Nadia Dubois
October 12, 2026
13 min read
Google is pushing back publicly against mounting community criticism of its $15 billion artificial intelligence data-hub project near Visakhapatnam, in India’s Andhra Pradesh state, as the facility’s role in the global AI compute race comes into sharper focus. The company, building the gigawatt-scale campus with the Adani Group, says the site will become one of the largest AI infrastructure investments it has made anywhere outside the United States, feeding the compute pipeline that trains and serves Google’s Gemini model family across Asia. As of October 12, 2026, the dispute has become a test case for how hyperscalers justify massive AI buildouts to the communities that host them.
The stakes go beyond one campus in coastal India. Every major AI lab, from Microsoft and Meta to Google, is racing to secure power, land and water for the data centers that train frontier models and serve billions of inference requests a day. Visakhapatnam has turned into a flashpoint precisely because Google’s AI ambitions there are colliding with a coastal city already short on water and power.
Don’t miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Google Announced in Visakhapatnam
Google and the Adani Group first announced the Visakhapatnam AI hub in October 2025, describing a joint venture, built through the AdaniConneX partnership, that would invest roughly $15 billion over five years. The companies describe the project as gigawatt-scale, spanning more than 600 acres across three sites in the region, with public filings pointing to a 1-gigawatt data-center campus as the core specification. Google has said the hub is expected to become operational in 2028.
Thomas Kurian, CEO of Google Cloud, framed the investment in blunt competitive terms when the deal was unveiled, calling it “the largest AI hub that we are going to be investing in anywhere in the world outside of the U.S.,” a comment that underscored how central India has become to Google’s AI infrastructure strategy. Kurian separately said the company is investing in the hub “to unlock India’s massive potential in the AI age,” positioning it as critical foundation to let “businesses, researchers, and creators build and scale with AI.” Google has laid out its own framing of the project, including its community-impact commitments, on its official company blog.
The timing matters. Google announced the project roughly a year before the current controversy flared, and construction has advanced steadily since, surviving a legal challenge at the Andhra Pradesh High Court in August 2026 that questioned the project’s environmental clearances but did not halt building. That history, now intersecting with active protests, is why the story has resurfaced as a defining AI-infrastructure story heading into the final quarter of 2026.
The AI-Model Capacity Angle: Why This Is Compute, Not Just Concrete
Data center fights tend to get covered as real-estate or environmental stories. The Visakhapatnam hub is different because of what it is actually built to run: the accelerator clusters that train and serve Google’s Gemini models and other Google DeepMind systems. Hyperscale AI campuses of this scale are not general-purpose server farms. They are purpose-built for the dense GPU and TPU racks that large language models require, both for the computationally punishing pre-training runs behind each new Gemini generation and for the inference traffic that keeps Gemini-powered Search, Workspace and enterprise cloud products running for hundreds of millions of users a day.
Google has not disclosed which specific Gemini training runs, if any, will be assigned to Visakhapatnam, and no public filing confirms a named model workload for the site. What Google has said is that the campus is meant to support what it calls the most demanding AI workloads in India, a description that fits both frontier-model training and the high-volume inference layer that regional cloud customers depend on, the same category of workload driving demand for high-throughput AI networking hardware worldwide. That ambiguity is itself notable: hyperscalers increasingly treat the location of specific model training runs as a competitive secret, even while trumpeting the capital numbers behind the facilities that could house them.
The broader pattern is unmistakable. As AI labs push out ever-larger models, the compute bottleneck has shifted from chip design to the physical footprint, power supply and cooling capacity needed to run thousands of accelerators continuously. India, with its large engineering talent pool and growing cloud market, has become one of the two or three countries hyperscalers are racing to build AI capacity in outside their home markets, alongside Southeast Asia and the Gulf states.
Google’s Defense: Air Cooling, Power Promises and Local Benefits
Facing organized local pushback, Google has leaned on a specific technical and economic defense, detailed in reporting by Techxplore. The company says the Visakhapatnam facility will rely on advanced air cooling rather than water-based cooling systems, a choice it frames as a direct response to the region’s water stress. Alexander Smith, Google’s head of infrastructure and energy for Asia-Pacific, said the company’s own water assessment found pressure on the local watershed, and that “that finding shaped our choice of cooling configuration from the outset.”
On the power side, Smith laid out what amounts to a conditional commitment: “If there is insufficient power in a location, or if operating a data centre would mean taking power away from a community, Google will either find ways to develop more power, or it will not build there.” That statement, delivered amid scrutiny over whether the gigawatt-scale campus could strain the regional grid, is Google’s clearest public commitment that it will not simply draw down shared electricity capacity to run its AI hardware.
