Everyone says AI is drinking the planet dry, right? Well, the numbers say your morning coffee uses 29,600× more water than a ChatGPT prompt!
In this episode, I sit down with Alex Passini to pressure-test the dominant narrative around AI's water footprint - and what we found completely flips the story the media is telling.
Here's what most "AI water crisis" headlines miss:
→ A single ChatGPT prompt uses roughly 16 milliliters of water. One cup of coffee uses ~140 liters when you count the beans. That's a 29,600× gap.
→ Of the water a hyperscale data center consumes, ~75% isn't used by AI at all — it's used upstream for the energy that powers it. Blame the grid, not the GPU.
→ One banana = ~6,250 ChatGPT prompts. One almond = ~12 prompts. The water-per-prompt math is rounding-error territory next to your lunch.
→ Florida already reuses 800 million gallons/day of treated wastewater. Data centers aren't the threat - they're an accelerant for the water reuse capex the sector has been waiting twenty years for.
If you're an investor trying to figure out whether AI's water story is a real thesis or a media artifact, this episode gives you the framework (and the numbers) to decide.
🧠 What you'll learn:
- Why "data center water consumption" and "AI water consumption" are not the same thing
- The actual per-prompt footprint of generative AI (modeled, not estimated)
- How water reuse technology turns the alarmist narrative into an investment opportunity
- Where the real bottleneck sits - and which sub-sectors stand to benefit
34:53
49:29
31:52
1:01:12
1:17:01