Let modern geothermal energy power our AI ambitions

Such units could come up faster than nuclear plants and satisfy the huge appetite of data centres

- RAHUL MATTHAN

If data is the fuel that powers the digital revolution, data centres are the tanks in which they are stored, primed and kept ready for use when needed. This is why any country looking to play a role in the digital future of the world has no choice but to make sure it has the capacity of a data centre to support that ambition.

Today, India's operational data centre capacity stands at about 1.25 gigawatts (GW). This is likely to grow to 3.4-3.5GW by 2030 at best. We trail far behind the US, where, in Ashburn, Virginia, alone, there is 4.5GW worth of data centre capacity that is on track to grow to 15GW. If India wants to be competitive in artificial intelligence (AI), it must scale up to at least 5GW by 2030 and 35GW by 2047. We must not only navigate the current global shortage of graphics processing units (GPUs) and implement significant policy reforms, but also find sufficient sources of clean energy to power these facilities.

Most experts believe nuclear energy is the answer. Of all the commonly available sources of power, this offers the sort of high-density, zero-emission baseload power that data centres require. This is why many AI companies in the US have begun work on restarting decommissioned nuclear power plants. It is also why so many believe that India should follow suit. When the finance minister announced in this year's budget speech that the government was going to amend both the Atomic Energy Act and the Civil Liability for Nuclear Damage Act, it was widely believed that a major impetus for this long-awaited reform was the digital ecosystem's power demands.