Advertisement
X

Politeness to ChatGPT Costs OpenAI ‘Tens of Millions’ in Operational Expenses, Says Sam Altman

An X user took to social media to share his thoughts on polite phrases increasing energy costs for OpenAI’s chatbot. “I wonder how much money OpenAI has lost in electricity costs from people saying ‘please’ and ‘thank you’ to their models,” he stated. Altman responded with humour and insight: “Tens of millions of dollars well spent,” he remarked. “You never know.”

Getty Images
Politeness to ChatGPT Costs OpenAI ‘Tens of Millions’ in Operational Expenses, Says Sam Altman Getty Images

OpenAI CEO Sam Altman recently disclosed that users’ politeness toward ChatGPT is causing an unexpected rise in the company’s operational costs. A post on X highlighted the issue, with a user speculating about electricity costs from people saying ‘please’ and ‘thank you’ to their models.

Advertisement

An X user took to social media to share his thoughts on polite phrases increasing energy costs for OpenAI’s chatbot. “I wonder how much money OpenAI has lost in electricity costs from people saying ‘please’ and ‘thank you’ to their models,” he stated.

Altman responded with humour and insight: “Tens of millions of dollars well spent,” he remarked. “You never know.”

The Cost

When a corporation trains an AI model to recognise images or interpret language, it utilises large datasets and advanced hardware such as GPUs (graphics processing units), TPUs (tensor processing units) or other high‑performance CPUs. This process consumes a substantial amount of electricity, with training complex models like GPT requiring hundreds of megawatt‑hours.

AI hardware generates significant heat during training. Cooling systems such as air conditioning or liquid cooling can consume nearly as much electricity as the computers themselves.

Advertisement

Ghibli Melts GPUs

Sam Altman has urged customers to “chill out” as the overwhelming demand for ChatGPT’s new upgrade is straining his team’s capacity. “Can y’all please chill on generating images. This is insane, our team needs sleep,” Altman wrote.

In response to a post by Y Combinator founder Paul Graham, Altman explained, “We just haven’t been able to catch up since launch so people are still working to keep the service up. Biblical demand, I have never seen anything like it.”

Shortly after the image generator was released, Altman commented, “It’s awesome to see people enjoying images in ChatGPT, but our GPUs are overheating. We’re rolling out some temporary rate limits while we optimise things — shouldn’t take too long! Soon, the free tier of ChatGPT will get three generations daily.” While he didn’t detail the specific restrictions, Altman expressed hope that this measure is temporary as OpenAI works to better manage the surge in demand.

Advertisement
Show comments