Palo Alto Networks CEO Nikesh Arora has proposed a solution to address what he calls the “AI Business Model Trap,” as companies like Amazon, Uber, and Meta restrict AI use for employees due to skyrocketing costs. Arora argued on X that Large Language Model (LLM) makers need massive cash flow to fund the race toward Artificial General Intelligence (AGI) but lose enormous sums serving consumers with free AI tools to feed post-training data needs. He warned that tech giants are pointing monetization challenges at enterprises, creating a risk of charging exorbitant token prices to businesses while consumers get the same models for free. Arora outlined a three-point approach, stating that Phase 1 involved value capture in coding with low customization, while Phase 2 requires building depth with harnesses, context, memory, and deterministic guardrails. He suggested that if companies want to win enterprise, they should forward price tokens, making them cheaper to allow experimentation and workflow reimagination. Arora advised: “Cut token pricing now, else send enterprise.”