Saying ‘please’ and ‘thank you’ to ChatGPT costs tens of millions of dollars, OpenAI CEO Sam Altman admits

OpenAI CEO Sam Altman admitted that simple courtesies such as saying “please” and “thank you” to his company’s ChatGPT bot are surprisingly expensive — costing the firm substantial sums in electricity expenses.

Altman made the revelation when a user on the social media platform X asked about the financial impact politeness towards AI might have on OpenAI’s operating costs.

Altman responded by noting the loss was “tens of millions of dollars well spent,” adding cryptically: “You never know.”

Chatbots such as ChatGPT run on large language models (LLMs), which rely on extensive computational infrastructure hosted in data centers.

These models require thousands of high-performance GPUs (graphics processing units) to operate efficiently.

The GPUs perform vast amounts of parallel processing to interpret prompts and generate responses in real time.

Powering these data centers demands enormous amounts of electricity.

It’s estimated that generating a single AI-written response, such as a short email or paragraph, can consume as much as 0.14 kilowatt-hours (kWh) of energy — comparable to keeping 14 LED bulbs lit for one hour.

When scaled across billions of interactions daily, the cumulative energy usage becomes significant.

Globally, data centers already account for about 2% of total electricity consumption.

With the rapid expansion of AI applications and increasing demand for generative AI services like ChatGPT, experts warn that this figure could rise sharply in the coming years.

While some might view courteous interactions with chatbots as unnecessary, several AI experts argue that politeness significantly shapes AI interactions positively.

Kurtis Beavers, a director on the design team for Microsoft Copilot, has advocated for respectful prompts, stating they “help generate respectful, collaborative outputs.”

According to Beavers, polite phrasing doesn’t merely reflect good manners but actively influences how the AI replies, setting a more constructive and professional interaction tone.

“When it clocks politeness, it’s more likely to be polite back,” according to Microsoft WorkLab, a digital publication produced by the software giant that is specifically devoted to integration of AI in the workplace.

Politeness towards AI has become increasingly common.

A 2024 survey revealed that approximately 67% of American users regularly employ courteous language when interacting with chatbots.

Within that group, a majority (55%) believe politeness is ethically correct, while another 12% humorously indicated that their polite language serves as insurance against potential AI rebellion.