This session bridges the gap between technology and business by framing AI infrastructure costs as a new variable in startup economics. It begins with a concise overview of the modern AI/ML stack and the AI Spectrum (Consume → Customize → Build → Orchestrate) as a practical roadmap. The core focuses on token economics: how LLM billing works (input, output, and reasoning tokens), the four most common "token leaks" that silently drain startup budgets, and a concrete optimization playbook with strategies that deliver 30-90% cost savings — including model routing (79% savings), semantic caching, and prompt engineering. Concludes with a 90-Day Optimization Playbook that founders can implement immediately. Ideal for anyone who needs to understand both the technology choices and their financial impact.