Building a house = 250 M AI prompts
Building a house equals roughly 250 M AI prompts on the CO₂ side: 50,000,000 g, while an average text prompt weighs just 0.2 g. The math, the source and detailed equivalents are right below, with public, verifiable numbers.
Figures updated on June 30, 2026 · Source : MIT Climate Portal : empreinte de la construction d'une maison
The CO₂ of a prompt reflects the emissions of the electricity used while answering, on an average power grid. On the action side, figures include production (food, goods) or combustion (transport). Here again, we compare orders of magnitude, not gram-perfect measurements.
Why 250 M prompts?
The math fits in three lines, using public, sourced values:
- 1.Building a house: 50,000,000 g (MIT Climate Portal : empreinte de la construction d'une maison)
- 2.1 text prompt: 0.2 g (published range: 0.15 to 1 g)
- 3.50,000,000 g ÷ 0.2 g ≈ 250 M prompts
In other words, “Building a house” weighs as much as 250 M questions asked to a generative AI. It is an order of magnitude, not a to-the-decimal measurement: sources publish ranges, and we use a median value, explained in our methodology.
What about a generated image or a reasoning prompt?
Not every AI use weighs the same. A generated image costs about 6 g and a “reasoning” prompt about 13.3 g, versus 0.2 g for a simple text prompt. So “Building a house” is also roughly 8,333,333 times one generated image, or 3,759,398 times one reasoning prompt.
Convert your own usageWhere do these numbers come from?
The value we use for “Building a house” comes from “MIT Climate Portal : empreinte de la construction d'une maison”, referenced like all our data on the Sources page. No number on this page is made up: every value can be traced back to a public document.
On the prompt side, there is no official meter: we convert the electricity used with the average grid intensity (about 480 g of CO₂ per kWh), which lands around 0.15 g in operation. We use 0.2 g per prompt, rising towards 1 g over the full life cycle.