Background

Data centres are mushrooming in India, and generating controversy, as has become almost routine in the US and Europe, due to their impact on energy, water and other environmental resources. Google’s hyperscale 1GW (1000 mega watts) data centre in partnership with Adani Connex coming up in the coastal city of Vishakhapatnam (Vizag) in AP, as the early entrant to India in this size bracket, has brought these issues to the forefront. The project envisages investment of around $6 billion (INR 53.25 lakh crore) over 5 years with facilities in 3 locations around Vizag. Airtel is another partner providing user interface and broadband connectivity. This is said to be Google’s largest investment and data centre outside the US. Together with landing centres for undersea cables connecting to many countries, and other new Data Centres and facilities, $10 billion to $15 billion investments are expected in the Vizag AI hub alone over 2026-30. These herald a wave of AI and cloud investments appear headed for India, According to Union government spokespersons, investments of upto $200 billion are expected in India nationwide with tax waivers up to 2047.

Data Centres

Data centres have long been the back-end internet infrastructure for storage, processing, distribution and related cloud services. The rapid advance of AI has exponentially expanded demand for data centres in both size and spread.

Globally there are around 10,000-12,000 data centres, with about 5,000 in the US alone, of various sizes from micro, through average (100 MW), large (200 MW) and hyperscale (1000 MW i.e. 1 GW or more). (Note: size of a Data Centre is depicted in terms of the power it requires). Of the total global capacity of around 122 GW in 2024, the US accounted for just under half, with China and the EU accounting for about 20% and 12% respectively. The global market for data centres was estimated to be about $243 billion in 2024, projected to grow to about $585 billion by 2032 at an annual growth rate of 11.7%.  

The ‘big four’ internet data generators namely Amazon (29%), Microsoft (20%) and Google (12%) together accounting for 63% of global cloud services, along with Meta (parent company of Facebook, Whatsapp and Instagram) which has an extensive network of data centres albeit mostly for its own use, dominate the data centres market. China stores most of its data within the country.

  Major expansion of data centres to outside the US is underway due to: (a) demands in different parts of the world; (b) shorter data latency i.e. time gap in transmission of data and services; (c) demands by many countries for data localization, i.e. storage of national data within the country itself; and (d) lower energy and infrastructure costs. India is an important and growing market increasingly pressing for data localization and sovereignty. With 900 million internet users, India generates about 20% the world’s data, while currently hosting only around 3.6%.

Data localization, sovereignty and regulation

Vast volumes of data generated by individuals, corporations and governments are stored in data centres owned and run mostly by giant US corporations. This enormous data is for commercial gain, now additionally through development and training of dominant AI models and autonomous agents. Experience has shown that data is also used for political manipulation, intelligence gathering and military purposes, all with minimum transparency. Regulation of this data thus assumes enormous significance.

Whereas the EU does not compel data localization, it has stringent regulations governing personal data of EU nationals, under which EU privacy laws apply to EU-origin data irrespective of where it is stored. India has several overlapping regulations but mainly governed by the Digital Personal Data Protection Act (2023). This does not compel all India-origin data to be stored in India, but has rules governing such data, and requires that specific categories of personal financial, health, bio-metric etc data must be stored in India, along with governmental and national security data.

Yet the operation of these regulations and actual practices regarding use of data even when stored in India, remain problematic. There is little transparency in corporate use of this data. Use of personal data by government is also opaque, and concerns have been expressed by many civil society organizations working in the legal and IT sector space about surveillance, censorship and political use and misuse of personal data. A nexus between government and corporate interests is also of much concern.

Growing backlash 

While struggles continue on all these issues worldwide, the US and Europe are witnessing a growing backlash against data centres, mainly due to environmental concerns. In the US, data centres worth $64 billion have been blocked or stalled by bi-partisan local opposition. In Europe too, protests have led to denial of industrial permits in Holland, Spain, Ireland etc impacting data centres worth $42 billion.

Opposition to data centres have been due to: (a) subsidies by local authorities at a time of public austerity; (b) strain on power grids, electricity supply and costs; (c) huge demand on water resources for cooling, often in water-stressed localities; (d) noise from constantly whirring servers; (e) loss of green spaces. With hyperscalers, these issues multiply manifold. India is set to join this trend, exacerbated by lack of transparency and regulatory failure.

