OpenAI threatened with losses of 5 billion US dollars
The business models of AI providers are on shaky ground. OpenAI is even threatened with a loss of several billion euros.
Take my money and shut up: OpenAI is still getting enough money from investors.
(Image: TierneyMJ/Shutterstock.com)
OpenAI is set to spend around seven billion US dollars on the development of new AI models. They pay their employees 1.5 billion US dollars. According to a report by The Information, however, only about 3.5 to 4.5 billion US dollars are apparently coming in. The magazine cites informants and unpublished financial reports that suggest OpenAI will run out of money in the next twelve months. The report quotes possible losses of up to five billion US dollars this year.
And in fact, what The Information is reporting is no great surprise. At the beginning of the year, it was already reported that OpenAI would generate around two billion US dollars in revenue in 2024- not enough to cover its costs. However, the sum was reportedly a mere extrapolation of the revenue from December 2023 for the entire year. Even then, OpenAI was already said to need significantly more money.
In addition to the development of new AI models and employees, the operation of current services also consumes a lot of money. ChatGPT alone is said to have 100 million active users per month. More people are likely to join soon when Apple integrates ChatGPT. However, Apple does not pay any money for this.
New AI models cost a lot of money
The development of new AI models also requires a lot of computing power - and therefore chips. OpenAI CEO Sam Altman is already looking for investors in order to almost completely reshape the semiconductor industry. He wants to significantly increase chip production capacity. And then there is electricity. OpenAI is said to be in talks with Helion, a company that wants to build fusion power plants. Altman himself is a major investor there. The energy start-up also wants to supply Microsoft from 2028. However, there is not yet even a building permit for a power plant, let alone certainty that the technology will work.
Shortly after the hype surrounding ChatGPT flared up, Altman himself said that the business models surrounding AI were difficult. He said that he had no idea how the services would be financed. Ultimately, this has only been possible so far with the help of investors. Microsoft has invested several billion US dollars in OpenAI.
Read also
The operation of AI models must therefore become significantly cheaper in order to be worthwhile. All known providers are working on this. OpenAI's GPT-4o, for example, is designed to run much more resource-efficiently than GPT-4. One way to increase cost efficiency is model distillation. This involves transferring the knowledge of a large AI model to a smaller one, which is then cheaper to operate. Meta sees the newly published open source model Llama 3.1 405B as such a possibility for developers. However, the development of large models remains expensive.
The AI applications themselves may have to become significantly more expensive for end users in order to finance further development and operation.
(emw)