Here’s how much power and water ChatGPT might be using per month

OpenAI CEO Sam Altman just published a new blog post titled “The Gentle Singularity,” which follows the established format of the positive AI manifesto that we have read before.

“We are past the event horizon; the takeoff has started.”

Altman’s post is full of the usual accelerationist predictions for the 2030s: colonizing space, brain-computer interfaces, and abundant energy.

It also includes some of the fruits that he thinks superintelligent AI will bring us: robots building robots, data centers building data centers, and “fake jobs” that humans will have in the future that feel “incredibly important and satisfying.”

But one of the most noteworthy things in the post is a pair of numbers subject to much external speculation — the amount of energy and water that a typical ChatGPT query uses.

Now we have some hard numbers, right from the CEO’s blog, which we can examine. Altman wrote:

“As datacenter production gets automated, the cost of intelligence should eventually converge to near the cost of electricity. (People are often curious about how much energy a ChatGPT query uses; the average query uses about 0.34 watt-hours, about what an oven would use in a little over one second, or a high-efficiency lightbulb would use in a couple of minutes. It also uses about 0.000085 gallons of water; roughly one fifteenth of a teaspoon.)”

That doesn’t sound like much! But let’s add some context to these figures.

The Information recently reported that OpenAI is serving 500 million active users per week. We’re going to engage in some extrapolation here, but let’s assume each user submits 10 queries each per week.

So based on these (admittedly rough) numbers, one month’s worth of ChatGPT queries might use:

• 🏠 Enough electricity to power about 8,200 average American homes (~7.4 gigawatt hours)

• 💧Enough water to fill about 90 residential swimming pools (~1.8 million gallons)

Of course, these estimates could be different than ChatGPT’s actual usage — feel free to extrapolate your own guess at the number of queries done in a given time frame — but hey, at least we now have hard numbers on how much power and water are consumed for the average query.

Things are indeed moving fast, and who knows how many of Altman’s predictions might come true. But he’s aware how crazy some of his ideas sound:

“Intelligence too cheap to meter is well within grasp. This may sound crazy to say, but if we told you back in 2020 we were going to be where we are today, it probably sounded more crazy than our current predictions about 2030.”