Smith also framed the project in terms of long-term local payoff rather than a one-time construction event, saying “the watershed, the grid and the communities around our facilities must be better off for our being here,” and describing the investment as “a multi-decade commitment.” On the economic side, Google has said the hub will generate roughly $688 million in local economic activity annually once operational, and has pointed to approximately 4,200 construction jobs tied to the build-out, with hundreds of specialist roles expected after operations begin.
Why Residents Are Not Reassured
Local groups and environmental activists have organized against the project through a campaign called Stop Dirty Data Centres, which has canvassed villages and neighborhoods near the three construction sites, according to CBC News. Their core objections track three resources: water, electricity and land, plus broader concerns about the project’s ecological footprint and whether the promised jobs will actually reach the surrounding communities rather than skilled workers brought in from elsewhere.
The legal fight has not gone away even though it has not stopped construction. The Andhra Pradesh High Court challenge filed in August 2026 raised questions about the basis for the project’s environmental clearances, and campaigners have argued separately that projects of this scale have been processed through a lighter-touch regulatory category that limits the scope of public hearings and environmental-impact review compared with what critics say a gigawatt-scale AI campus should require. Google has not disputed that the legal challenge exists, but the company has continued building while the case proceeds.
One detail that has not been independently confirmed deserves a caveat: some environmental-clearance paperwork, as reported by Ground News, has cited a capacity figure as high as 2.51 gigawatts, well above the 1-gigawatt specification Google has publicly described. That gap between the number in the public announcement and the number reportedly appearing in clearance filings is itself a point of contention for critics, who argue it reflects uncertainty about how much power and water the completed campus will actually consume.
Visakhapatnam in the Context of India’s Data Center Boom
Visakhapatnam is not an isolated case. Protests against large AI data centers have also emerged in Thane, near Mumbai, as campaigners there raise similar concerns about electricity and water allocation tied to new hyperscale projects from competing cloud providers. The pattern mirrors a global shift: AI infrastructure announcements that once generated almost uniformly positive local reception, because they promised jobs and tax revenue, are now drawing sustained organized opposition once communities start asking how much water and power a gigawatt-scale campus actually needs. That tension has already surfaced closer to home for Google’s rivals, too, including in zoning fights over proposed US data center campuses.
India’s position in this story is structural rather than incidental. The country has emerged as one of the fastest-growing cloud and AI infrastructure markets in the world, and hyperscalers are competing for sites with reliable power access, fiber connectivity and favorable land costs, a dynamic covered in detail by ETDatacenters. That competition is intensifying even as reporting on the sector has flagged growing concern about freshwater use tied to data-center cooling across multiple Indian states, a tension that predates the Visakhapatnam dispute but has been sharpened by it.
Competitive Landscape: How Google’s India Bet Compares
Google is far from alone in treating India as a priority AI infrastructure market, and the Visakhapatnam dispute is forcing comparisons with how other hyperscalers have approached similar buildouts. The table below lays out what has been publicly confirmed about Google’s project alongside the general shape of the AI infrastructure race in the region, without extending beyond what has been officially disclosed.
Historical Context: From Server Farms to AI Megaprojects
Google’s data-center footprint in India predates the current AI boom by more than a decade, but the scale and framing of this project mark a clear break from earlier buildouts. Prior-generation data centers were typically sized to serve regional cloud and search traffic. The Visakhapatnam hub is explicitly framed, by Google’s own account, as infrastructure for frontier AI workloads, a category of demand that did not meaningfully exist in India’s data-center planning conversations before the generative AI wave that began in 2023.
That shift explains why the water and power stakes have escalated so quickly, and it is the same dynamic fueling demand for the high-capacity networking gear and grid-scale battery capacity now required to keep AI campuses running continuously. A traditional hyperscale data center built five years ago drew meaningfully less power per rack than a facility designed around dense accelerator clusters built for model training. The same physical footprint today needs to support far higher power density, which is part of why Google’s own framing leans so heavily on the air-cooling decision as its central defense against water-stress criticism, rather than simply pointing to its historical track record operating data centers in the region.
What Happens Next: Five Predictions
- Expect the Andhra Pradesh High Court case to continue running in parallel with construction rather than halting it, given that the August 2026 challenge did not stop building and Google has shown no sign of pausing the timeline toward a 2028 operational date.
- Google is likely to publish more granular water and power usage commitments for Visakhapatnam specifically, moving beyond company-wide sustainability figures, as pressure from campaigners and local officials continues to build through the rest of 2026.
- Other hyperscalers with active or planned Indian AI campuses will face similar community scrutiny, following the same pattern already visible in Thane, as AI data-center announcements increasingly trigger organized local opposition rather than uncomplicated economic-development enthusiasm.
- Watch for Google to disclose, at least partially, what kind of Gemini-related training or inference workloads the completed Visakhapatnam campus is expected to carry once it nears operational status in 2028, something the company has so far avoided specifying.
