Ai is kicking up more controversy at every step of the way. Pope Leo’s call to prevent AI from being deployed for malicious uses, including replacing human relationships, depicts AI as a general-purpose technology capable of causing significant harm. The burden of AI adoption can have significant social, political, and economic repercussions for India and the Global South.
India has been emphasising safe, trusted, and ethical use of AI. Preventing AI from becoming an instrument of domination will require the world to make rules for serving the purpose. Such rules, though, are currently elusive. India and the Global South must be ready to address challenges arising from building larger AI capacities.
Resource Strain
Globally, AI infrastructure growth is causing major strains on natural resources. Demand for electricity is increasing at exponential rates. Along with electricity, there is high demand for use of water. Both demands are compounded with rising appetite for vast tracts of land. All three will manifest in India as it builds more AI data centres.
AI data centres require huge computing power. Generative AI like ChatGPT and Claude need to continuously process data for staying advanced and updated. Computing capacities of data centres, especially the hyperscale ones that host thousands of servers and run accelerated graphics processing units (GPUs), require high uninterrupted supply of electricity, and water for cooling the overworked systems. The big data centres with large computing capacities also need sizable pieces of land to grow.
The International Energy Agency estimates servers to account for more than 60% of energy demand coming from data centres. The share rises to almost 80% for servers in hyperscale data centres. The latter also require billions of litres of water annually, for cooling the chips used for computing. Water is needed for energy generation as well.
Looked at in the context of India becoming an important destination for building AI infrastructure and large data centres, the above conditions highlight some concerns.
Triple Threat
First, electricity demand by AI data centres pressurises electricity grids. The state grids will need to ramp up overall output since supplies to households, agriculture, and other industries can’t be compromised. Electricity producers will need to expand capacities for generating more power, bringing to head a critical problem: can more electricity be generated without resorting to greater use of fossil fuels? The ostensible answer is no. Without using more coal and natural gas the electricity output can’t be increased making for carbon footprints an unavoidable outcome.
The data centres in India involving the US big tech — Google and Meta — that are coming up in Visakhapatnam in Andhra Pradesh and Jamnagar in Gujarat are to run on captive renewable energy. They might also go off-grid for fetching power directly from renewable producers. While this can limit additional emissions by containing use of fossil fuels, it can also create different incentives for renewables: independent, especially smaller producers of the latter, might be more incentivised to sell off grid to data centres instead of national or state grids. Furthermore, not all large data centres will have the capacity to internally finance captive renewable capacities. They will have to rely on the common grids, thereby making use of more fossil fuels for generating electricity an almost unavoidable option.
Second, where will the water for cooling servers come from? The proposed Meta AI facility in Jamnagar is expected to use desalinated sea water for cooling. This can ensure preservation of water supplies for local use. Other data centres coming up close to coasts might be able to do the same, unless they take to more expensive techniques like liquid cooling of AI chips. This, however, brings up another important issue in the development trajectory of data centres. Coastal states of the country are likely to attract more data centres due to the advantage of using sea water. As it is, coastal states like Andhra and Gujarat are closer to underwater subsea cables that make them digitally more attractive for US big-tech promoted data centres. Might this generate a hinterland-coast divide in concentration of AI centre investments?
Third, where will the land for data centres come from? The electricity capacities of data centres in the form of power generators and substations, and water-cooling systems, require large spaces. The US experience of building these data centres shows that the hyperscale ones can require up to as much as 1,000 acres or more. Such tracts of barren land are not easily available in coastal Indian states. Might it mean that farmlands, and ecologically fragile land precious for local communities, need to be taken up for the data centres?
Globally, discontent against AI data centres is rising because of the pressures they are generating on energy, water, and land. India is walking a delicate path in this regard. It is yet to be hit as badly on jobs by AI as several other countries of the world, including the Global North. The hit, though, might just be around the corner, with many white-collar tech jobs awaiting the chop. Indeed, several businesses are citing AI for justifying layoffs. Coupled with the stress put on scare and sustainable natural resources, the employment squeeze can make AI politically unacceptable.
It is not just Pope Leo who’s called for disarming AI. Conservative ideologues are describing AI as new immigration with serious cultural, social, and economic consequences and urging action to stop its progress. Tech experts spelling out virtues of AI are being protested in global university campuses. The politics of AI is assuming highly sensitive proportions.
India needs to be careful in managing the economic downsides of AI infrastructure growth. The consequences might aggravate the political backlash on AI.
Disclaimer: The views expressed are the author’s own and do not reflect the official policy or position of Financial Express.