After the period of promoting the application of artificial intelligence (AI), many technology businesses are facing a new problem when the cost of using AI increases faster than expected.
From startups to large corporations, more and more units are concerned about the huge token bills (the smallest basic data unit used by the system to analyze and process information) arising from the use of advanced AI models.
According to TechCrunch, cost pressure has become a hot topic in the technology world. Some businesses said they have used up their AI budget for the whole year in just the first few months of the year.
Some companies are even said to have to spend up to 500 million USD on AI services after not setting usage limits for employees.
The reason lies in the explosion of new generation AI models and agents. Although the price of each token has decreased compared to before, the amount of tokens consumed has increased sharply as businesses deploy chatbots, AI assistants and large-scale automation tools.
New models such as Claude, GPT or Gemini have better work processing capabilities, but at the same time also cause demand for AI to skyrocket.
Alexander Embiricos - Senior Product Director at OpenAI, said that customer interactions have now changed significantly. If businesses were previously interested in what AI could do, now the common question is how to monitor costs, audit usage levels and control token consumption.
Faced with this situation, the Linux Foundation has announced plans to establish Tokenomics Foundation, a new organization to build common standards for token AI cost management. The goal is to create a system similar to FinOps, which is a method to help businesses control cloud computing costs for many years.
JR Storment - CEO of FinOps Foundation, said that since the beginning of this year he has continuously received feedback from businesses about exceeding the AI budget many times compared to the plan. According to him, companies are shifting from the mindset of "using as much as possible" to "how to effectively control spending".
However, the problem is not simple. A two-year study of 20,000 programmers conducted by Faros AI showed that AI actually helps improve work productivity.
However, the number of errors and times having to rewrite the source code also increased accordingly. Meanwhile, data from Jellyfish shows that engineers who use AI the most have twice the productivity of the remaining group but have to spend 10 times more tokens.
What makes businesses headache now is that there is no clear way to measure whether that expenditure really brings corresponding value or not.
Meanwhile, according to forecasts by investment bank Goldman Sachs, the global use of AI tokens may increase 24 times by 2030.
