“AI inference costs per agentic workflow will increase more than fivefold through 2028, despite improving per-token cost economics — the 'Inference Paradox'.”

Gartner notes routing a task to an agentic reasoning model already raises provider inference costs at least fivefold versus a basic chatbot interaction.