A Disaster in the Making: Red Carpet for Hyper-scale Data Centres
- 2 days ago
- 8 min read
Updated: 14 hours ago
So why are global tech billionaires zeroing in on Visakhapatnam?
By Ajith Pillai in Chennai
The impulsive and pushy rush in where the more cautious and considerate think twice before treading. This perhaps holds true for the haste with which India is rolling out the red carpet for international tech companies and domestic conglomerates keen to explore opportunities in AI and hyper-scale data centres.
A prime example of this welcoming approach is Andhra Pradesh, which aims to become a leading hub for AI data centres. To achieve this ‘exalted’ status, the state government has decided to ignore the considerable environmental damage caused by these facilities.
This includes carbon emissions, the humongous consumption of power and water, and the disruption they cause in surrounding areas by depleting the local water table, which impacts daily life and agriculture.
The first hyper-scale data centre, with an investment of $15 billion, will be built near Visakhapatnam. This facility will be the largest project of its kind in India and reportedly represents a partnership between Google, AdaniConnex and Airtel Nxtra.
According to the Human Rights Forum, which has been monitoring the establishment of data centres in the region, the project has been prioritised by both the state and central governments. Allegedly, environmental clearance was granted in just nine days, without conducting a proper environmental impact assessment, which is usually required for projects of this scale.
The forest department swiftly issued a No Objection Certificate (NOC), despite the project being located in an ecologically sensitive, biodiversity-rich area near a wildlife sanctuary. Additionally, there was no effort to record the concerns of local communities in the villages affected by the data centre's construction.
Ironically, this push in Andhra Pradesh comes at a time when there is growing public concern and outcry against hyper-scale data centres across the world, particularly in the United States. According to multiple media reports, including one in The Guardian (London), 48 data centre projects worth an estimated $156 billion were “blocked or stalled” in the US last year by local opposition. A Gallup poll last month revealed that seven in ten Americans oppose the construction of data centres for artificial intelligence in their local area.
Two US lawmakers, Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez, are responding to the prevailing public sentiment by promoting the legislation known as the Artificial Intelligence (AI) Data Centre Moratorium Act. This proposed Act aims to temporarily halt the development of AI technologies and the construction of new data centres.
Here are some points raised in the proposed Act pertaining to data centres, which should set our politicians thinking:
*New or upgraded AI data centres must not increase the electric and utility bills of consumers, harm the environment, or make climate change even worse;
*The American people must be empowered to approve or reject the construction or upgrading of AI data centres in their local communities;
*New AI data centre construction and upgrades must use union labour to create union jobs with strong labour standards, including paying prevailing wages, utilising registered apprenticeship programmes, and establishing project labour agreements;
*AI data centres must not receive subsidies from US taxpayers for construction, upgrading, or operations.
*The Department of Energy shall collect and publish data related to AI data centers, including all financial vehicles involved in the facility’s operation; water usage; energy usage and infrastructure needs; on- and off-site greenhouse gas emissions associated with facility operation; fenceline air quality monitoring; worker wage and benefit information; temporary and permanent jobs created; wastewater discharge and thermal outputs; cooling chemicals used; noise levels; certification of no utilization of federal, state, and local subsidies…
Due to the influence of Silicon Valley billionaires in the Trump administration and the Republican Party, the Act may not be passed by Congress. However, lawmakers in 14 states, including Michigan, Maine, Pennsylvania, New York, Vermont, and Virginia, are considering a moratorium on new hyper-scale data centres.
In sharp contrast, protests from lawmakers and the public in India have been largely minimal or nonexistent. This lack of opposition and the ready availability of land have encouraged major tech companies from Silicon Valley to plan data centres in the country and take advantage of various incentives, including a tax holiday that extends until 2047, as announced in this year's Union budget.
The port city of Visakhapatnam and its surrounding areas will be the first to be affected by the Andhra Pradesh government's data centre push. Reports indicate that by 2030, the state government plans to provide a total of 5 gigawatts (GW—one gigawatt is one billion watts) of data centre capacity in Visakhapatnam.
This capacity, which represents the total electric power available to operate IT equipment, cooling systems, and critical infrastructure, has reportedly been promised to major companies, including Google, Meta, Reliance, Tata Consultancy Services, and six others. Additionally, the rest of Andhra Pradesh is expected to add another 1 GW of capacity within the same time-frame.
Predicting the exact number of hyper-scale data centres in Andhra Pradesh by the end of the decade is challenging. Projections typically focus on capacity growth, measured in gigawatts, rather than on providing a specific count of these facilities.
When Andhra Pradesh sets a target of 6 GW data centre capacity by 2030, with Visakhapatnam expected to contribute 5 GW of that total, it is significant, especially considering that India's current total data centre capacity is only 1.3 GW. This means that Visakhapatnam will eventually have nearly five times the data centre capacity of Mumbai, Delhi, Bangalore, and Chennai combined.
So, what does a 1 GW hyperscale data centre supporting AI require to function? For a start, it is a massive project requiring large inputs of power and fresh water. It would also require between 500 and 800 acres of land.
