The Water We Forgot to Measure.
AI data centers have sparked conversations about water use. What happens if that electricity comes from nuclear power? This isn't a debate about whether nuclear energy is good or bad. It's a question about measurement.
One widely cited AI study estimates the electricity-generation portion of AI's water footprint at about 480 gallons per megawatt-hour using the average U.S. electricity mix.
Typical nuclear power plants consume roughly. 400–720 gallons per megawatt-hour to generate electricity.
If AI data centers were powered entirely by nuclear, the electricity-related portion of their water footprint could be similar or higher than the average value used in the AI study.
I'm not arguing for or against nuclear power.I'm asking a different question:
Why do we talk about the water footprint of electricity when the conversation is about AI... and rarely when it's about electric vehicles, heat pumps, or nuclear power?
Sometimes the most interesting part of a story isn't the number - it's why we started measuring one thing and not another.
What other technologies do we evaluate using one metric—while ignoring the same metric everywhere else?
Watch the video here.
Reader Responses
After publishing this short on platforms such as LinkedIn, YouTube, and Instagram, people shared questions, critiques, and alternative perspectives. I've selected a few excerpts and included my responses below. Names have been removed to keep the focus on the ideas rather than the individuals.
Context Matters More Than Comparisons - Location Changes the Equation
Reader: Ok someone please clarify this: nuclear plants are usually built alongside existing water sources like lakes are they not? While AI data centers seem to be in places like Utah and Texas that don’t have much water.
My reply: That's an interesting question. It raises another one: Is that always true, and can we assume it will always be true? It reminds me of the Palo Verde Nuclear Generating Station near Phoenix, which is notable because it isn't located on a lake or river. Rather, it uses treated wastewater from surrounding communities for cooling. To me, that's an example of why context matters. Before drawing conclusions about any technology's water footprint, I'd want to understand where it's located, how it's cooled, and what water source it's actually using.
What Do the Numbers Actually Measure?
Reader: Thermal power plants of any kind do not consume water; they heat it and/or evaporate it. It is returned to the hydrologic cycle. Further, thermal power plants do not use potable water, they use the surface water available - seawater, brackish or fresh, but NOT potable water.
My reply: I appreciate this comment as it touches on the question behind my post. Distinctions such as potable vs. non-potable water, withdrawals vs. consumption, and where the water comes from matter. It makes me wonder how often those distinctions are preserved when numbers are reported publicly. When a headline says, "AI uses X gallons of water," what exactly does that number represent? That's the question I was hoping people would begin asking.
Who Benefits From the Narrative?
Reader: THANK YOU!! This is one of the most intellectually honest posts I've seen in this space in a long time. You're not arguing the numbers — you're exposing the frame.
We measure what we've decided to scrutinize. Data centers became the villain in the water conversation not necessarily because they're the worst offender but because they're a visible, concentrated, politically convenient target. Nuclear gets a pass. EVs get a pass. Heat pumps get a pass. Same metric, different story.
I'd take your question one step further — who benefits from the selective measurement? Because in my experience, when an industry gets singled out for scrutiny while others don't, there's usually a narrative being protected somewhere upstream.
The water conversation around data centers is real and worth having. But it should be honest. And right now it isn't.
My reply: That's the question I found myself asking, and it led me to write a longer article exploring it. One possibility I considered was the space industry, as several companies are now proposing orbital data centers as a solution to concerns about electricity, land use, and water consumption. They would benefit. The next question is whether they played any role in creating the controversy in the first place.
Based on the research I did, I couldn't find evidence of that. My impression was that they recognized an existing concern and positioned their technology as a potential solution.
What I did find interesting was how the original water-use research was reported. Once journalists picked it up, much of the discussion focused on the headline numbers rather than the assumptions, definitions, and context behind them. To me, that's where the interesting questions begin.
I ended up writing an article about that line of thinking—Data Centers, AI in Space, and How Narratives Shape the Future—if you're interested in the reasoning behind it.
Not All Water Use Is the Same
Reader: Thermal power plants of any kind do not consume water; they heat it and/or evaporate it. It is returned to the hydrologic cycle. Further, thermal power plants do not use potable water, they use the surface water available - seawater, brackish or fresh, but NOT potable water.
My Reply: I appreciate this comment as it touches on the question behind my post. Distinctions such as potable vs. non-potable water, withdrawals vs. consumption, and where the water comes from matter. It makes me wonder how often those distinctions are preserved when numbers are reported publicly. When a headline says, "AI uses X gallons of water," what exactly does that number represent? That's the question I was hoping people would begin asking.
Your Thoughts?
I'm interested in hearing your perspective. Share your thoughts, questions, or alternative viewpoints in the comments below.