How Energy Intensive Are Data Centers?

In 2022, the global electricity consumption stood at roughly 27,000 terawatt-hours, according to the U.S. Energy Information Administration (EIA). While countries like China and the United States had the highest electricity demand by a considerable margin, a new player that rivals the consumption of some of the biggest economies and is bolstered by the ongoing AI hype has entered the stage: data centers. As the International Energy Agency (IEA) points out in a recent report, estimated data center energy consumption could increase rapidly over the next two years.

While the median estimate for data centers, AI-connected offerings and cryptocurrency mining stood at 415 TWh in 2022, already significantly higher than the United Kingdom's electricity demand and on par with the electricity consumption of France, this figure could rise to anything between 620 and 1,050 TWh in 2026. Per the IEA's report, this would be the equivalent "to adding at least one Sweden or at most one Germany" to the most recent estimates. Calculating the median estimated energy use, data centers and associated industries would rival the electricity demand of Japan, which as of 2022 had the fifth-highest in the world after China (8,540 TWh), the United States (4,128 TWh), India (1,463 TWh) and Russia (1,026 TWh).

Even though data centers are getting more energy-efficient and are required by law to do so in some countries like Germany, the increased demand fueled by the AI hype cycle might be too high to be fully mitigated. This is especially true for big tech companies like Alphabet, Amazon or Meta. The most recent Digital Economy Report by the UN suggests that data centers run by GAMAM companies alone were responsible for energy consumption of upwards of 90 TWh, more than countries like Finland, Belgium, Chile or Switzerland.

Alphabet's CO2 emissions, which are largely connected to an increase in data center activity, have risen by 48 percent compared to 2019. The company claims that "reducing emissions may be challenging due to increasing energy demands from the greater intensity of AI compute, and the emissions associated with the expected increases in our technical infrastructure investment" in its 2024 Environmental Report. Microsoft, which partnered with OpenAI to integrate its large language models into all of its products, reported CO2 emission increases of 30 percent compared to 2020.