Gartner, the world's leading IT research company, predicts that by 2028, AI coding tools could cost more than the average developer's salary. The reason is that companies will buy more licenses, and AI vendors will start charging for the amount of work performed and for the actual result.
Gartner, the world's leading IT research company, predicts that by 2028, AI coding tools could cost more than the average developer's salary. The reason is that companies will buy more licenses, and AI vendors will start charging for the amount of work performed and for the actual result.
Since each code generation, refinement, or bug fix consumes tokens, companies have to overpay for the use of AI that improves productivity.
At the same time, analysts warn that many companies are significantly underestimating how rapidly token consumption will grow once AI coding becomes common practice.
“Engineering team leaders are increasingly concerned as token-based AI spending becomes increasingly difficult to justify, and budgets are often depleted earlier than expected,” explained Senior Principal Analyst Nitish Tyagi.
This comes as businesses move from pilot projects to large-scale implementation of the technology, and more developers are beginning to rely on AI in their daily workflows.
Although companies face higher costs, they ultimately get better results. Developers cite faster code delivery, convenience, and higher quality code generation as key benefits. Therefore, they are unlikely to voluntarily reduce token consumption now if it slows down their work.
“‘Token discipline’ will not emerge solely at the whim of developers themselves, as they tend to choose speed and convenience over cost savings,” Tyagi added, emphasizing the need for company-wide control and governance.
The report also points to some shortcomings in the industry, including the fact that vendors do not provide sufficient transparency about how tokens are calculated. As a result, it is difficult for companies to predict their costs and optimize usage accordingly.
