Overnight, tech giants' earnings reports fell short of expectations, triggering investors' pessimistic sentiment towards AI and causing a broad decline in U.S. stocks. Now, the news that OpenAI may suffer significant losses this year could add even more pressure to an already strained AI industry.

According to a report by tech media The Information, based on undisclosed internal financial data and information provided by relevant business personnel, they believe that OpenAI may incur a loss of $5 billion this year. The media outlet stated that if their analysis is correct, OpenAI may need additional financing within the next 12 months.

The question arises, if AI companies are consistently unable to reduce the costs of developing and operating AI, do they have no choice but to charge higher prices, effectively passing the costs onto the consumers? However, the industry is confident that the cost of inference will continue to decrease significantly.

Where exactly does OpenAI spend its money?

Let's first look at the expenditures. The Information cited insiders who said that as of March this year, OpenAI is expected to spend nearly $4 billion this year on renting Microsoft's servers to support ChatGPT and its underlying LLM (i.e., inference costs).

Advertisement

In addition to running ChatGPT, this year's training costs, including data costs, may also rise to $3 billion.

Last year, OpenAI accelerated the training of new AI models, surpassing the initial plan, resulting in final costs far exceeding the company's expected $800 million. Therefore, this year, as OpenAI trains new versions of its flagship LLM and begins training new flagship models, the associated training costs may double.

Furthermore, OpenAI's workforce has reached around 1,500 people, and to compete with companies like Google for top technical talent, the overall salary costs are approximately $1.5 billion. Currently, OpenAI still has about 200 vacancies listed on its official website, which indicates that more employees may be recruited in the second half of the year.

Overall, OpenAI's operating costs for this year may reach as high as $8.5 billion.

Now let's look at the revenue. In addition to the approximately $2 billion annual revenue from ChatGPT, OpenAI also charges developers. As of March this year, this business generates more than $80 million in revenue per month.Building on this foundation, the company has recently achieved a monthly total revenue of approximately $283 million, which, when calculated, suggests an annual revenue that could range between $3.5 billion and $4.5 billion. However, the specific situation will still depend on the sales performance in the second half of the year.

In other words, after deducting the operating costs of around $8.5 billion from the revenue of $4.5 billion, the result may indicate a loss of between $4 billion and $5 billion.

Nevertheless, in the face of such a significant discrepancy between income and expenditure, the overall attitude within the industry seems to be relatively calm.

Ion Stoica, co-founder of enterprise software company Databricks, stated:

The industry is confident that the cost of inference will continue to decrease significantly, which means that OpenAI may be able to generate reasonable profits from its models in the coming years... The cost of training AI is more complex, and companies may continue to invest as much funding as possible to train larger models.