Skip to main content


By 07.10.2023October 8th, 2023No Comments

07.10.2023 |

What the leading AI startup’s business shows about the future of technology


When it was released last November, no one would have imagined what a phenomenal success the flagship product of leading AI startup OpenAI would be. Chat GPT outperformed the competition, such as Google’s Bard, by leagues. But will it be able to hold the lead?

Interest in AI technology remains high. Japanese tech investment company SoftBank has committed to OpenAI, providing them with significant financial backing. The basis of this commitment is OpenAI’s collaboration with Microsoft, whose Azure cloud computing platform provides the computing power they need. To receive this support, Microsoft will receive a significant portion of OpenAI’s future profits. It is vital for OpenAI to receive sufficient funding to develop increasingly sophisticated models. Training the newest model, GPT-4, is estimated to cost $100 million. It is interesting to note that OpenAI does not plan to develop their next major model (GPT-5), instead they will create a more economically used model, GPT-4.5. The main reason for this decision is that the cost of the models is an important factor and OpenAI is looking to balance its costs and revenues.

On the other hand, OpenAI is not a Chat GPT-only development company. Their main revenue comes from developing bespoke AI solutions for large corporate clients, such as investment bank Morgan Stanley.

OpenAI may have valuable advantages, but they don’t guarantee its dominance. The network effects that have helped big companies like Alphabet, Amazon and Meta become quasi-monopolies in the areas of search, e-commerce and social networking have yet to materialise. Despite the large amount of users, GPT-4 is not a significant improvement over the previous version six months ago. When optimized using user data, the overall performance of GPT-4 has changed in unpredictable ways and in some cases for the worse.

The struggle to be first in creating models is also met with drawbacks. For model builders, the biggest expense is experimentation, which often does not yield results before a successful model is trained. OpenAI is estimated to have lost $500 million last year, even though training GPT-4 alone cost only a fifth of that amount. In the world of artificial intelligence, news of failed ideas spreads quickly, helping OpenAI’s competitors avoid costly mistakes.

Ultimately, the market is projected to contain a few large universal models and a host of specific ones. If AI turns out to be what it is believed to be, the role of oligopolist could bring OpenAI a lot of money. And if its backers only really see some of that money once the company has created a human-like thinking machine, then all bets will be off.



Target audience

Digital skills for the workforce

Digital technology

Artificial Intelligence