How much do your ChatGPT prompts impact the planet?

Nowadays, ChatGPT and other AI tools have become a ubiquitous part of everyday life. From drafting emails to coming up with quick weekday recipes, AI has taken over our daily routines. The aforementioned benefits of AI as an accessible tool for various types of aid come with heavy environmental costs. The processing of our AI requests is housed in vast data centres that require substantial amounts of electricity and water to operate effectively. The environmental cost of AI often goes unnoticed by many but has a larger impact than we might consider.

When accounting for the water usage for both direct cooling and electricity generation, it is estimated that a singular ChatGPT prompt uses up to half a litre of water

Every ChatGPT prompt consumes roughly 0.0025 kilowatt-hours (kWh) of electricity, which is ten times the energy used for a typical Google search. What seems like a minor difference becomes alarming when multiplied by ChatGPT’s massive usage and layered onto an already overburdened environment. ChatGPT’s daily consumption of energy is estimated to be an alarming 39.98 million kWh – enough to charge eight million smartphones. In addition to this, the global electricity demand from these data centres is predicted to at least double by 2030, according to the International Energy Agency (IEA).

The imminent rise in the demand for AI further necessitates conversations addressing the environmental implications of these AI technologies. The energy consumption of AI operations can also be translated into carbon emissions to contextualise the impact on climate change. Every ChatGPT query costs an estimated 4.32 grams of carbon dioxide, making its monthly carbon emissions comparable to that of 260 transatlantic flights. Another overlooked aspect of these large AI models in terms of energy consumption is their training phases. Training GPT-3, for example, consumed around 1,287 megawatt-hours of electricity and produced 552 tonnes of CO₂.

On top of electricity consumption, the water required for cooling the servers in data centres proves to be quite expensive. When accounting for the water usage for both direct cooling and electricity generation, it is estimated that a singular ChatGPT prompt uses up to half a litre of water. Globally, AI operations are predicted to consume between 4.2 and 6.6 billion cubic meters of water annually by 2027 – which is comparable to the water usage of Denmark. The outsized demand for water and energy in AI operations underscores the imperative for sustainable solutions moving forward.

Simple initiatives like pushing for the development of energy-efficient algorithms and hardware or the implementation of reuse technologies in data centres to mitigate water consumption can make a world of difference

In addition to resource consumption and emissions, the environmental impact of AI also poses some ethical and societal considerations. The placement of these data centres commonly affects marginalised communities as they can be a massive strain on local resources. The Microsoft data centres in West Des Moines, Iowa contributed to 6% of the area’s freshwater consumption in just a month. Such resource allocation raises critical questions about the burdens of technological innovation and highlights why those making these decisions often remain indifferent – they are often not directly affected. The lack of transparency in making these decisions brings forth a recurring problem with big tech firms: their limited accountability as a result of a reduced public awareness about their operations.

The journey to addressing the environmental effects of AI usage is a long but rewarding one. Informing the public and demanding that large companies provide more transparent reports about their consumption in this regard is vital. Simple initiatives like pushing for the development of energy-efficient algorithms and hardware or the implementation of reuse technologies in data centres to mitigate water consumption can make a world of difference. A precedent for the results of this form of call for transparency is how the LEED certification system – a framework for the reporting of energy and material efficiency in architecture and construction – pushed for the prioritisation of greener practices through the motivation of market competitiveness and legitimacy.

In essence, as AI technologies like ChatGPT become increasingly woven into our daily lives, it’s imperative to recognise and address their environmental implications. We must ensure that the pursuit of innovation does not come at the cost of our planet’s health.