AI Inference Costs Will Increase More Than Fivefold Through 2028

The expenditure associated with AI inference per agentic workflow is projected to increase more than fivefold through 2028, according to Gartner. As AI products transition from assistive features to multi-step autonomous execution, product leaders are facing a significant margin challenge, with declining model prices effectively subsidizing more complex workflows and escalating overall AI expenditures. Consequently, the management of inference costs has emerged as a primary strategic priority.

“Product leaders cannot rely on more efficient token economics to rationalize AI costs,” stated Will Sommer, Senior Director Analyst at Gartner. “Each successive generation of AI capability will necessitate more, and often more expensive, tokens. There is no reliable, economical one-size-fits-all model on the horizon. Producing competitive AI products will require developing and maintaining complex multimodel ecosystems.”

Gartner has identified three fundamental trends influencing token economics:

1. Foundational model cost efficiencies are rapidly improving.
2. Enhanced AI efficiency is enabling the deployment of more sophisticated, higher-cost models to facilitate advanced AI applications.
3. Complex AI workflows consume significantly more tokens than basic chatbot interactions, leading to higher aggregate inference costs.

These dynamics indicate that while tokens are becoming more cost-efficient, they are not doing so at a rate that offsets the increasing costs associated with advanced AI capabilities. The velocity of innovation is currently outpacing the cost curve, resulting in what Gartner defines as the "Inference Paradox"—a scenario where improved unit economics actually escalate the total cost of AI without providing a clear path to predictable value.

“The harsh economics of the Inference Paradox are exemplified by the differences between a simple chatbot and an AI agent,” Sommer remarked. “Where a simple chatbot must read and interpret a query and quickly respond with a probabilistically reasonable answer, an AI agent must constantly reason, negotiate, and question itself.” These responsibilities are cumulative; compared to basic chatbot interactions, routing a task to an agentic reasoning model increases inference costs by at least five times, with costs rising further as task complexity increases.

Securing a return on investment from advanced AI necessitates significantly higher returns relative to basic models, or highly optimized tiering and orchestration to balance complex tasks against more cost-efficient intelligence. While both outcomes are achievable, they require substantial effort across intricate workflows. “Defaulting to generic autonomous intelligence will result in unbounded costs orders of magnitude higher than those of optimized product ecosystems,” Sommer concluded.