Usage-based SaaS Tokenization Model

📌 Background

In 2026, the cost of large model inference dropped sharply, making traditional subscription models unable to reflect actual usage value. Enterprises have shifted toward billing based on Token consumption or usage volume. Xunce Technology's positive profit forecast shows that its revenue for the first half of 2026 was approximately 967.02 million yuan, a year-on-year increase of 388.76%, and it expects to achieve its first semi-annual profit (based on company announcements); revenue rises naturally with the frequency of AI calls.

👤 Target Customers

Developer teams with high-frequency AI usage and enterprise clients requiring elastic scaling.

💰 Revenue Streams

1) Usage-based billing: Pay-as-you-go based on actual API Token consumption or data usage; 2) Usage packages: Tiered usage packages to lock in prepayments from large clients; 3) Private deployment: One-time fees for localized deployment and implementation per project; 4) Value-added modules: Subscription-based seat pricing for usage monitoring and cost governance tools (opportunity item; revenue scale for seat subscriptions is not publicly disclosed).

🧮 Cost Structure

Cloud computing and GPU compute costs, underlying large model API call costs, and maintenance fees for the billing system architecture.

🛡️ Moat

Extreme compute cost control capabilities, infrastructure for high concurrency and stability, and the industry data flywheel accumulated through early-mover advantage.

🔑 Keys to Success

• Deep optimization of inference compute costs
• Provision of transparent and visual usage monitoring dashboards
• Building a customer success ecosystem that drives usage growth

⚠️ Risks

• Periodic spikes in compute costs eroding profits
• Customer spending cuts leading to a collapse in usage

🏢 Cases

• Xunce Technology: Enterprise data paid by usage, revenue surged by 389%
• Zhipu AI: GLM Coding Plan switched to transparent, unified credit-based billing

📊 SWOT Analysis