RE: LeoThread 2025-11-15 01-05

Physical AI, however, depends on massive supply chains and factories, which will remain a huge bottleneck for the next ~20 years

This explains why physical AI companies (like Tesla) look significantly undervalued, while digital AI firms—especially bolt-ons to ChatGPT, Grok, etc.—seem likely overvalued

Physical AI firms can deliver a much lower cost per unit of labor than human equivalents (e.g., drivers) and hold super-defensible moats such as factories and supply chains

The investment thesis: most AI application startups are likely to be overwhelmed by rapid expansion of foundational model providers

App features will be folded into foundational models because the major players are simply huge, creating a fast-moving wave that can obsolete new apps almost as soon as they're invented

That leaves almost no time to build and scale a standalone company
Two realistic monetization routes for AI app startups:

• Launch a short-lived hit that generates strong cashflow and bank it; estimated cashflow window ~12–18 months
• Build an app good enough to be acquired by a major provider for meaningful equity

The environment is highly unstable — whether it crashes or soars, both outcomes make it unlikely many AI app startups will become generational supercompanies

The best odds come from targeting a niche with extremely specific data barriers, ideally tied to physical-world (hardware or real-world) data rather than pure software/finance