Experts warn:

AI technology consumes as much electricity as entire countries!

AI-generated symbolic image of a nuclear power plant: recognizing and generating patterns in the data center is a power-hungry undertaking.

(Bild: KI-generiert – stock.adobe.com/Ben)

Artificial intelligence (AI) is on the rise - and so is electricity consumption, as the technology requires extremely powerful data centers. AI applications already consume as much electricity as entire countries. Experts warn of the consequences for the environment.

As an end user, you often don't even notice or realize that energy is consumed every time you use a search engine or surf the internet. This is because behind the websites are huge data centers with powerful servers that consume a lot of electricity to operate. They heat up and need to be cooled, which further increases energy consumption.

Why does AI need so much energy?
Data centers that run artificial intelligence applications require significantly more power, as the computing power required for AI is enormous. For comparison: a query from ChatGPT or another generative AI requires on average around ten times as much energy as a simple Google search.

A data center of the US internet company Google

(Bild: AP)

According to the International Energy Agency (IEA), data centers consume 40 percent of their electricity for operation and the same amount for cooling. However, normal data centers can be operated in rooms with good air conditioning and cooled sufficiently, but AI applications are so powerful that they often have to be cooled with water, says Fabrice Coquio from the data center service Digital Realty.

How much energy does AI need?
Before the triumph of artificial intelligence, data centers were responsible for around one percent of global electricity demand. The IEA assumed that data centers, cryptocurrencies and AI applications would consume 460 terawatt hours in 2022, which corresponds to a share of two percent. This share could double again by 2026. Artificial intelligence would then consume as much electricity as Japan.

And it's not just the servers and data centers that are driving up the energy consumption associated with AI. There is an entire industry behind the technology that produces chips and manufactures server capacity. Researcher Alex de Vries, who runs the online portal Digiconomist, looked at the consumption of the company Nvidia, a world leader in AI hardware. In his calculations, he came up with a consumption of 85.4 to 134.0 terawatt hours of electricity - that's how much Argentina or Sweden consume in a year.

How does this fit in with climate targets?
Major players in the AI sector such as Amazon, Google and Microsoft are trying to reduce their carbon footprint by purchasing large amounts of electricity from renewable energy sources. Amazon's cloud provider AWS claims to be the world's largest buyer of renewable energy.

AWS aims to be climate-neutral by 2040, while Google and Microsoft want to achieve this goal ten years earlier. However, building new data centers will not help. On the contrary: Google and Microsoft recently declared that their greenhouse gas emissions have continued to rise in recent years. Google reported a 48% increase since 2019, Microsoft a 30% increase since 2020.

Both companies blamed the AI boom for this. Microsoft President Brad Smith told the Bloomberg news agency in May that the promise of climate neutrality was made before the "AI explosion". Today, the target is five times further away than in 2020.