​Clean AI: on AI data centres and energy demand

The International Monetary Fund’s report that pointed to the likelihood of the economic gains of Artificial Intelligence (AI) applications outweighing the environmental costs of the increased energy demand that AI data centres will require is reassuring. It underscores that this transformational technology is not fundamentally at odds with the global imperative to pursue sustainable growth strategies across the board. Countries that are better prepared with renewable energy generation are bound to see a lower social and environmental cost to pursuing their AI ambitions. India’s AI infrastructure — at least the part of it that the government is indirectly funding through the IndiaAI Mission — does not rise to the level of weighing at a macro level on the nation’s energy mix. Still, the need for pursuing renewables specifically for AI is necessary to follow. This is already in a sense the government’s approach to the issue, outlined at the AI Action Summit in Paris earlier this year. While AI is not the sole industry where a push for renewable energy and sustainable practices is important, the sector nevertheless offers itself up for two main reasons. The first is the sheer volume of electricity that it is set to consume. The IMF’s report indicates that in the United States — the single largest home for AI computing capacity globally — “AI expansion alone could increase electricity prices by up to 9 percent, adding to price pressures coming from many other sources”. As such, renewable energy could play a major role in cushioning against a multifold rise in emissions that conventional energy sources would likely entail. The second is that data centres lend themselves uniquely to captive renewable infrastructure. Some Indian firms have already made moves to purchase renewable energy, and the hundreds of acres that data centres occupy are ripe for complementing equipment with solar cells. Nuclear energy may also turn out to be a welcome contribution: small modular reactors at emerging data centre clusters, in conjunction with other renewable sources, would avert a sizeable quantity of emissions.

Electricity use is not the only environmental footprint that the AI age will leave behind — the technology requires large-scale mining of minerals and water use, and produces effluents in the manufacturing of the electronics supply chain. As electronics manufacturing is another space that India aspires to grow in, these opportunities must be pursued in this area too. India’s net zero target for 2070 will demand a scaling down of conventional sources of emissions, and managing the rise of industries that are poised to expand their share of global energy consumption.