Editorial Charter

Sustainability Question Looms Over Make AI in India Mission 

Data centres are key if India has to propel itself ahead in the AI supremacy race. Given their significant water and carbon footprint, is the environmental impact of data centres being given the short shrift?

Data centres are popping up in India almost like blades of grass after a heavy downpour on parched land.

Home to 163 data centres now, more are on the way. In the melee for opening new data centres, what gets often overlooked is their environmental impact. 

Data centres leave a big carbon footprint since they consume significant amounts of power and water. Ajith Sahasranamam, founder and CEO of Ongil.ai, used the example of ChatGPT to illustrate the carbon emissions from training a large language model, likening it to the absorption capacity of 2,500 acres of forests.

It is true that data centres are the backbone in the race for dominance in the artificial intelligence (AI) realm, since they power algorithms driving AI development. And it makes sense to have more data centres in India since proximity to data means it can be processed faster. Revenue from the data centre market is expected to grow 6.26 per cent each year until 2028, reaching almost US$ 10 billion, say experts.

As India floors the pedal in the AI era for greater profit and revenue, are we knowingly fuelling an environmental crisis?

Data Centres And Their Carbon Footprint

It’s estimated data centres globally contribute between 1 per cent to 3 per cent of total global emissions. Especially in India, the industry is more carbon intensive because fossil fuels are the main electricity source powering data centres.

The International Energy Agency (IEA) projects power consumption from data centres, AI and the cryptocurrency sector will double by 2026.

The surge in data centres can't be blamed exclusively on AI. There are other factors like growth of internet usage, fast-paced digitisation, expansion of cloud computing, big tech data analysis and higher performance requirements. All this ends up contributing to a larger carbon footprint than the entire airline industry. Think about that!

Take the example of Singapore-based STT Global Data Centres, which has one-third of its data centres in India. In its Environment, Social and Governance (ESG) report, STT Global Data Centres acknowledged it produced a total 790,112 tons of CO2e (carbon dioxide equivalent) in India in 2022. 

This is equivalent to the carbon footprint of 188,048 petrol cars or from electrifying 155,933 homes a year, per the US Environmental Protection Agency’s Greenhouse Gas Equivalencies Calculator.

President and Group CEO Bruno Lopez said STT’s operations run on 52 per cent renewable energy sources globally. But as per their 2022 ESG report, STT is yet to hit the 50 per cent renewables mark in India and would achieve the target by 2026.

STT data centres in Singapore and the UK don't have as much emissions as in India. This is likely because there are not as many data centres in Singapore and the UK as STT has in India, and these countries have stricter regulations on declaring carbon emissions.

One of the closest things India has to declaring ESG is the Securities and Exchange Board of India’s mandate for the top 1,000 listed companies to file the Business Responsibility and Sustainability Report (BRSR) with information of their ESG initiatives and carbon emissions.

“We have a lot of good-to-haves, but not must-haves,” said Shantanu Sharma, co-founder of Pro Zero Carbon, a company that helps companies minimise their carbon footprint. “If you go through the BRSR reports filed by companies, you’ll realise it’s more of a documentation process rather than a compliance aspect.”

In STT's case, it’s unclear how much energy is consumed from the grid, and how much from renewable sources. The Secretariat reached out to STT and this story will be updated once we get a response.

Another prominent India-based data centre company CtrlS Data Centres runs 15 data centres across Mumbai, Bengaluru, Noida, Hyderabad, and Kolkata, all Tier-1 cities and mostly where data centres set up shop.

CtrlS boasts a 275 MW data centre capacity and aims to build 600 MW total capacity of AI and cloud-ready data centres. Aiming to achieve carbon neutrality six years from now, some of CtrlS’ sustainable practices include setting up solar power plants, recycling water, paperless offices and offsetting carbon emissions.

Unlike STT Global Data Centres, CtrlS hasn’t made its ESG report available on its website but claims to have eliminated 225,000 metric tons of CO2 by adopting sustainable practices. 

