ChatGPT is on the rise, with more and more people using its text generation services. Its operation, however, costs an arm to OpenAI. The company must provide extremely expensive maintenance to keep the chatbot operational.
Since its launch last November, ChatGPT has attracted an impressive number of users. According to a recent count, the site has already received more than 4 billion visits. In March, the chatbot had 186 million accounts. To work, the conversational robot relies in particular on solid servers, which is obviously far from free. A new report gives an idea of daily operating costs.
Chief analyst at SemiAnalysis, Dylan Patel reveals to The Information that ChatGPT would require $700,000 a day to remain operational. In other words, each query would cost 36 cents. “Most of that cost is based on the expensive servers they need”, underlines the expert. This would also be a low estimate since it takes GPT-3 as a reference even though paying subscribers also benefit from the GPT-4 model.
ChatGPT needs $700,000 a day to operate
Although training the large language models of ChatGPT likely costs tens of millions of dollars, operational expenses or inference costs “far exceed training training costs when deploying a model on such a scale”Patel and Afzal Ahmad, another SemiAnalysis analyst, tell Forbes. “Ultimately, ChatGPT inference costs exceed training costs on a weekly basis”.
To read > ChatGPT Plus: the $20 per month subscription is launched
In order to reduce costs, Microsoft is developing a chip known as Athena. Under construction since 2019, this would help in the daily operation of ChatGPT. More than 300 Microsoft employees are now said to be actively working on its development. The chip could be ready for internal use by Microsoft and OpenAI as early as next year.
