We're short PLTR with roughly 580bps of negative edge: the stock embeds ~43% long-run free-cash-flow growth, a bar we think the business clears in the near term but cannot sustain for a decade. On AIP-driven U.S. commercial (~30.4% implied, ~18% of attention — the heaviest single weight), the boot-camp-to-pilot-to-production engine is genuinely working and explains the 85% YoY print this quarter, but the price requires that conversion AND ontology differentiation AND durable enterprise AI budgets all hold for ten years; we think ontology is a real but narrowing moat as hyperscalers and open frameworks commoditize the data-integration layer, so we're modestly below. On operating leverage (~5.7% opex growth against ~30% revenue), the lean, automation-heavy headcount model is real — op margin has gone from 13% to 46% in eight quarters — but pricing in opex barely growing while AI engineering talent inflates is too aggressive. On U.S. government/defense (~30.4% implied, ~10% of attention), Maven and ShipOS momentum and preferred-source status are durable, but defense software budgets don't compound at 30% for a decade and concentration (one customer at 31% of receivables) is a real fragility. On net dollar retention and large-deal/TCV bookings (~30.4% each), the proven ROI on flagship deployments is the strongest part of the bull case, yet indefinite consumption expansion with no budget consolidation onto hyperscalers is the assumption we'd fade. Across the board the qualitative read is the same: the business is excellent and the near-term trajectory justifies a premium — just not one that capitalizes a decade of ~30%+ compounding plus 43% FCF growth. The biggest risk to the short is that international (U.K. government and commercial) is a still-untapped TAM that re-accelerates the top line and validates the long runway.
| Business factor | What the market prices in | Attention | Our view | Why we are right |
|---|---|---|---|---|
| AIP-driven U.S. commercial demand | ~30.4% annual revenue growth sustained ~10 years | ~17.6% — the single heaviest weight | lower; modestly below ~30% | Boot-camp pilot conversion is genuinely working and drives the current 85% YoY print, but ontology's data-integration moat narrows as hyperscalers and open LLM frameworks commoditize the layer; a decade of ~30% requires conversion, differentiation and AI budgets all to hold simultaneously. |
| Operating leverage / lean model | ~5.7% opex growth while revenue compounds ~30% | ~0% explicit, but underpins the FCF math | lower margin upside; opex grows faster than priced | Margin expansion from 13% to 46% op margin in eight quarters is real, but pricing opex to barely grow while AI engineering talent costs inflate and go-to-market scales is too optimistic. |
| U.S. government / defense AI software | ~30.4% annual growth ~10 years | ~10.4% | lower; defense software budgets don't compound at 30% for a decade | Maven/ShipOS momentum and preferred-source status are durable near-term, but 31% receivables concentration in one customer and finite defense software budgets cap the long-run rate. |
| Net dollar retention / existing-customer expansion | ~30.4% growth from seat/usage expansion ~10 years | ~3.4% | lower; fade indefinite consumption growth | Proven flagship ROI supports strong NDR, but pricing in indefinite consumption expansion with no budget consolidation onto hyperscalers ignores how enterprises eventually rationalize platform spend. |
| Ontology / platform differentiation | ~30.4% growth, moat amplifies LLM value as token costs fall | ~6.5% | lower; moat real but eroding at the margin | Near-term LLM-value translation through ontology is real, but falling token costs expand reach for competitors too, pressuring the pricing power the long-run growth depends on. |
| Large deal closings / TCV bookings | ~30.4% growth as big-deal pipeline scales and converts on time | ~2.4% | lower; conversion timing and budget funding risk | Government funding supports bookings near-term, but the path requires booked TCV to convert to recognized revenue on schedule for a decade — a lumpy, deal-by-deal dependency. |