Two of the world’s leading artificial intelligence companies, OpenAI and Anthropic, are preparing for possible stock market listings. Financial data shared before these expected IPOs shows a major challenge: building advanced AI systems is extremely expensive.
A large part of their spending goes into training AI models. This process teaches machines how to understand language and respond intelligently. However, each new version requires more computing power, more data, and much higher costs than before.
OpenAI expects its computing costs for AI research to reach about $121 billion by 2028. Even with strong revenue growth, it may still face losses of around $85 billion in that year. These projected losses are far larger than what most companies experience.
Anthropic is expected to spend less than OpenAI, but its costs are also rising quickly. Its projections show that training advanced AI systems is becoming more expensive over time. As AI improves, each step forward costs more than the last.
Investor charts show that yearly spending on AI training is increasing sharply for both companies. In some cases, these costs take up a very large share of their revenue, highlighting the scale of investment required.
Russia VPN restrictions linked to payment system disruptions, claims telegram founder
Profits, Losses, and Revenue Growth
Both OpenAI and Anthropic report two versions of their profits. One includes the cost of training AI models, while the other excludes it.
When training costs are excluded, both companies appear close to profitability. OpenAI is on track to post a small operating profit this year under this method. Anthropic also shows positive results in its best-case scenario.
However, when training costs are included, both companies report large losses. OpenAI does not expect to break even until the 2030s. Anthropic is expected to reach that point sooner, but still faces significant losses in the near term.
At the same time, both companies are experiencing rapid revenue growth. They expect their income to more than double within a year. This growth is mainly driven by businesses adopting AI tools for daily use.
Revenue comes from enterprise clients, consumer subscriptions, and new products. There are also differences in how revenue is counted. For example, Anthropic includes sales through cloud partners, while OpenAI does not, making direct comparisons more complex.
High Operating Costs and Cash Burn Pressure
In addition to training costs, both companies face high expenses to run their AI systems. These are called inference costs, which occur whenever users interact with AI tools.
Currently, inference costs take up more than half of the revenue for both OpenAI and Anthropic. This means a large portion of their earnings is spent on maintaining services.
A key issue is that many users do not pay for access. Free users still generate costs, especially with frequent usage. This creates pressure on finances, as companies spend money without earning directly from a large share of users.
To manage this, companies focus on converting users into paying customers and expanding enterprise services. Financial data also shows that both companies are expected to continue burning cash for several years, with negative cash flow due to heavy investment in AI development.
To support these rising costs, both companies are preparing to raise large amounts of money through stock market listings. Financial institutions such as Nasdaq are adjusting rules to help newly listed companies access investment funds more quickly.
These financial details show the scale of spending required to build and operate advanced AI systems, as both companies continue to invest heavily ahead of their expected IPOs.