The cost of enterprise-class AI models is falling rapidly. But describing tokens themselves as a commodity is only half right. A token is a billing meter, not a standardized unit of intelligence. One million tokens from one model may not deliver the same business result as one million from another, from a capability, reliability or latency perspective. What’s becoming commodity-like is the metered intelligence that enterprises can increasingly source across model providers. DeepSeek’s V4 pricing illustrates this shift. V4 Flash mode offers discounts of up to 99% off frontier model pricing for comparable coding work, alongside peak and off-peak pricing….

The post Why Are Enterprise AI Bills Rising as Model Prices Collapse? appeared first on Interconnections - The Equinix Blog.