- The gap between the publicly announced 1-gigawatt specification and the higher capacity figures reportedly appearing in clearance paperwork is likely to remain a recurring point of contention in local reporting until Google or Indian regulators clarify it on the record.
Market and Industry Impact
For Google, the Visakhapatnam dispute is a reputational test at a moment when the company is simultaneously trying to demonstrate it can scale Gemini-class AI infrastructure responsibly and compete with Microsoft and Meta’s own massive AI capital spending commitments. The three companies’ combined AI capital expenditure has become one of the most closely watched figures in the industry, and any one of them mishandling a community relations fight at this scale risks becoming a template that regulators and activists elsewhere point to when new projects are proposed.
For India, the stakes cut the other way. The country wants to be seen as an attractive, fast-moving destination for AI infrastructure capital, not a jurisdiction where environmental and community objections can tie up multibillion-dollar projects in court for years. How the Andhra Pradesh state government and the courts resolve the clearance questions raised by the Visakhapatnam case could shape how aggressively other states court similar projects, and how carefully hyperscalers scope their public commitments before breaking ground next time.
Data Snapshot: Verified vs. Unconfirmed Project Figures
Given how much conflicting data has circulated around this story, it is worth laying out plainly what has been officially confirmed by Google or named outlets, versus what remains reported but unverified.
The Regulatory Question That Will Not Go Away
At the center of the legal challenge is a question that extends well beyond Visakhapatnam: whether gigawatt-scale AI data centers should be reviewed under the same regulatory categories designed for conventional industrial projects, or whether their power and water intensity warrants a dedicated environmental review track. Critics argue that processing a project of this magnitude under a lighter clearance category allowed it to avoid the fuller environmental-impact-assessment and public-hearing process that larger industrial developments typically face.
That question is now effectively in front of the Andhra Pradesh High Court, and however the case resolves, it is likely to become a reference point for how Indian states classify AI infrastructure projects going forward. Given how many hyperscalers, including Google, Microsoft and companies racing to secure AI compute capacity more broadly, are competing for sites across India, a definitive ruling on clearance categories could reshape how quickly future projects can move from announcement to groundbreaking.
Why This Matters Beyond India
The Visakhapatnam standoff is a preview of a conflict that is going to recur wherever hyperscalers try to site the next generation of AI infrastructure. As AI companies push to add gigawatts of new compute capacity to train and serve increasingly capable models, the physical footprint of that ambition, land, water, power and local labor, is going to collide with communities that have not historically had a seat at the table when infrastructure decisions get made. Google’s air-cooling commitment and its conditional power pledge at Visakhapatnam may end up functioning as an informal template other hyperscalers reach for when facing similar objections elsewhere, precisely because it gives the company a concrete, falsifiable claim to point to rather than a vague sustainability pledge.
Whether that template holds up will depend on what happens after 2028, when the campus is supposed to go live and residents will be able to check Google’s promises against actual water and power draw. Until then, the Visakhapatnam dispute remains a live test of whether a $15 billion AI infrastructure commitment can survive sustained local scrutiny without either stalling in court or straining the resources it promised not to touch.
Frequently Asked Questions
How much is Google investing in the Visakhapatnam AI data center?
Google has committed approximately $15 billion over five years, from 2026 to 2030, to the project, developed jointly with the Adani Group.
When will the Visakhapatnam AI hub become operational?
Google has said the facility is scheduled to become operational in 2028.
Will the data center use water for cooling?
Google says the facility will rely on advanced air cooling rather than water-based cooling systems, a choice the company has said was shaped directly by its own assessment of pressure on the local watershed.
What is the capacity of the project?
Google and Adani describe the project as gigawatt-scale, with public reports identifying a 1-gigawatt data-center campus as the core specification. A higher figure of 2.51 gigawatts has appeared in secondary reporting citing clearance paperwork, but this has not been independently confirmed against Google’s own public statements.
Why are local residents protesting the project?
Residents and environmental groups, organized partly through the Stop Dirty Data Centres campaign, have raised concerns about pressure on water supplies, strain on the regional electricity grid, proximity to protected wildlife habitat, and whether promised jobs will meaningfully benefit the local community.
Has any court blocked construction?
No. An Andhra Pradesh High Court challenge filed in August 2026 raised questions about the project’s environmental clearances, but it did not halt construction, which has continued.
Does the data center train Google’s Gemini AI models?
Google has not confirmed which specific AI workloads will run at the Visakhapatnam campus. The company has described the facility as intended to support its most demanding AI workloads in India, a description broad enough to cover both model training and large-scale inference, but no named Gemini training run has been publicly tied to the site.
What economic benefits has Google promised the region?
Google has said the project will generate approximately $688 million in local economic activity annually once operational, alongside roughly 4,200 construction jobs and hundreds of specialist roles after operations begin.
