Google secured environmental clearance for 2.51GW of capacity for its $15bn AI datacentre hub near Visakhapatnam in Andhra Pradesh, more than double the 1GW announced in October 2025. The figure comes from clearance papers seen by the Guardian, which compares the load to the output of two large nuclear reactors and more than 30% of the state's current yearly power use. The gap between the announced and approved scale matters because access to power, land and permits now shapes AI costs as directly as hardware.
Approved capacity and nine-day clearance
The hub covers three sites around Visakhapatnam and is being developed with AdaniConneX and Airtel. State authorities granted environmental approval in nine days with no public hearing, according to reporting from Tarluvada by Hannah Ellis-Petersen and Aakash Hassan. The support package of tax breaks and cheap land, power and water is estimated at 220bn rupees ($2.2bn) over 20 years. Asked about the scale, Google described the project as a gigawatt-scale hub comparable to other projects, without confirming the 2.51GW figure.
Google says the site will use air cooling rather than water cooling, so local drinking water supplies will not be affected. That choice moves the main burden to the power system, since 2.51GW implies grid connections and long-term supply contracts well beyond a first 1GW phase. One site lies a few metres from a reserve forest, a drinking water reservoir and a wildlife sanctuary, which increases environmental oversight. The design therefore trades lower water risk for higher exposure to electricity pricing, grid readiness and compliance review.
Legal challenges are already in motion. The Human Rights Forum has filed three petitions at India's National Green Tribunal, and a separate case is before the Andhra Pradesh High Court, with none decided so far. The Adani Group says all legal approvals are in place, while the state says no land was taken by force. The Guardian notes the dispute mirrors datacentre protests in the US and elsewhere, where communities contest power demand, land use and consultation. For the industry, the case tests how fast gigawatt-scale AI infrastructure can advance under standard environmental procedures.
What this means for buyers of AI capacity
For companies that rent AI compute, the practical point is that permitted capacity can exceed announced capacity and allow expansion without fresh approval. If built out, 2.51GW in Andhra Pradesh would add a large pool of power-backed space for cloud, colocation and AI services in India. Large customers might negotiate direct hyperscale placements supported by subsidized inputs, while smaller firms would feel the effect through availability and pricing from Google and its partners. In both cases, delivery dates will hinge on power allocation and court outcomes rather than on servers alone.
Land and community risk needs separate checks before planning around the hub. In Tarluvada, 520 families had land taken back that the government had allocated about 20 years ago, and residents said Adani Group security guards erected fences in April blocking access to farmed plots. Among 47 Dalit farmers who accepted payment, some report they have not received promised replacement plots and jobs; P Venkat Rao said he received $40,000 for his only acre and now has no livelihood. Google says it expects no impact on the local community and points to thousands of jobs around Visakhapatnam. Buyers should ask for petition status, guaranteed power tranches and written resettlement terms.
The signal to follow is the sequence of court decisions and construction steps across the three sites, plus any update on movement from 1GW toward 2.51GW. Confirmation of the clearance and steady build-out would establish subsidized gigawatt-scale hubs as a repeatable model for AI capacity in India. A court-ordered review or new hearing would instead reset schedules and costs for similar projects.
