According to estimates based on OpenRouter data, the average token price weighted by model usage has dropped from approximately $3 to $1 per million tokens over the past year, a decline of about 66%. The JPMorgan equity macro research team estimates that the annualized revenue run rate related to AI could exceed $300 billion by the end of 2026, more than triple the figure by the end of 2025. The price decline reflects changes in user preferences. This summer, some companies reached their annual token budgets ahead of schedule and began allocating models based on task difficulty, increasing the use of cheaper models. JPMorgan believes that improvements in hardware and software efficiency are also reducing token generation costs, while applications in intelligent agents and cybersecurity may continue to drive demand for calls. However, whether revenue can sustain growth depends on whether call volumes can offset price reductions. JPMorgan's 2027 scenario suggests: if the average price drops another 70% and token consumption increases tenfold, the revenue run rate could still grow by approximately threefold.