Breaking

OpenAI CEO Sam Altman Says ChatGPT Water Use Is Tiny Next to a California Almond

September 3, 2026 at 03:14 EDT

Foundation Models
Infra & Chips

OpenAI CEO Sam Altman has argued that fears over AI data-center water use are overstated, saying a single California almond takes more water to grow than thousands of ChatGPT queries and that modern facilities use about as much water as an office building.

September 2026 · OpenAI

One almond vs thousands of ChatGPT queries

Sam Altman says modern AI data centers are not the water hogs of public memory — and that growing a single California almond uses more water than thousands of ChatGPT prompts.

0.32 mL

Altman: average ChatGPT query

0.26 mL

Google: Gemini text prompt

3.56 L

USGS-linked 2019 almond average

~11,000×

Almond vs 0.32 mL query

Water per item (same unit, drawn to scale)

One ChatGPT query at 0.32 mL versus one almond at 3.56 L — about 11,000 times more water for the nut. A circulating 38,000-query figure does not match those two published estimates.

0.32 mL

1 ChatGPT query

3.56 L

1 California almond

Takeaway: per prompt, ChatGPT water use is a drop; the almond is a jug. Scale of the industry is a separate question.

The case for small prompts

Altman: modern halls use about as much water as a large office. Per-query figures from OpenAI and Google sit well under a milliliter.

The case for local impact

Critics: a teaspoon per chat is the wrong unit when campuses draw millions of gallons a day, and the 38,000 figure does not match 0.32 mL × 3.56 L.

What still has to hold up

A “robust meme,” he said — and still the industry’s job to change minds.

Virginia’s 2024 review found most data centers at or below large-office water use, but some sites used far more. Gallup: 71% opposed a data center nearby. Altman’s prescription is product value, not a better analogy.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.