Uploaded March 2026 | Updated September 2026, 2 weeks ago
How much water does AI use?
We asked people that in the latest episode of Water Street Questions from the Reservoir Center.
The reactions were… all over the place.
Some people immediately admitted they had no idea: “Oh God, I’m not good with AI.”
Others threw out a range of guesses — 200 gallons, three bottles of water, maybe a gallon per query. One cybersecurity expert said they probably should know the answer but didn’t. Another pointed out that it likely depends on the size of the data center doing the computing.
Most people knew the key point: AI actually uses water — data centers require cooling systems to keep servers from overheating, plus rely on energy sources that often use water.
So what’s the answer?
Research from Lawrence Berkeley National Laboratory estimates that a single AI query may use about 17 milliliters of water — roughly one tablespoon.
That might sound small. But when AI systems process millions or billions of queries, it adds up — making water an important part of the AI conversation.
What question should we ask next on Water Street Questions? Drop it in the comments.
waterloop is a nonprofit news outlet exploring solutions for water sustainability. Visit waterloop.org
Subscribe to catch every video: youtube.com/@waterloop
Never miss an episode! Subscribe to waterloop:
🎧 Spotify:bit.ly/waterloopSpotify
🎧 Apple Podcasts:bit.ly/waterloopApple
🎧 YouTube Podcasts:bit.ly/waterloopYouTubePod
Follow waterloop for more stories on water sustainability:
🔹 Instagram:instagram.com/the_waterloop
🔹 LinkedIn:linkedin.com/company/thewaterloop
#water #ai #artificialintelligence #datacenters #waterandtech #waterinnovation #watersecurity #watermanagement #climateandwater #sustainability #watersector #waterprofessionals #waterpolicy #washingtondc #waterloop
How much water does AI use?
We asked people that in the latest episode of Water Street Questions from the Reservoir Center.
The reactions were… all over the place.
Some people immediately admitted they had no idea: “Oh God, I’m not good with AI.”
Others threw out a range of guesses — 200 gallons, three bottles of water, maybe a gallon per query. One cybersecurity expert said they probably should know the answer but didn’t. Another pointed out that it likely depends on the size of the data center doing the computing.
Most people knew the key point: AI actually uses water — data centers require cooling systems to keep servers from overheating, plus rely on energy sources that often use water.
So what’s the answer?
Research from Lawrence Berkeley National Laboratory estimates that a single AI query may use about 17 milliliters of water — roughly one tablespoon.
That might sound small. But when AI systems process millions or billions of queries, it adds up — making water an important part of the AI conversation.
What question should we ask next on Water Street Questions? Drop it in the comments.
waterloop is a nonprofit news outlet exploring solutions for water sustainability. Visit waterloop.org
Subscribe to catch every video: youtube.com/@waterloop
Never miss an episode! Subscribe to waterloop:
🎧 Spotify:bit.ly/waterloopSpotify
🎧 Apple Podcasts:bit.ly/waterloopApple
🎧 YouTube Podcasts:bit.ly/waterloopYouTubePod
Follow waterloop for more stories on water sustainability:
🔹 Instagram:instagram.com/the_waterloop
🔹 LinkedIn:linkedin.com/company/thewaterloop
#water #ai #artificialintelligence #datacenters #waterandtech #waterinnovation #watersecurity #watermanagement #climateandwater #sustainability #watersector #waterprofessionals #waterpolicy #washingtondc #waterloop
