Per-second billing model for edge infere Diagram

Strengths

• Low-latency inference experience enabled by edge node distribution
• Granular per-second billing reduces trial-and-error costs for developers

Weaknesses

• High GPU resource costs, with gross margins susceptible to compute price fluctuations
• High functional overlap with platforms like Hugging Face and Together AI

Opportunities

• Continued growth in LLM inference usage, driving strong demand for aggregation and hosting
• Low-cost customer acquisition through conversion of existing Cloudflare clients

Threats

• Price cuts on direct APIs from model providers squeeze hosting premiums
• Lower barriers to self-deploying open-source models weaken platform value