Developing ChatGPT was so expensive that OpenAI made a loss last year. In accordance with The Informationthe company lost 540 million dollars in 2022. Spending soared in the months leading up to the AI chatbot’s launch and in a year that the company hired key personnel who previously worked for Google.
The figure was leaked by three people with knowledge of the startup’s finances. The report accounts for the huge investment that OpenAI had to make to train its machine learning models, before it began to sell access to ChatGPT.
The company’s revenues have increased, reaching an annual rate of hundreds of millions of dollars. Costs, however, will continue to rise as more customers use the AI technology developed by the company. The more users, the more technological infrastructure is needed to keep the platform active.
The popularity of ChatGPT made it the fastest growing consumer app in history. In less than two months, summoned more than 100 million monthly active users.
OpenAI CEO Sam Altman privately suggested that they would try raise up to $100 billion over the next few years to offset losses. This, with the aim of developing a general artificial intelligence advanced enough to improve its own capabilities, the sources said to The Information. They said it will be the most capital-intensive startup in Silicon Valley history.
The balance between gains and losses of OpenAI
OpenAI’s million-dollar investment that led to losses last year is just the beginning of the story. Keep ChatGPT operational it could cost the company at least $700,000 per day. Most of the money goes into maintaining its servers, the semiconductor research firm estimated. Semi-Analysis.
Training giant language models like GPT-4, which brings ChatGPT to life, can require an investment of tens of millions of dollars. However, in the end, the operating costs far outweigh this item.
The OpenAI chatbot uses what AI experts call “inference,” a skill that allows it to calculate responses based on input from each of its users. And according to another report published by Semi-Analysis, the costs to infer from ChatGPT exceed training costs on a weekly basis. This estimate is based on the GPT-3 model, so it is likely that the operating cost today will be even higher, after the launch of GPT-4.
The search for lower costs
OpenAI launched a paid version of ChatGPT last February, as one of many strategies to lower costs and control money losses. But its financial position has been strengthened, mainly, by the multimillion-dollar agreement it reached at the beginning of the year with Microsoft.
Within the framework of this alliance, Microsoft has started developing an AI chip it calls the Athena. This move will speed up the development of proposals such as OpenAI’s ChatGPT and GPT-4 and would considerably reduce implementation costs. In this project they have the support of AMD.
OpenAI revenue is projected to increase significantly this year. In accordance with ReutersIt is expected that reach $200 million. In 2024, the expectation is that rise to $1 billion.