“If they (data centre companies) are taking anything from the grid and offsetting it, that means they are not balanced in their scope 2 emissions,” added Sharma. Scope 2 emissions are greenhouse gases indirectly released when an entity buys electricity or other off-site operations.

To put the elimination of 225,000 metric tons of CO2 into perspective, it’s close to eliminating carbon emissions from over 55,000 petrol cars. This is commendable. However, if this is just the eliminated emissions, we were curious to find out the total emissions generated by CtrlS in India annually. 

The Secretariat sent a detailed questionnaire to CtrlS and asked for a copy of their ESG report and details of carbon accounting. A senior member of CtrlS’ corporate communications team said they won’t be able to participate in the story. 

Industry experts said Indian companies won’t voluntarily hand out details of carbon accounting given the amount of emissions generated by the data centre industry.

“Just saying ‘we are shifting to renewable’ is not the solution. You will have to look at your heating, ventilation, cooling, water usages, heat escape mechanisms,” said Sharma.

At data centres, there are servers that are simultaneously cooled by various systems, another reason why data centres are located in areas with a cooler climate.

There is also concern about their "water footprint". Large quantities of water is required to cool down servers inside data centres. Water is also used indirectly as a power source. It has been estimated that a medium-sized data centre (15 MW) uses as much water as three average-sized hospitals, or more than two 18-hole golf courses.

India already has a water crises on its hands. STT Global Data Centres in its ESG report said its water consumption was 419,204 cubic metres or 419.2 million litres in 2022. 

If an average Indian is estimated to use 49,275 litres a year (135 litres per day), then STT’s annual water consumption would equal the water needs of over 8,500 people annually.

“Water utilisation will be a major concern especially in India where fresh water is scarce,” added Sahasranamam. “Unlike emissions, which will be spread over the entire population, water exploitation will be localised.”

Is it even possible to have fully sustainable data centres in the future? “It’s not impossible to run fully on sustainable sources of energy. But, for context, Google and Apple have not been able to do it yet,” said Sharma.

AI’s appetite for electricity is bound to grow. In order to meet its hunger, OpenAI CEO Sam Altman recently said they might require an altogether new energy breakthrough like nuclear fusion to power it.

Mega Data Centre Deals Inked In India

Over the last few months, a spate of deals and announcements have come about to develop AI infrastructure at an unprecedented scale in India. The world’s biggest Graphics Processing Unit (GPU) maker Nvidia has inked a deal with Reliance to build more infrastructure to aid AI growth.

Also, Meta will soon inaugurate its first data centre in India in Reliance Industries’ Chennai campus to enhance data processing speed in its largest market. The government, in a bid to further its ‘Make AI in India’ project, has also allocated Rs 10,372 crore to enhance AI infrastructure in India to meet the growing demand.

This government investment is the result of recommendations made by seven expert panels in a report  titled 'India AI 2023' submitted in October 2023. These panels comprised representatives from government and industry.

“There are challenges to building AI in India. And one of the biggest challenges is lack of compute,” said Chirag Jain, co-founder of Textify AI and a member of the expert panel. “We don’t have our own commercial cloud compute and we don’t have enough GPUs.”

The expert panel also recommended the government create computing infrastructure of 24,500 GPUs at 17 centres. One of the things discussed in the panel meetings was the need for more data centres in India and that 50 additional data centres will be built as part of the government investment.

Among the things not discussed in the panel meeting was the environmental factors affecting AI development. “We did touch upon it (sustainability), but it was not the centre of discussion,” added Jain. 

The government has set a goal of reducing 50 per cent carbon emissions by 2030 and be net zero by 2070. The Data Centre Policy draft released in 2020 by the Ministry of Electronics and Information Technology encourages use of solar and wind-based renewable sources of power, but fails to address how they will tackle the impending environmental effects.

Data centres are going to be major contributors to global gas emissions, it’s imperative for India to address their long-term environmental and social impact. 

This is a free story, Feel free to share.

facebooktwitterlinkedInwhatsApp