AI has yet to write its business model

© iStock.

Large tech companies such as Microsoft and Google are investing heavily in AI-based tools, but the revenue model still escapes them. Generative AI models require lots of power, which is an expensive proposition: computing power and powerful servers with high-end, energy-consuming chips. The Wall Street Journal cites services such as Microsoft’s GitHub Copilot coding assist as an example, which is incurring large financial losses. Though GitHub has attracted more than 1.5 million users who each pay a 10 USD monthly subscription, the actual average cost per user exceeds 20 USD monthly. This service alone burns a 15 million USD hole through Microsoft’s pockets each month.

Clearly, there’s no return on investment on the horizon yet for most of these services. Their profitability depends on finding a way to lower operating costs, by leveraging advances in hardware and opting for better LLMs, as some are less resource-intensive than others. Microsoft, for example, uses GPT-4, OpenAI’s most complex model, for many of its AI functionalities. The WSJ quips that using this model for simple tasks such as summarizing an e-mail is akin to “getting a Lamborghini to deliver a pizza.” The most powerful AI models in the service of basic business functions might be overkill.

⇨ Ars Technica, Benj Edwards, “So far, AI hasn’t been profitable for Big Tech.”

2023-10-10