2.8 Gigatons of CO₂ by 2031 Due to AI Expansion - If AI Is Not Embedded in More Eco-Friendly Processes

Artificial intelligence consumes a lot of CO₂ but could reduce emissions in the long run. A new study reveals that without embedding AI in eco-friendly processes, we may face significant environmental damage.

Home
News
Future & Science

Add t3n as a preferred source on Google

Don't miss any news!

Artificial intelligence consumes a lot of CO₂ but could reduce emissions in the long run, it is often said. However, by the time that happens, it might already be too late. This is the conclusion of a new study. Yet, there is still hope.

AI as a Solution to Climate Change? Study Contradicts

In the long term, artificial intelligence can contribute to reducing CO₂, a common argument goes. After all, AI could help analyze processes and measure its own CO₂ footprint, it is often stated.

A study by climate scientist Yassine Charabi published in the journal Communications Earth & Environment has critically analyzed this thesis. The gist of the study: By the time AI balances emissions annually, we may have already accumulated total damage that is hardly repairable. At least not if we want to achieve the goal set at the UN Climate Conference in Paris in 2015 to limit human-caused global temperature rise to 1.5 degrees Celsius.

Mathematical Model Simulates Future Scenarios

For his study, Yassine Charabi designed a mathematical model. He fed data such as global energy forecasts, growth rates of data centers, and timelines for hardware replacement into it. The model also calculated daily electricity consumption and its CO₂ costs as well as the emissions generated during the production of AI chips.

He compared these figures with the scenarios of the International Energy Agency regarding artificial intelligence and energy up to the year 2035.

Average CO₂ Emission of 2.8 Gigatons

Yassine Charabi conducted simulations with his model over 10,000 times. On average, these resulted in a CO₂ emission of 2.8 gigatons.

According to the study, artificial intelligence will only be able to offset the impacts caused by its production by the end of 2031. This is in the best, or fastest, case scenario that the simulations produced. However, before that, AI technology remains in a realm that Charabi calls "Carbon Valley". A valley where the advancement of AI systems consumes more carbon than it can save.

Every Year of Hesitation Costs 0.45 Gigatons of CO₂

"Efficiency improvements alone do not guarantee an absolute decoupling between the expansion of AI and electricity demand in rapid expansion scenarios," the study states.

However, according to Charabi, there is one sector where the use of AI is worthwhile. Namely, in the integration of artificial intelligence into eco-friendly technologies. Every year of hesitation to invest here costs 0.45 gigatons of CO₂.