Saying 'please', 'thank you' adds millions in OpenAI's costs: Sam Altman

While seemingly trivial, polite phrases require ChatGPT to interpret and generate full responses, thus increasing the computational load

OpenAI CEO Sam Altman has revealed that user politeness, such as saying "please" and "thank you" to ChatGPT, contributes to millions of dollars in operational expenses for the company.  A user on X wondered how much the OpenAI might be spending on polite exchanges. "How much money OpenAI has lost in electricity costs from people saying 'please' and 'thank you' to their models," the user asked. 

 

“Tens of millions of dollars well spent,” Altman replied, “You never know.”

 

While seemingly trivial, these polite phrases require ChatGPT to interpret and generate full responses, thus increasing the computational load. This, in turn, raises energy consumption and costs. 

 

AI's energy demand

 

The electricity used by AI data centers, which power models like ChatGPT, already accounts for about 2 per cent of global electricity consumption. According a Goldman Sachs report, each ChatGPT-4 query requires approximately 2.9 watt-hours of electricity, which is about ten times more than a standard Google search. With OpenAI handling over 1 billion queries daily, this translates to a daily energy consumption of approximately 2.9 million kilowatt-hours.

 

The Electric Power Research Institute (EPRI) projects that by 2030, data centers could consume up to 9.1 per cent of US electricity, driven by AI workloads. Similarly, the International Energy Agency (IEA) forecasts that data centers will account for more than 20 per cent of the growth in electricity demand in advanced economies by 2030.

 

In light of these challenges, Altman has expressed the need for clean energy breakthroughs to power AI systems sustainably. He has invested in nuclear fusion company Helion Energy and solar startup Exowatt to address the growing energy demand. OpenAI is also working on enhancing its data center capabilities to meet future demand.