Frontier AI labs see their lead as only one or two models ahead of open-source alternatives. Their moat relies on talent, compute and proprietary user data. They now seek to expand into new markets like ads, robotics and wet labs to sustain revenues. Other firms can still enter with billion-dollar training budgets. Corporate AI spending is hitting budget limits and shifting to cheaper models. Labs may either compete on cost or push for regulatory barriers to protect their lead. Recursive self-improvement remains the long-term goal to secure enduring advantage. Current models excel in narrow domains but struggle with broad real-world tasks. Economic constraints may limit how long the frontier can stay ahead. The pace of AI progress will likely continue but with persistent domain-specific limits.