According to techtrends.com, a credible US technology platform, a 1 GW data centre is a facility capable of drawing 1,000 megawatts of power continuously to support AI workloads on a scale. That is enough electricity to power roughly 700,000 to 800,000 US homes. In short, if a data centre of that size runs for a year, it will consume as much power required to power upwards of 700,000 homes for a year. That is the size of a mid-sized city.
Now imagine 5 GW of capacity provided in the vicinity of a Tier-2 city like Visakhapatnam!
A 1 GW facility also requires between 500 million and 1 billion gallons of water each year. Official figures indicate that Visakhapatnam has the lowest groundwater level in the state, measuring at 2.12 TMC (thousand million cubic feet). Consequently, the operation of even a single 1-GW-capacity plant could lead to significant water shortages.
However, the government has assured us that desalination plants will help address these shortages by drawing water from the Bay of Bengal. But there is no clarity about the amount of water that will be drawn from the sea, and how.
As in the US, hyper-scale data centres have been sold in Andhra Pradesh on the development plank. But will these facilities provide as many jobs as promised?
Yes, at the construction stage, it will provide employment to workers. But once it is set up, a mega data centre requires fewer than 200 people to run.
John Dinsdale, chief analyst at Synergy Research Group, made this pertinent observation in the Wall Street Journal: “The reality is data centres can employ more than 1,000 people in the several months or years it takes to build them, but rarely need more than one or two hundred once they open.”
This view was echoed in the same publication by John Johnson, chief executive of data centre operator Patmos Hosting: “Data centres have rightly earned a dismal reputation of creating the lowest number of jobs per square foot in their facilities.”
So why are tech billionaires zeroing in on Visakhapatnam?
Ready availability of land is one explanation. An AI and tech-friendly Andhra Pradesh Chief Minister, Chandrababu Naidu, is the other. His party is indeed propping up the BJP regime in Delhi, which does not have a majority in Parliament. Union IT Minister Ashwini Vaishnaw added another perspective during the foundation-laying ceremony for the Google facility on April 28.
He focused on three subsea optical fibre cable landings along the Visakhapatnam coast to enhance global digital connectivity. This will provide substantial bandwidth and reduce latency, resulting in minimal delay between an input or command and the corresponding response or action.
Said Vaishnaw: “One cable will go all the way to Australia, then via the Pacific Ocean to the west coast of the US. The second one will go to West Asia, to Europe, then to the US. The third one will go around the Cape of Good Hope, the African region, and then go straight to the US.”
As a result, the data centres on the Andhra coast will serve as a gateway for Silicon Valley to connect with India, disrupt local businesses and even governance, similar to what has occurred in Britain, where entire government departments are now reliant on US-based AI companies. The consequences in India could be significant job losses in a country already struggling with an unemployment crisis.
When it is pointed out that hyper-scale data centres create only a limited number of jobs, it is noted that they will significantly benefit AI companies and the country's IT sector, enabling them to scale up operations.
In this context, it is important to consider two recent developments involving Microsoft and Uber that have sparked discussions in Silicon Valley. According to the tech magazine The Verge, Microsoft has started to cancel most of its Claude Code licenses and has instructed its employees, including engineers, to stop using the AI agent. The reason for this decision is that the cost associated with employees using Claude was deemed economically unviable. This action follows just six months after Microsoft encouraged its workforce to adopt Claude.
Similarly, Uber is also reducing its reliance on AI. Andrew Macdonald, Uber's operations chief, indicated that the company is finding it increasingly difficult to justify the costs associated with AI. He noted that he does not see proportional productivity gains that would offset these rising costs. In simple terms, the expense of operating AI systems has become economically impractical for the company.
In conclusion, it would be worthwhile to refocus on the proposed Artificial Intelligence (AI) Data Centre Moratorium Act. Here is what Mitch Jones, managing director of policy and litigation at Food and Water Watch, had to say about AI data centres: “We need a halt to the explosive growth of new AI data centre construction now, because political and community leaders across the country have been caught completely off guard by this aggressive, profit-hungry industry. It has yet to be determined if – not how – the industry can ever operate in a manner that sufficiently protects people and society from the profusion of inherent hazards and harms that data centres bring wherever they appear.”
Is anyone in the establishment or the Opposition in India listening?
Editor’s Note: This is the first part of a story on Vishakhapatnam’s hyperscale data centres.
Ajith Pillai is member, Editorial College, senior editor and writer, independentink.in. A seasoned journalist working in the profession for 40 years, he has reported out of Delhi, Mumbai, Chennai, Andhra Pradesh and Kashmir on a broad spectrum of events related to politics, crime, conflict and social change. He has worked with leading publications, including The Sunday Observer, Indian Post, Pioneer, The Week and India Today, where he headed the Chennai bureau. He was part of the team led by Editor Vinod Mehta that launched Outlook magazine and headed its current affairs section till 2012. Under his watch, Outlook broke several stories that attracted national attention and questioned the government of the day. He has written two books—'Off the Record: Untold Stories from a Reporter’s Diary,’ and a novel, ‘Junkland Journeys’.
Also see by Ajith Pillai:
When Tech Billionaires Come Marching In
AI to the left of them, AI to the right
AI is not reliable. Especially in a conflict zone. QED. Hence proved.
To err is AI: Bubble burst?