The energy consumption of the world’s data centers is expected to double in the next two years. Above all, due to the development of cryptomining and artificial intelligence, explains the latest report from the International Energy Agency (IEA).
2% of global electricity demand in 2022 came from data centers, including those used to mine cryptocurrencies and artificial intelligence. Together, they needed an estimated 460 terawatt-hours (TWh). Almost a quarter of this consumption was taken by crypto mining alone, points out the annual report on electricity from the IEA.
The demand for data centers had been representing a more or less stable 1% of global energy since 2010, according to a study published in October last year in the journal Joule. THE IEA recognizes that The rapid advancement of technological developments complicates projections. However, he maintains that artificial intelligence and crypto mining could make the electricity consumption of these facilities will range between 620 and 1,050 TWh by 2026.
To put it in perspective, this additional spending is like adding up all the consumption of countries like Germany or Sweden, the agency points out. The final size of the increase will depend on the pace of implementation, the extent of efficiency improvements and new developments.
“The rapid expansion of the data center sector and high demand for electricity may pose challenges for the electricity system,” the IEA says in its report. “Improvements in efficiency and regulations will be crucial to curb consumption.”
The leap in consumption of artificial intelligence in data centers
40% of its data center electricity demand is used for computing operations. Another 40% is used to cool all that equipment. The remaining 20% is related to other associated information devices.
The IEA estimates that consumption specifically related to crypto mining will grow at least 40% by 2026 —making the comparison with last year—. There are still some uncertainties regarding the adoption of this type of currency and its supposed methods to save energy, the agency insists.
For example, he mentions the case of Ethereum, the second largest cryptocurrency on the market, which reduced its electricity demand by 99% in 2022 after changing its mining mechanism. The opposite is true for Bitcoin, which has not wanted to do the same and is responsible for more than 90% of the energy consumption of crypto mining. In the end, “energy savings may be negated by increases in other energy-consuming operations, such as other cryptocurrencies, even as some become more efficient,” the IEA explains.
The projection for artificial intelligence is much greater: could multiply by 10 in the next two years. The calculation takes into account the number of artificial intelligence servers that are estimated to be sold in the future and their nominal power. NVIDIA, the current market leader in artificial intelligence processors, could be shipping around 1.5 million units by 2027, even if its market share declines, as expected.
The impact of a key player: Google
The IEA makes special mention of the impact that relevant market actors could have. Among them, Google. The total electricity consumption of the search giant was 18.3 TWh in 2021. Artificial intelligence then represented between 10% and 15% of the total.
Now that Google is incorporating artificial intelligence into most of its products, it is estimated that this consumption will skyrocket. John Hennessy, president of Alphabet (Google’s parent company), told Reuters in February last year that interacting with a large language model (LLM) could “probably cost 10 times more than a standard keyword search.” A Google search uses 0.3 Wh of electricity. If necessary, the calculation for each search powered by artificial intelligence would be 3 Wh.
“The worst-case scenario suggests that Google’s artificial intelligence alone could consume as much electricity as a country like Ireland,” warned an analysis by Alex De Vries, founder of Digiconomist, a company dedicated to researching digital trends.
The key could be in how technology companies manage to take advantage of the momentum of clean energy to power their data centers. The IEA reported that clean energy grew by 50% in 2023, reaching almost 510 gigawatts (GW). It is the highest peak in the last 20 years.