Gartner predicts agentic AI expenses could surge over five times by 2028

Gartner’s latest outlook suggests that the expense of deploying agentic AI systems could increase by more than five times by 2028, driven by the adoption of more elaborate AI-driven processes. While the price per token for foundational models is falling, the token demand of complex agent workflows grows faster, eroding cost savings. Senior director analyst Will Sommer contrasts the lightweight nature of chatbots with the heavy reasoning loops required of autonomous agents, which inflate inference costs.

Gartner also points to the shift by AI vendors toward usage-based billing, which can lead to runaway expenses for token-intensive tasks. As a result, the firm predicts that a sizable share of firms may either scale back or decommission their AI agents unless they achieve substantially higher returns or improve orchestration efficiency. Earlier forecasts from Gartner indicated that many generative AI projects could exceed budgets due to poor architecture and lack of expertise.