It was meant to revolutionize finance, and free us all from the tyranny of centralized control of our banking. But it turns out, Bitcoin just might end up killing our planet.

That’s the findings of a new study by Alex de Vries, a researcher at the Vrije Universiteit Amsterdam, who has put forward an estimate for how much water is consumed every time someone buys, sells, or mines Bitcoin on the blockchain.

Writing in the journal Cell Reports Sustainability, de Vries calculated that in 2021 mining Bitcoin consumed more than 1,600 gigaliters of water worldwide, with every single transaction on the blockchain using 16,000 liters of water—6.2 million times more than is used every time we swipe a credit card to pay for something.

Water is used in these transactions to cool the computer equipment used to mine Bitcoin, as well as to cool the power plants that supply the electricity for those computers. “In order to generate electricity, you also need a lot of water to cool the fossil fuel-based power plants,” he explains.

Only between 10% and 20% of the water usage involved with every Bitcoin transaction is direct—that is, used to cool the equipment that actually mines the Bitcoin, says de Vries. The remainder is the transaction’s indirect footprint, where water is used elsewhere along the supply chain.

“Water footprint is a very important, but often overlooked, environmental cost of computing systems, including AI and Bitcoin,” says Shaolei Ren, associate professor of electrical and computer engineering at the University of California, Riverside, who tries to track the water impact of various technologies. Ren was not involved in the Cell Reports Sustainability study.

Ren welcomes the paper putting forward a concrete estimate of the overall water footprint of Bitcoin. “The operational water footprint for server and power-plant cooling is already huge,” he says. “If we further factor in the embodied water footprint for manufacturing Bitcoin servers, the total water footprint could be easily doubled. But, at this point, there’s been very little, if any, data available for the embodied water footprint.”

However, Ren points out that the calculations put forward by de Vries might actually underestimate water usage in one comparison: The paper compares the water footprint of Bitcoin in the United States to the water usage of 300,000 U.S. households—shocking enough. But Ren believes that household water usage has been conflated with household water withdrawal. Households only consume around 10% of their water withdrawal—meaning that the equivalent number of households should be 10 times higher, or three million households.

Regardless of the numbers involved, de Vries believes it’s incumbent on the crypto industry to do better—both as a collective, and as individual miners. “The miners themselves can bring down their own water consumption by relying on cooling methods that don’t use that much water,” he says, pointing to immersive cooling, where computer equipment is cooled using special fluid, not water. How you source your power is also important, de Vries adds: Wind, solar, and other renewable sources are preferable to fossil fuel plants, which are often cooled by water.

But the best way, he suggests, is to rewrite how Bitcoins are mined. He points to Ethereum, which recently changed the way it produces cryptocurrency from an energy-intensive proof-of-work mining to a proof-of-stake method. “By making that change, they got rid of the need for computational power altogether,” he says. “In their case, they cut their power demand by 99.85% as a result of just changing the software. You can make this happen too in Bitcoin—if you really want to.”