Generative AI models such as OpenAI's ChatGPT, Google's Bard and Anthropic's Claude are largely dependent on the significant computing power required to perform the complex mathematical algorithms necessary to generate answers to user queries. Large technology companies such as Microsoft and Google face the problem of how to turn AI products into profitable businesses.

AI often lacks the economies of scale that standard software has because it requires intensive computation for each individual query. As product usage increases, so do the costs associated with maintaining the infrastructure. This exposes companies that charge fixed fees for AI services to potential financial losses.

In particular, the Wall Street Journal report said Microsoft's GitHub Copilot service, which helps app developers by automatically generating code, is operating at a loss despite amassing more than 1.5 million users and integrating into nearly half of their coding projects. Users pay a flat fee of $10 a month for the service, but Microsoft's average cost per user is more than $20 a month, according to an insider source. In some cases, individual heavy users cost the company as much as US$80 a month.

To reduce costs, companies are exploring the use of less powerful and more cost-effective AI tools that are sufficient for specific tasks. For example, they choose ChatGPT 3.5 AI instead of the latest GPT-4 if it meets the requirements. Another way to reduce the cost of AI services relates to hardware, with some AI firms exploring the development of more affordable and efficient processors with lower power consumption to handle the workload.

Given these factors, it is not surprising that investors are increasingly cautious about investing in AI. According to a forecast by analyst firm CCS Insight, generative AI will face financial hurdles in 2024 as the costs associated with the technology rise.