Sam Altman, CEO of OpenAI, has indicated that simply being courteous towards ChatGPT could be leading to expenses amounting to “tens of millions of dollars” in additional computing resources.
When asked how much OpenAI has incurred in electricity expenses due to users saying “please” and “thank you” to their AI models, Altman noted, “tens of millions of dollars well spent. You never know.”
tens of millions of dollars well spent–you never know
— Sam Altman (@sama) April 16, 2025
Every interaction with ChatGPT—no matter how minor, including polite expressions—demands extra processing power. Even small exchanges can accumulate rapidly across billions of conversations, resulting in higher electricity costs and server demand.
AI systems heavily rely on large data centers, which account for around 2% of global electricity consumption, a figure that is expected to rise. The energy consumption of AI could potentially match that of an industrialized nation, such as Japan.
A Washington Post investigation, in collaboration with researchers from the University of California, found that producing a single 100-word email through AI consumes 0.14 kilowatt-hours of electricity—sufficient to power 14 LED lights for an hour. Collectively, these seemingly minor interactions generate a substantial energy footprint.
Water consumption is equally notable. Researchers at UC Riverside discovered that utilizing GPT-4 for generating 100 words can necessitate up to three bottles of water for cooling the servers. Even a brief three-word response, like “You are welcome,” can use about 1.5 ounces of water.
Why show politeness to machines?
While many joke about the need to be polite to AI to avoid potential domination, the evidence shows that users genuinely strive to adhere to proper AI etiquette. A survey conducted in late 2023 revealed that 67% of respondents in the US felt compelled to be courteous to their chatbots.
Out of those who engage in digital politeness, 55% indicated they do so “because it’s the right thing to do,” while a more cautious 12% confessed they engage in this behavior to “appease the algorithm in the event of an AI uprising.” Given AI’s rapid advancement since 2023, it’s likely that this percentage would be even higher today.
Beyond being a mere psychological or behavioral trend, a study by researchers at Waseda University revealed that polite prompts can lead to higher-quality responses from large language models (LLMs).
After all, LLMs are trained on human interactions; as the researchers put it, “LLMs reflect the human desire to be respected to some degree.”
The human aspect
Some experts argue that maintaining politeness towards AI serves a greater purpose beyond technical efficacy. Using disrespectful language with AI could potentially normalize rudeness in our interactions with one another.
This trend has already been noted with previous voice assistants. Some parents have reported that their children exhibit less respect after frequently issuing commands to Siri or Alexa, prompting Google to launch their “Pretty Please” feature in 2018 to encourage politeness among children.
So, whether you’re courteous to your AI for performance enhancements or as a matter of etiquette, it’s important to recognize that every interaction incurs a cost.
