# Insights on AI unit economics

Per-customer margins, pricing strategies, MCP finance workflows, and the operational realities of running LLMs in production.

[**The AI Gross Margin Crisis Nobody Talks About**](/content/tokenops/blog/ai-gross-margin-crisis/index.html)  
The average AI-native B2B company runs 52 percent gross margin. Classic SaaS runs 80. The gap is inference cost, and it is getting wider.

[**Your Inference Bill Is Not R&D. It Is COGS.**](/content/tokenops/blog/inference-is-cogs/index.html)  
LLM API spend is classified under engineering at most AI companies. That misclassification hides per-customer margins, distorts pricing, and delays the conversation your board will have.

[**Voice AI's Hidden Margin Problem**](/content/tokenops/blog/voice-ai-margin-problem/index.html)  
Voice AI companies stack STT, LLM, and TTS costs per call. At scale, the per-minute COGS compresses margins faster than any other AI modality. Here is the math.

[**AI Gateways Cannot Price Without Margin Attribution**](/content/tokenops/blog/gateway-margin-attribution/index.html)  
AI gateway and routing companies resell tokens. Their gross margin is the spread between what they pay vendors and what they charge customers. Without per-customer cost tracking, that spread is invisible.

[**Agentic Workflows Burn 100x More Tokens. Your Pricing Does Not Know.**](/content/tokenops/blog/agentic-token-cost/index.html)  
Multi-agent reasoning tasks use orders of magnitude more tokens than simple completions. If your pricing has not adjusted, your margin is shrinking with every new feature you ship.