AI’s Energy Demands: A Sustainability Dilemma

The Rising Energy Consumption of AI

To answer a query, ChatGPT uses 10 times as much electricity as a search on Google. This report by Goldman Sachs doesn’t just shed light on AI’s burdensome energy demands, it questions the very rationality of using AI as frequently as it is, and more importantly - the rationality of investing hundreds of billions of dollars in developing data centres for it.

AI’s Potential in Sustainability

I need not mention the rising capabilities and opportunities AI presents, as I’m sure all of us are, to a certain extent, informed about this. Also among these possibilities lies the chance to make significant strides in reducing emissions. Through the implementation of AI-driven research and autonomous renewable systems, the race towards net-zero carbon emissions may be accelerated remarkably.

The Energy Challenge: How Are AI Data Centres Powered?

Nonetheless, the problem of how to power the data centres required for AI to function and satisfy the thousands of queries it receives remains. At the moment, we use sources like coal to power these data centres, which require around 100 MegaWatts of power, enough for 400,000 electric cars! Some think that as data centres develop rapidly, renewable energy will grow tremendously to fulfil their energy demands. But what if sources like coal just continue to be utilized instead?

The Problem of Availability

Well, this is a very real possibility. Why? Two reasons - Firstly, availability. Most, if not all, of these data centres need to be located in cities or urban landscapes (as that is where most IT/computing companies are). But here’s where the problem comes in: these data centres need huge chunks of energy to operate, and these can be provided only by a dedicated grid from wind and solar power plants. The catch? These power plants are located far from cities, and need to be so to efficiently capture energy.

The Reliability Factor

Second, reliability. Renewable energy sources aren’t reliable enough to self-sufficiently power a data centre. In fact, they can at most, just play a supplemental role to traditional sources.

The Future of AI and Sustainability

So we end up with a scenario where AI has the potential to revolutionize sustainability through its expertise, but using AI to do anything, for that matter of fact, is hugely harmful to the climate. How exactly will this play out? Will the world just ignore sustainability and climate concerns in the face of short term progress and advancement, or will we figure out a way to make these data centres, and hence the usage of AI, more eco-friendly?