As AI agents consume vastly more tokens while operating autonomously, the industry is shifting from flat-rate subscriptions to consumption-based pricing models tied directly to token usage.
The economics of artificial intelligence are undergoing fundamental change. Traditional generative AI operated on a simple model: users paid monthly subscriptions for access to chat interfaces and query limits. Agentic AI workflows shatter this arrangement.
Agentic systems run autonomously for extended periods, executing complex tasks without constant human intervention. This continuous operation consumes tokens at rates many times higher than traditional chatbots. The math no longer works for providers offering unlimited access at fixed prices.
From Subscriptions to Consumption
Providers are shifting toward consumption-based billing, where costs scale directly with token usage. This transition reflects the economic reality: agentic workflows generate unpredictable, often substantial token bills that flat-rate models cannot absorb.
The emerging token economy is becoming more granular. Pricing now varies based on multiple factors: inference speed, model specialization, and the economic value generated by the output. A high-value task producing significant business results may command premium token pricing, while commodity operations cost less.
Market Implications
This evolution creates new pricing tiers across the AI landscape. Providers must balance affordability for users against profitability—a challenge that commodity pricing alone cannot solve. Token pricing becomes a direct reflection of computational cost and market demand.
For enterprises deploying agentic systems, this means budgeting shifts from predictable monthly expenses to variable, output-dependent costs. Organizations benefit when agents deliver high-value results efficiently, but face escalating costs if operations prove inefficient or token-intensive.
The third edition of Frontier Radar maps this emerging token economy, documenting how the industry is pricing autonomy and computational intensity. As agentic AI becomes production-standard across enterprises, token economics will increasingly determine which implementations succeed and which prove unsustainably expensive.
OpenAI is preparing a major redesign of ChatGPT, repositioning the chatbot as a gateway to higher-margin products before a potential public offering. The company is moving beyond the basic chat interface that defined its viral success.
Xiaomi unveiled MiMo-v2.5-Pro-UltraSpeed, a 1 trillion parameter model capable of generating 1,000 tokens per second. The release marks a significant step in large language model inference speed.
Google is rolling out a major upgrade to NotebookLM, its AI-powered note-taking app, featuring the new Gemini 3.5 model and improved source discovery tools. The update aims to deliver more accurate research capabilities across the platform.
Major AI companies preparing for public offerings could drive significant price increases across the sector. Analysts warn of potential cost escalation as firms seek market valuations.