If India is to achieve its data centre ambitions, it needs to find new sources of efficient power. While nuclear is a solution, it will take too long to come on stream. We need alternatives, and enhanced geothermal might just be it.

This is a link-enhanced version of an article that first appeared in the Mint. You can read the original here. For the full archive of all the Ex Machina articles, please visit the website.

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. This will call on us to 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.

The Nuclear Solution

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.

As glad as I am that we will finally remove hurdles in the way of building nuclear power plants in the country, we cannot rely on nuclear energy alone to solve our data centre problem. It takes anywhere from eight to 10 years to build even a small modular reactor, and that is way more time than we have. We need energy solutions that scale in years, not decades—and offer the reliability of nuclear energy with none of the risk, the cleanliness of renewables without intermittency, and the locality of distribution without any need to first acquire large tracts of land.

Enhanced Geothermal

One solution that fits the bill is enhanced geothermal energy. Unlike its conventional counterpart (which uses heat from naturally occurring hot-water reservoirs near the earth’s surface), enhanced geothermal systems (EGS) use sophisticated drilling techniques to create reservoirs deep below the surface of the earth where rock layers are dry and the temperature can go as high as 200° Celsius. Water is then pumped into these man-made reservoirs, and, once it gets converted into steam, is used to drive turbines that generate electricity.

To be clear, the upfront development cost of an EGS plant does exceed that of a solar or a wind project, and there is also a not-insignificant exploration risk in finding an ideal location to sink a geothermal well. Even though the development timeline for EGS plants—typically 3-5 years from exploration to production—is shorter than nuclear, it is still much longer than the time it typically takes to get a solar or wind project up and running.

That said, this technology is highly efficient, with a capacity factor of 90% (as compared to 25-40% in wind and solar). What’s more, EGS projects use far less land than other renewable energy projects, producing as much energy as a solar farm from roughly 1% of the land. This means these plants can be built in and around urban centres that have easy access to high-capacity fibre networks and various telecom providers. This is perfect for data centres that need high-quality, uninterrupted power supply and connectivity that is good enough for low-latency applications.

A number of regions in India have been identified as being suitable for EGS plants. This includes the Puga Valley in Ladakh, the Cambay basin in Gujarat, the Godavari Basin in Andhra Pradesh and Telangana and parts of the western coastal regions of Maharashtra and Gujarat. Among these, the best location for a well that supports data centres might be somewhere along the west coast of the country, where most of our cable landing stations terminate.

In 2016, the Union Ministry of New and Renewable Energy published a draft geothermal policy suggesting that geothermal energy would form a significant part of India’s long-term energy mix. Not much has been done about this since then, which is unfortunate, given that Indonesia, with similar resources, is already generating 2.3GW of geothermal power.

EGS for AI

Our AI ambitions depend heavily on our ability to augment our data centre capacity rapidly. This, in turn, depends on us making sure that we have an abundance of clean and reliable power to do so. EGS presents a rare alignment of technological, environmental and strategic feasibility, and we should take concrete steps to make it happen.

The first thing we need to do is resurrect the 2016 draft geothermal policy and allow companies to set up captive EGS plants to power their data centre facilities. If the government were to subsidize the cost of geothermal exploration, that would help us identify suitable locations for these plants. Done right, we should be able to create data centre corridors in and around our geothermal hotspots and then leverage our cable landing stations accordingly.

If we start now, by the time these facilities come on-stream in 3-5 years, the accelerated expansion of our data centre capacity will be ready to absorb their output.