Will Artificial Intelligence Solve or Actually Worsen the Climate Crisis?

Artificial intelligence (AI), in the form of generative AI, has reached the masses at breakneck speed. Whereas it took TikTok nine months to reach 100 million users, ChatGPT by comparison needed only two months to do that. This has resulted in additional electricity consumption by the data centers that power AI.

At the same time, enormous hopes are riding on AI. Many experts expect a lot from the new technology, including solutions for tackling the climate crisis. So, the big question is: Will AI help to solve the climate crisis, or will it make it worse?

In December 2024, ChatGPT had over 300 million weekly users. They send approximately 7 billion queries to the large language model (LLM) each week. That takes electricity and accordingly causes emissions.

According to a calculation by Piktochart, each ChatGPT query produces approximately 4.23 grams of CO₂. The range of estimates varies widely depending on the assumptions applied. But let’s stay with this numerical example. Extrapolated to the 7 billion weekly queries, this means that ChatGPT produces around 1.5 million tons of CO₂ per year, and the trend is rising.

In comparison, an average internal combustion engine (ICE) car emits approximately 4.73 tons of CO₂ per year, according to the United States Environmental Protection Agency. This means that at the current user count, ChatGPT queries alone emit as much CO₂ per year as 350,000 ICE cars do.

Generative AI increases the amount of electricity consumed by data centers for a number of reasons. For one thing, it takes vast quantities of data to train an AI model. During training, the model learns how to behave in a wide variety of situations based on a large set of data and examples. Depending on the size of the model, the training process can take from a few minutes to several months and requires special hardware that works continuously the entire time. Training a large-scale model can consume as much electricity as thousands of households do in a month.

But using the finished product also consumes resources. For every single query, billions of computations are performed on high-performance servers located in large data centers. Since they have to be cooled around the clock, not only is electricity consumed, but also water.

Data centers in their full scope of services currently account for 1% to 2% of the world’s total electricity consumption. That electricity can originate directly or indirectly from fossil fuels. If one puts credence in a study by Goldman Sachs, that percentage of total power consumption will rise to around 3%–4% by 2030. The projected increase primarily owes to generative AI.

At the same time, huge hopes are being pinned on AI to help fight the battle against climate change. AI is already being deployed in a wide array of different ways today in this context. Below is a short list of just a few of those applications:

• Weather Forecasting
  AI is used in climate modeling and weather forecasting. This results in more accurate forecasting of mounting extreme weather events such as storms, droughts, or floods. Moreover, the forecasting helps to organize renewable energy sources more efficiently. Precise weather data and historical patterns can be used to minimize energy grid overloads, but also overcapacities at solar or wind power plants, for example.
  
  
• Building Operation Technology
  Did you know that the operation of buildings accounts for around 30% of worldwide electricity consumption? Here, too, AI can help to recognize patterns and save energy wherever it is not needed. At the same time, AI can better predict energy demand – on the basis of weather forecasts, for instance –, facilitating the integration of renewable energy. This “smart buildings” technology enables buildings to optimize their energy consumption themselves to increase energy efficiency.
  
  
• CO₂ Footprint
  Businesses can use AI to measure their respective CO₂ footprints more accurately. AI can help here to collect and analyze data from a wide range of sources, such as from logistics operations or supply chains. That helps to identify and eliminate weak spots. In addition, precise AI models can estimate emissions, point out savings potential, and simulate scenarios to devise sustainability strategies.
  
  
• Meltwater in the Arctic and Antarctic
  AI is also already being deployed in the Arctic and Antarctic. With the aid of this technology, researchers can ascertain how much ice is melting and when and where it melts. That helps to quantify how much meltwater gets released into the ocean.
  
  
• Agriculture
  Another application area is agriculture. Here, AI analyzes data on weather, soil quality, and crop conditions, thereby increasing the efficiency of tillage, pest control, and harvests while at the same time protecting soil from erosion and overuse.

Those are just a few of the many examples of how AI is being employed to make the world more sustainable.

AI is an extra consumer of electricity, and that is not going to change for the time being. In recompense, though, AI can contribute, for instance, to making the management of data more sustainable. The onus here is mainly on businesses to get a handle on “dark data” that companies collect but don’t utilize. In a survey of 1,300 businesses conducted by software provider Splunk, 60% of the respondents disclosed that they do not utilize more than half of their data. A lot of energy gets consumed to needlessly store it. Businesses have a duty here to manage their data with greater awareness.

In 2023, around 30% of the world’s electricity came from renewable sources, up from 20% in 2012. The trend is thus headed in the right direction. The question, though, is can it keep pace with the rising consumption? The primary problem with huge electricity consumption is fossil fuels. Hypothetically, if all electricity worldwide were derived from renewable sources, AI and other resource-intensive technologies like cryptocurrencies, for instance, admittedly still would not be emission-free (e.g. due to the construction of data centers and cooling with water), but would have a vastly better climate footprint. Some studies say it is realistic that up to 100% of worldwide energy production will originate from renewable sources by 2050.

Other studies, however, do not consider that feasible. What is clear, though, is that technologies like generative AI raise global electricity consumption. So, switching to renewable energy sources is a priority now more than ever.

Whether AI worsens or alleviates the climate crisis depends on how we utilize it. While AI in its present form contributes to increased electricity consumption, its potential lies in targeted deployment to improve energy efficiency and devise sustainable solutions. Businesses, governments, and private individuals are called on in equal measure to use the technology conscientiously and responsibly.