OpenAI, supported by Microsoft, is anticipated to remain unprofitable until 2029 due to the high costs associated with training its AI models. The company's financial documents indicate that by 2026, annual training expenses could reach $9.5 billion, and by 2030, cumulative spending might exceed $200 billion, with 60% to 80% allocated for model training and operation.
These significant expenditures are projected to result in a $44 billion loss from 2023 to 2028, with a $5 billion loss expected this year and possibly escalating to $14 billion by 2026. Notably, these estimates do not account for stock-based compensation, which is one of OpenAI's major expenses, despite it not being a cash outlay.
The company anticipates that annual revenue could reach $100 billion by 2029, potentially enabling profitability. OpenAI remains optimistic that ChatGPT will continue to be its primary revenue source, surpassing income from selling AI models to developers via APIs. Additionally, by the end of 2025, new products are projected to generate nearly $2 billion in sales, exceeding API revenue.
Although the specifics of these new products are not detailed, insiders suggest OpenAI is working on developing intelligent agents and research assistants for personal computers, among other technologies. The company has also explored offering subscription services at higher prices for its most advanced technologies. Other products still in development include the Sora video generator, aiming to compete with Google Search, and software specifically for robotics developers.
In contrast, growth in API sales is expected to slow, likely due to competition from Anthropic, Microsoft, and Google. Last week, OpenAI concluded a funding round that raised $6.6 billion, bringing the company’s valuation to $157 billion, with participation from Microsoft and Nvidia.
