AI's real-world-data gap → daily-global Earth archive → Planet data moat

LLMs trained on internet text are "blind" to the physical world; capability is data-bound. Cheap launch + satellite miniaturization make a daily-refresh global Earth archive economical, and Planet already operates the largest one. If "large earth models" become the grounded-data layer for AI, Planet's archive is the scarce input — re-rating PL from an imagery vendor into an AI-data company.

• Step 2: A competitor (e.g., a SpaceX/Starshield-class or Chinese constellation) matches daily-global refresh and depth of archive, eroding the scarcity claim.
• Step 4: Through 2027, Planet's revenue mix stays dominated by traditional government/imagery contracts with no material "AI-data / large-earth-model" line — the re-rate premise fails to show up in the financials.

• **Single, self-interested source.** Every step rests on the Planet CEO's framing in a friendly podcast. The "large earth models" demand category is asserted, not yet demonstrated by a paying frontier-lab customer.
• The defense/intel revenue that funds the business today (~60%) is not obviously the same demand as "LLMs need grounded data" — the AI-data thesis may be a narrative layered over a government-imagery business.
• Imagery is increasingly commoditized (multiple constellations); "largest archive + daily refresh + full history" is the claimed moat, but the durability of that moat against newer entrants is untested here.
• **Challenged in-source by alexander-wissner-gross** in 2026-06-26-podcast-moonshots-the-10b-satellite-empire-putting-ai-in-orbit-why: "You really think that's true? You don't think there are like millions of first person videos of people seeing trees on YouTube?" — i.e. frontier multimodal models may already carry enough physical-world grounding that an orbital archive is not the scarce input. will-marshall's rebuttal is embodiment, not data volume: "they haven't gone outside the library". Unresolved; it is the load-bearing objection to Step 4.
• Marshall is Planet's CEO describing Planet's moat. First-party and self-interested at once.

• Tradeable: **PL** as an AI-data play rather than a pure imagery vendor — the re-rate hinges on whether "large earth models" become a real demand category.
• Adjacent to terrestrial-power-flat-to-orbital-dc-arbitrage: the same launch-cost decline that makes the fleet cheap is what eventually makes orbital compute viable — Planet is positioned on both legs (data now, compute later).
• A grounded-real-world-data thesis generalizes beyond Planet (any proprietary physical-world dataset gains value as AI saturates internet text), but Planet is the cleanest listed instance.