Enterprise Token Costs Drop 67% YoY: AICC Report Shows DeepSeek V4, Multi-Model Routing, and Prompt Caching Driving Deflation — But Agentic Consumption 5-30x Offsets Savings

AICC·high signal

The AICC report through April 30, 2026 shows enterprise token costs fell 67% year-over-year, driven by open-source model pricing (DeepSeek V4, Qwen 3.6-Plus) and multi-model routing adoption. However, agentic AI token consumption is 5-30x higher per task than chatbot interactions, and context window compounding can balloon a session from 5K to 200K tokens by turn 50. Net effect: per-token savings eaten by volume growth. Teams need explicit token budgets per agent workflow.