Gartner predicts AI use will become more expensive than human devs
By 2028, the costs of AI-driven programming will exceed the average salary of a developer, according to market research firm Gartner.
The increase is mainly caused by the explosive consumption of tokens in large language models and the shift to licensing models where you pay based on usage.
Organisations are switching ever faster from experimenting to large-scale deployment of AI coding agents. From the perspective of progress and process efficiency this is not objectionable, but the financial impact of token consumption is often underestimated, according to Gartner.
“Developers typically optimise for speed and convenience, not for cost,” says senior principal analyst Nitish Tyagi in a press release. Without a tightly organised operational model, expenses can rise faster than the productivity gains the tools promise.
This is exactly the situation that Dutch financial newspaper Het Financieele Dagblad described on Monday for the legal sector. The paper describes a situation in which law firms are caught off guard by the rising costs of AI use. The issue is not the price per token as such; that is on a downward trend. The major problem lies in the fact that employees are using AI more and more, firstly, and secondly that AI providers are increasingly being paid based on usage. In Gartner’s words, this leads to unpredictable cost structures and financial surprises. According to the consultancy, companies need to think more carefully about which models are deployed for which tasks.
System architects can also consider splitting tasks between Western models such as GPT, Claude and Gemini on the one hand, and Chinese open models Qwen, DeepSeek, MiniMax and GLM on the other. The Western ones are often more expensive but higher quality overall, while the Eastern ones are cheaper and sometimes also offer the option of self‑hosting. That in turn brings lower costs at scale.
Emerce
Subscribers 0
Fans 0
Followers 0
Followers