous political work: it is the number state officials cite to justify environmental trade-offs, and it is the number that makes water and wildlife objections look, to some observers, like a small price for a large economic prize.

But data centres are not labour-intensive facilities once they're running. A hyperscale campus employs an army of electricians, welders, and construction crews for the two-to-four years it takes to build — and then, once operational, requires a comparatively tiny permanent headcount to keep servers cooled, powered, and secured. That is not a criticism unique to this project; it is simply how data centres work everywhere in the world. The question a number like 188,000 is designed to obscure is: 188,000 jobs of what kind, for how long?

Construction Surge vs. Permanent Payroll

Neither Google nor Andhra Pradesh state officials have published a detailed breakdown distinguishing temporary construction employment from permanent operational roles — a gap that is itself worth noting. Based on how comparable hyperscale campuses elsewhere have been staffed once fully operational, the overwhelming share of any large job figure attached to a data centre build is virtually always concentrated in the construction phase, not the decades of operation that follow.

Job Category	Typical Duration	Share of Headline Figures Industry-Wide
Construction & site build-out	2-4 years, project-based	Typically the large majority of cited totals
Indirect & supply chain roles	Duration of build plus early operations	Significant secondary component of most estimates
Permanent operational staff	Ongoing, life of facility	A small fraction — typically low hundreds per hyperscale campus

None of this means the 188,000 figure is fabricated. It likely reflects a genuine, if generous, economic modelling exercise that folds in construction, indirect, and induced employment across the region. But treating a multi-year construction surge as equivalent to 188,000 durable careers is the kind of framing that lets a project claim transformative local employment impact while the actual long-term headcount on the payroll numbers in the hundreds. Residents losing water access and a sanctuary losing quiet are permanent costs. The job boom, as typically structured at facilities like this one, is not.

India's Stake in the Global AI Infrastructure Race

Strip away the local dispute and the Visakhapatnam project is also a data point in a much larger story: India's aggressive push to become a serious node in global AI infrastructure, rather than merely a consumer of AI services built elsewhere. Hyperscalers are racing to secure land, power, and water rights across the country, and Andhra Pradesh's willingness to fast-track a $15 billion Google-Adani partnership signals how badly state governments want to win that race — and how much regulatory latitude they're prepared to extend to do it.

The Adani Factor

Google's choice of Adani as a build partner is not incidental. Adani brings the land acquisition muscle, political relationships, and infrastructure execution capacity that a foreign hyperscaler needs to move a project of this scale through Indian regulatory and land processes quickly. That partnership model — global AI capital paired with a domestic conglomerate that can navigate local approvals — is likely to become the template for how the next wave of Indian data centre investment gets built, for better or worse. It also means the local political calculus around projects like this one increasingly runs through the same handful of large domestic players, concentrating both the economic upside and the reputational risk of environmental disputes.

For India, the stakes go beyond one city. Winning a larger share of global AI compute investment means jobs, tax revenue, and geopolitical leverage in a technology race that Washington and Beijing are already treating as a national priority. But every contested project like Visakhapatnam's raises the cost of the next one: environmental litigation, community organizing, and international press coverage of "we cannot drink data" banners are not exactly the pitch state governments want circulating as they court the next $15 billion commitment.

The Visakhapatnam dispute is a preview, not an anomaly. As AI compute demand keeps compounding, more of these projects will land in water-stressed, ecologically sensitive, and politically underprepared regions simply because that's where cheap land and permissive regulation intersect. Google's air-cooling pledge and the state's insistence that no residential water will be touched may well hold up in practice — or they may join a long list of infrastructure promises that quietly loosen once construction crews go home and the courts move on to the next case. Either way, the 188,000-job headline will keep circulating long after the far smaller permanent workforce settles into place, because impressive topline numbers travel faster than the fine print explaining what they actually mean.