OpenAI CEO Sam Altman stated that user politeness (i.e. saying "please" and "thank you" to ChatGPT) is a factor in their millions of dollars in operational costs.

A user on X (formerly Twitter) asked how much money OpenAI might be losing out on when people have polite exchanges. "How much money OpenAI has lost in electricity costs from people talking to their models and politely saying 'please' and 'thank you'."

Altman replied, "around $10 million well spent. You never know".

While it may sound trivial, these positively polite words require diction models to process and produce full responses which thus add to the series of determinations which, when combined (in architectures of multiple layers), add to the computational burden of the AI, ultimately increase the cost OpenAI pays costing $10-$13 per query which is based on the electricity and other commercial costs associated with that request.

AI data centers that are providing the power supply for these model responses, like that of ChatGPT, already consume electricity alone, ~2% of the world's electricity generation, according to Goldman Sachs. Each query (i.e. response) from ChatGPT-4 consumes about 2.9 watt-hours of electricity, which is about 10 times more than a regular Google search. This means if OpenAI has billion queries daily, approximately 2.9 million kilowatt-hours are generated each day (emphasis on "kilowatt-hours!" Not less - "watts!" as those numbers can be very misleading).

The Electric Power Research Institute (EPRI) expects data centers will account for as much as 9.1 per cent of US consumption by 2030 due to demands from AI workloads. The International Energy Agency (IEA) also projects that data centers will constitute more than 20 per cent of the growth in electricity demand by 2030 in the developed world.

In over of this challenge, Altman has called for clean energy advances to sustainably power AI systems. He has invested in Helion Energy, a nuclear fusion company and Exowatt, a solar startup to help meet the rising energy needs. OpenAI is also looking to enhance its data center capacity to meet this demand.