Artificial intelligence has been the hottest topic in tech since late 2022, when ChatGPT went viral. The AI race started almost immediately, with every big tech company in the US and elsewhere working on new AI systems of their own.
We quickly learned that software like ChatGPT requires massive resources. Datacenters packing thousands of expensive GPUs specialized in training and running AI chatbots were needed. The larger the data centers, the more energy the world would need to set aside for AI projects.
Some people worried about the impact AI infrastructure would have on the world. It wasn’t just about the electricity powering the chats, but also the water used to cool some of these data centers.
Two and a half years after ChatGPT went viral, we finally know how much energy and power a ChatGPT chat consumes. It comes from Sam Altman’s latest blog, titled The Gentle Singularity, which teases what the world could look like in the next five to ten years thanks to superintelligence:
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.
It’s unclear where Altman’s figure comes from. If accurate, I’ll probably consume about a teaspoon of water with AI queries every day. But the figure is also misleading, considering that OpenAI has hundreds of millions of monthly users.
Add to that all the energy and water Gemini, Claude, Meta AI, Deep Research, and all the other chatbots out there consume, and you’ll rack up quite a bill.
It’s unclear what prompted the CEO to pen the post, one he describes on X as maybe “the last one like this I write with no AI help at all.” It likely wasn’t to reveal the energy costs associated with ChatGPT chats, though energy is one of the big topics in the blog. On that note, I’ll point out that ChatGPT o3 just got dramatically cheaper than before.
Is superintelligence inevitable?
Unsurprisingly, Altman is quite optimistic about the future of AI. He presents superintelligence as inevitable. It’s a foregone conclusion. We’ll get to a world where smarter-than-human AI will make our jobs easier than ever, leading to potential massive discoveries to improve daily lives:
AI will contribute to the world in many ways, but the gains to quality of life from AI driving faster scientific progress and increased productivity will be enormous; the future can be vastly better than the present. Scientific progress is the biggest driver of overall progress; it’s hugely exciting to think about how much more we could have.
Altman expects AI to bring novel insights in 2026. A year after that, the world will start getting robots that can do tasks in the wild. Then, “the 2030s are likely going to be wildly different from any time that has come before,” Altman writes. That’s even though we, humans, will continue to enjoy our lives as we did before.
The OpenAI CEO also says that intelligence and energy will be “wildly abundant” in the 2030s. Once that happens, the world will be able to do things that weren’t possible before.
Speaking of energy, Altman also sees datacenter production becoming automated. AI and robots will power everything:
There are other self-reinforcing loops at play. The economic value creation has started a flywheel of compounding infrastructure buildout to run these increasingly-powerful AI systems. And robots that can build other robots (and in some sense, datacenters that can build other datacenters) aren’t that far off.
AI will help humanity achieve new “wonders” by 2035. Altman even sees a future where some people will choose to “plug in” via “true high-bandwidth brain-computer interfaces.”
Not so fast…
It all sounds amazing, and it certainly beats the gloomier pictures others paint about the future of AI.
Altman’s essay also downplays the downsides, like the massive job revolution we’re about to witness. The CEO isn’t ready to propose any solution for AI stealing jobs, other than indicating that humans will adapt and some sort of new social contract might emerge:
There will be very hard parts like whole classes of jobs going away, but on the other hand, the world will be getting so much richer so quickly that we’ll be able to seriously entertain new policy ideas we never could before. We probably won’t adopt a new social contract all at once, but when we look back in a few decades, the gradual changes will have amounted to something big.
Toward the end of the post, Altman also addresses the obvious challenges. AI has to be aligned to our interests to give us the rosy future he paints in the previous paragraphs. The other challenge is making sure superintelligence is “cheap, widely available, and not too concentrated with any person, company, or country.”
How will OpenAI and every other firm engaged in developing frontier AI ensure it’s safe, cheap, and widely available? Altman doesn’t say. We’ll just have to wait and see what happens next, while we continue to chat with chatbots like ChatGPT, a fifteenth of a teaspoon of water at a time.
The post How much energy and water does ChatGPT consume? appeared first on BGR.
