When the monsoon fails, India's AI dreams fail with it

**Snigdha Sharma** (0:01)
Yesterday morning, the Indian Med Department's Daily Bulletin showed the southwest monsoon's northern limit passing through a small coastal town in Maharashtra called Harnai. It's been there for days. And just a few hundred kilometers north, Mumbai is still waiting. The city's reservoirs are already down to 12% of capacity. Multiple news reports say that the city only has 45 days of water reserve left.
Its normal monsoon arrival date was June 11th. Now, predictions say that it is going to come by the 20th of June. It is the worst start to a monsoon season that Mumbai has seen in nearly two decades. And Mumbai is not alone.
Nationally, India is carrying a 35% rainfall deficit for this season. Between the 4th and 15th of June, the country received barely a third of the rain that it normally would.
A deficit like this does not stay confined to a weather report. More than 40% of India still depends on agriculture for its livelihood, which means the consequences of a weak monsoon ripple outward throughout the entire economy. They will show up in Kharif sowing numbers in August and in the hydropower output throughout the winter. And also in the groundwater reserves that cities like Bangalore and Mumbai will be drilling for next summer. And this matters because groundwater is no longer just sustaining farms, households and cities. It is increasingly being asked to sustain India's digital future. And these two cities most exposed to future water stress, Bangalore and Mumbai, also happen to be at the center of India's nearly 180 billion data center investment push. Last year alone, data centers in India consumed roughly 150 billion liters of water. By 2030, this figure is projected to more than double. And water is only half the story. Data centers also need uninterrupted electricity. And India's power system is more dependent on the monsoon than most AI investment presentations would suggest. In 2023-24, erratic rainfall contributed to more than a 16% drop in hydroelectric generation. And this was the steepest fall in nearly four decades. The shortfall was made up by a record coal consumption. And all of these strains are already becoming visible. Bangalore is extracting groundwater at 177% of what nature can replenish. Earlier this year, Karnataka's own IT minister told the state assembly that hyperscale data centers may not even be suitable for the city. In neighboring Maharashtra and Telangana too, data centers are legally classified as essential services, which means that in a scarcity scenario, it is possible that they are prioritized for water and for power over residential supplies.
And this brings us to a question that rarely appears in investment announcements or in budget speeches. You see, India is currently holding two visions of itself at the same time.
One is immediate and deeply familiar, which is a monsoon dependent economy where a bad rain year can mean crop losses, debt, migration, and in turn, political instability.
The other is aspirational. An AI-powered economy racing to build the infrastructure needed to get a seat at the table where the future is being decided. And it will do anything for this. If you look at them separately, maybe both of these visions do make sense. The problem is that they may not be compatible. What happens when the infrastructure required to power India's AI ambitions depends on the same water and same energy systems that are becoming increasingly vulnerable to climate change.
India's AI ambitions and its climate reality are on a collision course, and almost nobody wants to say it like it is.
Welcome to Daybreak, a business podcast from The Ken. I'm your host, Snigdha Sharma, and I don't chase the news cycle. Instead, every day of the week, my colleague Rachel Varghese and I will come to you with one business story that's worth understanding and worth your time. Today is Thursday, the 18th of June.
India has staked its AI future on infrastructure, specifically on becoming the data center hub that global AI runs on, because the frontier model race has been consolidated around American and Chinese labs, and the chip fabrication race belongs to Taiwan and the Netherlands. So, our only realistic shot at global AI is the physical layer underneath. The land, the power, the water, and we have land, plus a large engineering workforce and a government that is willing to move fast. Which is why the bet, the one that $180 billion in investment is riding on, is infrastructure. Become the data center hub that global AI runs on and capture value as the world's AI backbone. It is a credible bet, but it has a physical requirement that investment announcements tend to skip over.
As we already know, data centers are like gigantic hungry monsters of the future. They run 24 hours a day without interruption, consuming enormous quantities of water for cooling and electricity for computing. Last year, data centers in India consumed 150 billion liters of water, which is a figure that is projected to more than double by 2030 To give you a clearer picture, the Google Adani AI campus that is planned for Viswaka Patnam at 1000 megawatts of plant capacity would, when fully operational, consume roughly as much electricity as one-fifth of Bangalore's entire annual power supply. So these are the level of demands that are being layered onto a grid and on a water table that is already under stress. And that stress has a single point of origin, the monsoon. India's power grid has always been monsoon dependent. When reservoirs are full, hydroelectric generation is reliable and coal consumption stays in check. But when monsoons fail or arrive late, hydro output drops and coal fills the gap, at higher cost and with greater emissions. In the fiscal year 23 to 24, erratic rainfall caused more than 16 % decline in hydroelectric generation, which was the steepest drop in nearly four decades. This year, electricity demand hit an all-time record high of more than 270 gigawatts on the 21st of May, which is the highest for our country ever.