The average PC user will likely think of “ChatGPT” whenever generative AI is mentioned, as it is perhaps the most well-known and widely used LLM system in the world. Despite its popularity, the huge costs of training AI models and running the servers that host them are so high that the company is on track to post an operating loss of $5 billion, according to a research report on the financial health of its creator, OpenAI.
This comes from The Information (via Windows' Central), which based its projections on figures taken from unreleased internal financial statements and various industry data. Founded in 2015, OpenAI has grown in size and scope thanks to multiple investments, with Microsoft pumping billions of dollars into the AI company over the years.
The latest capital injection of $10 billion in early 2023 is rumored to include 75% of OpenAI's profits and 49% of the company's shares, as well as integrating ChatGPT into Bing and other Microsoft systems. In return, OpenAI is believed to get the investment, as well as access to Azure cloud servers and heavily discounted fees.
But a report from The Information claims OpenAI isn't making nearly enough money and could post an operating loss of $5 billion by the end of its fiscal year.
OpenAI is said to have spent around $7 billion on training and inference of its LLMs, and $1.5 billion on staff, though other analysts have estimated the costs to be around $700,000. Per day Running ChatGPT is quite expensive due to the cost of Nvidia's AI servers, but it is not realistic to determine the exact cost. All this combined, OpenAI should still be profitable. So why is it losing so much?
The problem seems multi-faceted. OpenAI is heavily invested in being the first to achieve AGI (artificial general intelligence), which will surely eat into some of its profits. And companies developing generative AI systems face competition not only from OpenAI, but also from Anthropic, Amazon, Google, Nvidia, Meta, xAI, and many others. ChatGPT is the best known, but it is getting an increasingly smaller share of the total available revenue.
The Information's claim that OpenAI may go bankrupt isn't the first time it has done so: It issued dire warnings about the AI company's future last year, but other analysts have noted that the relative costs of AI training, inference and chip manufacturing will fall over time as more companies enter the market.
Given Microsoft's total investment to date, it's hard to imagine OpenAI pulling in a manager anytime soon: Earlier this year, it was valued at $80 billion, which isn't a huge amount of money, but if things really start to go under, Microsoft could afford to buy it outright.
That said, such an acquisition would undoubtedly be investigated by antitrust commissions around the world. OpenAI may increase ChatGPT's query costs to boost revenue, but this would likely lead to a decline in daily user numbers.
Perhaps OpenAI's main hope for the future is that cheaper AI servers will become available, helping to reduce operational costs, as sales of the much-touted AI PCs appear to be driven by interest in longer battery life rather than AI capabilities.