It is indeed ironic that AI and data centres, considered to be at the forefront of advanced tech, no longer appear as clean, green or modern but as an environmentally malign presence. The general public may not object to these technologies per se, but would rather not have data centres in their localities --- the latest industry to join the NIMBY (not in my backyard) list.

Energy use and Emissions

The demand for electricity by data centres for both operations and cooling is a major problem internationally, and has not been discussed adequately by government here, for example in the Vizag hyperscaler.

The Vizag project is expected to be powered by renewable energy (RE) possibly to the extent of 80%. Some fossil-fuel energy may still be needed for base and peak loads, additional cooling during heat waves, and to even out fluctuations. AP government Ministers have said that Google’s $6bn investment includes $2bn for development of RE sources and related infrastructure including a dedicated grid. Some Adani and related sources have claimed that the Vizag Data Centre would be powered 100% by RE from Adani, including local storage to balance the load, and investments to improve the grid and local electricity infrastructure. However, no specifics are available in the public domain in Vizag, which is already struggling with a summer peak load of around 1,500 MW, exacerbating anxiety in the City and surrounding areas.

Globally, energy use by data centres are skyrocketing even if sources are mainly RE, straining power generation and grids.

Consequently, running counter to trends in other sectors, AI has driven up greenhouse gas (GHG) emissions significantly. Some estimates suggest that, while energy use would increase by 170% by 2030, global emissions from data centres would accumulate to about 2.5 billion tons of CO2 equivalent. And emissions of the top 5 data companies, already equivalent to the 30^th ranked country in 2022, would be rapidly climbing the ladder!     

Water the biggest issue

No regulation in India currently requires disclosure of information regarding water use by data centres. The Union Ministry of Electronics and Information Technology (MeitY) published a Draft Policy on data centres in 2020 but never finalized it.

Journalists who have accessed the Environment Clearance given to the Vizag project say there is no information given regarding water resources, and that neither Google nor the AP government has been forthcoming.

Several technologies are in use for cooling data centres that generate enormous heat which must be tackled for smooth operations and to prevent damage to sensitive chips and other electronics.  

The older and least efficient method using enormous quantities of water is evaporative cooling, where water is circulated to carry away the heat and then evaporated before the condensate is recycled. Air cooling is also used, but requires large quantities of electricity and poses problems in hot and humid conditions. More modern plants use closed-loop systems where water is loaded once, takes away the heat which is cooled in heat exchangers before being recycled, using less water. Options such as using recycled waste-water or using sea-water requiring use of expensive, non-corrosive piping materials are also available. Using sea-water from desalination plants is another alternative. A further improvement used in some systems is closed loop direct-to-chip cooling using special liquids circulated close to the chips and recycled. All these technologies are now available and in commercial use, but the more advanced systems call for significantly higher up-front costs. Limitations therefore are financial rather than technological.

The forthcoming 1.5GW Reliance data centre project in Vizag has applied for an additional 80 acres for a desalination plant, which at least tells us what it plans to do. But the public in Vizag and the wider region are totally in the dark about water use in the Google-Adani and other envisaged data centres, fueling even more anxiety especially with government looking the other way.

Regulatory collapse

All in all, the envisaged AI and cloud hub in Vizag involving at least 6.5GW of hyperscalers, will invariably have a huge environmental footprint if one counts agricultural including orchard lands, fisheries and coastal lands, forests including substantial areas of reserved forests, catchment areas of a major reservoir for water supply to Vizag, and the Kambalakonda Wildlife Sanctuary nearby.

Not much thought seems to have been given to these issues. On the contrary, the Project was examined under Category B2 for projects with “minimum environment impacts!” It is learned that the company had applied for Environment Clearance on 9 April and received it on 18 April from the State Environment Impact Assessment Authority (SEIAA) in astonishingly short time. Indeed, hyperscale data centres should be subject to thorough EIA under category A1 given their evidently extensive impacts.    

The regulatory compromises with regard to the Vizag data centre project comes as a rude reminder of the deliberate downgrading of environmental regulations and institutional mechanisms over the years since 2014 in order to speed up the “ease of doing business.” If the Vizag project is any indication, the country should brace itself for much resistance to future data centres in the months and years ahead.