AI Now Uses More Electricity Than Bitcoin

Artificial intelligence is now using more electricity than Bitcoin mining, making it one of the most power-hungry technologies in the world. A study published in December 2025 in Patterns by Alex de Vries-Gao from VU Amsterdam estimates that global AI systems could use up to 23 gigawatts of power in 2025. This is more than the total electricity Bitcoin mining used in 2024, showing how fast AI is growing in energy demand.

Bitcoin mining has long been criticized for using a lot of energy because of proof-of-work calculations. It used around 170 to 200 terawatt-hours of electricity each year recently, similar to the electricity use of countries like Thailand or Argentina. AI uses energy differently. Its demand comes from huge data centers running specialized hardware like GPUs to train and run large language models and other AI programs. The 23 gigawatt estimate could mean dozens to hundreds of terawatt-hours of electricity each year, depending on how the hardware is used and how efficient it is.

De Vries-Gao made these estimates using public information. He looked at hardware production numbers, earnings reports from big tech companies like Google, Microsoft, and Amazon, and analyst studies. He warns that the real numbers are probably higher because companies rarely report exactly how much electricity goes to AI.

AI also affects water use. Data centers need large cooling systems, which could use between 312 and 765 billion liters of water in 2025. That is about the same as all the bottled water people use in the world in a year. AI could also cause 33 to 80 million tons of CO₂ emissions each year, about the same as a small country produces.

This situation has caused debates with activists, tech enthusiasts and policymakers. Some Bitcoin miners are switching to AI tasks because they can earn more. Tech companies are also looking at nuclear power and ways to make data centers more efficient. But AI is growing so fast that it is putting pressure on electricity grids and natural resources worldwide.

As AI becomes a bigger part of everyday life, these hidden costs make us ask important questions about sustainability and what it really costs to run intelligence at a global scale.