OpenAI CEO Sam Altman stated that the average ChatGPT prompt consumes about 0.000085 gallons of water—roughly equivalent to one-fifteenth of a teaspoon—while discussing the energy and water costs of artificial intelligence.
Article contents:
- Introduction
- Altman’s statements
- Energy consumption
- Industry scrutiny
- Comparative studies
- Conclusion
Introduction
Sam Altman, CEO of OpenAI, said in a blog post published on Tuesday that the average ChatGPT prompt consumes about 0.000085 gallons of water, or “about 1/15th of a teaspoon.” Altman made the claim in a broader post discussing his predictions for how the world will change thanks to artificial intelligence.
Altman’s statements
“People often wonder about the amount of energy a ChatGPT prompt consumes; the average prompt uses about 0.34 watt-hours, which is roughly what an oven uses for one second, or what an energy-efficient lightbulb uses for a few minutes,” Altman wrote. He added that “the cost of intelligence should eventually approach the cost of electricity.” OpenAI did not immediately respond to a request for comment on how Altman arrived at these figures.
Energy consumption
AI companies have faced scrutiny over the high energy costs consumed by their technologies. For example, researchers predicted this year that AI energy consumption will reach a level surpassing bitcoin mining by the end of the year.
Industry scrutiny
The artificial intelligence industry is facing mounting questions regarding the environmental footprint of its technologies, especially with the accelerating growth in demand for generative AI services. The data centers powering these models require significant amounts of electricity and water for cooling.
Comparative studies
In a report last year, The Washington Post partnered with researchers to estimate that a 100-word email “generated by a GPT-4-powered chatbot” required “just over a single bottle of water.” The report also found that water consumption varies depending on the location of the data center running these intelligent models.
Conclusion
Although the figures provided by Altman indicate a relatively low consumption of water and energy per query, the sheer scale of global ChatGPT usage means the overall impact could be substantial. There remains an urgent need for greater transparency and independent studies to understand the true environmental impact of artificial intelligence technologies.
