Up to 1000 percent more: Gartner consultants warn of expensive AI bills
According to management consultancy Gartner, companies experimenting with AI run the risk of being surprised by bills that are many times higher.
(Image: Andrey_Popov/Shutterstock.com)
Companies integrating artificial intelligence (AI) into their processes must learn to manage costs and capture the productivity gains the technology enables, senior analysts from the consultancy said in the opening keynote at the start of this year's Gartner IT Symposium on Monday in Gold Coast, Australia.
Wasting money on generative AI is easy, said VP analyst Mary Mesaglio in the keynote, which she delivered with colleague Kristian Steenstrup. Just as IT departments sometimes miscalculate the costs of cloud usage after switching to a cloud provider, this can also happen in the case of artificial intelligence. She believes that miscalculations of 500 to 1000 percent are possible.
High costs possible
One reason for this is price increases by providers, another is that AI experimenters do not take into account the costs incurred for cloud usage. The fact that employees use AI for tasks that could also be accomplished with a simple web search or start experiments with long and complex queries could also drive up costs. This is more expensive for providers who charge by token.
The keynote speakers also mentioned the structure of the data accessed by the AI used by a company as another factor that affects costs. If artificial intelligence works with unstructured data, this may lead to better results, but it is expensive.
Whether the productivity advantage of an average of 43 minutes of saved working time per day per employee that generative AI is supposed to bring, according to a survey by the management consultancy, is worth the unpredictable financial investment seems unclear at the moment anyway. Another Gartner analyst predicted as recently as July that 30 percent of all generative AI projects could be scrapped by the end of 2025 due to unclear return on investment prospects.
According to the British online medium The Register, keynote speaker Mesaglio freely admitted during the presentation that instead of working harder, she would probably use the time she gained to get a coffee.
Meanwhile, the operators of the models also seem to be struggling to make a profit with generative AI. ChatGPT provider OpenAI is reportedly facing a loss of several billion euros. Nvidia, the main supplier of AI accelerators, on the other hand, is reporting record profits.
(kst)
