AI Water Use Estimates Reveal the Growth Play: When the Media Reports a Single Scary Number, Publish the Full Range of Estimates Instead.

The Play

California WaterBlog did not just correct AI water use estimates.

They published the full range.

That is the growth lesson.

The article starts by noting:

"Estimates of AI data center water use for California range from 2,300 acre‑ft/year to 400,000 acre‑ft/year."

Then it explains why the range is wide:

"California has about 15 million square feet of floor space for data centers… The energy dissipation needed for data center racks is about 2‑12 kw/square meter… Major industrial cooling systems seem to have efficiencies of 60‑90%…"

Finally, it places the estimate in context:

"AI use is about 0.055 percent of annual human water use in California."

"A recent study for Central Arizona found that beer production consumed more water than data centers in that region."

That combination—range, methodology, and benchmark—turns a scare story into a lesson in responsible estimation.

Why this matters

Environmental and technical reporting often suffers from endpoint cherry‑picking.

A reporter sees estimates spanning 2,300 to 400,000 acre‑ft/year and writes: "AI data centers could use 400,000 acre‑feet of water annually."

Readers think that is the number.

They do not see the 2,300 lower bound.

They do not know the range is 174× wide.

They do not understand why the range exists.

Showing the full range—and explaining its width—fixes that.

When the article says "estimates range from 2,300 to 400,000 acre‑ft/year," readers immediately understand that the number is uncertain. When it explains the methodological reasons, they see why.

What California WaterBlog got right

The article does several things especially well.

1. It leads with the range

The opening paragraphs present the estimate as a span, not a point.

That sets the tone: this is a story about estimation, not a story about a scary number.

2. It explains the methodology clearly

The article breaks down the calculation:

• Square footage of data centers
• Energy dissipation per square meter
• Cooling efficiency ranges
• Conversion to water evaporation

That shows readers where the uncertainty comes from.

3. It benchmarks against familiar things

The article compares AI water use to:

• Total human water use in California (0.055%)
• Beer production in Central Arizona (more than data centers)

Those comparisons ground the estimate in reality.

4. It uses AI to check its own work

The article notes:

"I did these calculations and then, perhaps appropriately, checked and explored these estimates using four AI models."

That meta‑move reinforces the theme: estimation is iterative, not absolute.

The growth play to steal

If you are reporting on a contested estimate, do not pick one number.

Publish the range.

The pattern looks like this:

1. State the range plainly ("Estimates vary from X to Y")

2. Explain why the range is wide ("Because of A, B, and C methodological uncertainties")

3. Compare to familiar benchmarks ("That is about Z% of total use, or equivalent to Q")

4. Show where media picked endpoints ("Outlet A reported X, Outlet B reported Y")

5. Provide tools for exploration (Links to data sources, calculators, etc.)

That sequence turns readers from passive consumers into critical thinkers.

Why founders miss this

Because ranges are messy.

It is easier to write "AI uses 400,000 acre‑feet of water" than "AI water use estimates span 2,300 to 400,000 acre‑feet annually, depending on cooling efficiency, data center square footage, and operational assumptions."

But that messiness is exactly what builds trust.

When you acknowledge uncertainty, you show intellectual honesty.

When you explain its sources, you demonstrate expertise.

When you provide benchmarks, you help readers judge scale.

The wording lesson

Notice how California WaterBlog frames the range:

"Estimates… range from…"

"The still broad… estimate seems reasonable."

"A narrower estimate supported by all four estimations would be…"

That is measured, not definitive.

It positions the article as a guide to estimation, not a declaration of truth.

Bottom line

The strongest trust play in technical reporting is transparency about uncertainty.

When you report on a contested estimate, do not cherry‑pick an endpoint.

Show the full range, explain why it exists, and help readers understand what the numbers really mean.

That turns sensationalism into education, builds authority faster, and makes your reporting much harder for critics to dismiss.

Sources:

https://californiawaterblog.com/2026/04/26/ai-water-use-distractions-and-lessons-for-california/

https://issuu.com/asuwattscollege/docs/kyl_center_-_industrial_water_use_placeholder

Hacker News discussion: 286 points, 261 comments (ID: 47977383)