The Fade Line, Measured

We published the test before the answer. Palantir cleared every threshold — growth accelerated to +93% — and under our own pre-registered rules, the fair value moves. The full valuation, from what the company actually does to what its price now demands.

Yesterday morning we wrote the test. Last night Palantir took it.

Hours before this print, we published The Fade Line: a valuation built backwards from Palantir’s price, compressed into one number — the fade rate, the fraction of each year’s growth that survives into the next — with revenue thresholds for this exact report written down in advance, so the print would grade our framework instead of our hindsight. The grades:

Revenue didn’t just clear the top threshold; growth accelerated, from +85% to +93% — a sixth consecutive quarter of acceleration, at a scale where every prior software company in history was already slowing. EPS came in at $0.41 against $0.33 expected, the sixth beat in seven prints. The stock gapped to $154.22 by mid-morning today — +22.7%, more than double the ±11.4% the options market had priced. And under the re-weighting rule the morning note pre-registered (“two consecutive quarters holding 80%+ growth = evidence for the slow-fade world”), our scenario weights are now obligated to move. This note re-measures everything — and explains, for readers arriving fresh, what this company actually is.

What Palantir actually does

Palantir builds the operating layer that lets a large institution — an army, a hospital network, a manufacturer — run its operations through software that understands the institution’s own data, people, and processes as one connected model. Four platforms make one machine. Foundry hosts the Ontology: a living digital model of the organization where data, meaning, and action are bound together — a data warehouse answers questions; the Ontology executes decisions. Gotham is the defense and intelligence version, wrapped in security accreditations and a twenty-year operational record that no competitor can buy. Apollo is the delivery system that ships all of it anywhere — public cloud, factory floor, air-gapped classified network — from one codebase. And AIP, the 2023-vintage AI layer, is why the company is growing 93%: it connects any large language model to the Ontology with permissions and guardrails, so AI can safely do things — reroute the shipment, flag the claim — rather than chat about them. Sold through week-long “bootcamps,” it closed 220 deals of $1M+ in this single quarter.

Revenue splits roughly half government (Q2 US government: $809M, +90%), half commercial (US commercial: $764M, +149%), with the US at 81% of the total — this is, for now, an American growth story. The quality of the P&L under it has become extraordinary: 47% GAAP operating margin, 55% GAAP net margin, 63% cash-flow margin, and — for the first time — GAAP earnings equal to “adjusted” earnings, the cleanest answer yet to the old stock-compensation critique. All of it is produced by roughly 4,400 employees — about $1.4 million of revenue per person, among the highest figures in software — and summarized by the industry’s “Rule of 40” health score (growth rate plus profit margin, where 40 counts as excellent): Palantir just printed 155.

The moat is real. Here is exactly where it is — and isn’t.

Four reinforcing moat sources, in declining order of strength. Switching costs: once operations run through the Ontology, leaving is an ERP-scale re-implementation, not a subscription cancellation — visible in remaining deal value expanding 27% in a single quarter. Regulatory intangibles: the accreditations, clearances, and two decades of live government history form a wall that funding cannot cross quickly — and recent procurement reform turned that wall into a weapon. Process power: forward-deployed engineers plus Apollo’s deliver-anywhere machinery produce software-company margins on services-company intimacy. Counter-positioning: Palantir sells AI sovereignty — your data never trains someone else’s model — a position the AI labs cannot copy without self-destructing and the cloud giants carry conflicts into.

The forces net out unusually well — two forces favor Palantir outright, and the genuinely adverse one (a government buyer that is ~42% of revenue and negotiates like a sovereign) is offset by how deeply the software is embedded in missions that cannot pause. Where the framework bites is the margin assumption: we model a 40% steady-state cash margin, far below today’s reported 63%, precisely because buyer power and the build-it-yourself substitute should compress pricing as the category matures. The moat protects the business. Nothing protects the multiple.

The pipeline is outrunning the fastest P&L in software

The single most important analytical fact in this report: every leading indicator of Palantir’s revenue is growing faster than revenue itself. Backlog (+137%), remaining commercial deal value (+124%), closed commercial contract value (+153%) — in a fade-rate framework, this is what a slow fade looks like from the inside:

Quality checks, honestly reported. The receivables stretch we flagged yesterday morning cleared: Q2 collections improved to ~70 days of sales outstanding, reversing the first-quarter stretch and back in line with Q4-2025’s best. Customers are paying ahead — deferred revenue up 42% in six months. The two structural costs stand and are charged in our numbers rather than excused: stock compensation (~14% of revenue) and its dilution — this note values the company on 2.571 billion diluted shares, not the 2.397B basic count, a ~7% per-share haircut most Palantir math quietly skips. One disclosure gripe: customer counts and net-dollar-retention have vanished from the company’s releases; for a story built on expansion, removing the cleanest expansion metric is a choice we flag every time until it reverses.

The fade line, re-measured: the demanded growth just got easier while the stock went up

Run the machine backwards on post-print inputs — $369.7B market value at $154.22, minus $9.2B cash, over $6.2B trailing revenue. The multiple fell on the print: 59× trailing sales versus 56× yesterday morning at a lower price, because the business grew faster than the stock. The decade-long growth demanded to earn 10%/yr at a 10×-sales exit eased from 30.7% to 31.3%; the breakeven fade moved from ×0.78 to ×0.78 — the market now needs Palantir to keep only ~78% of each year’s growth rate, seeded from a guide that just went up.

The chart’s left edge is the discipline in this framework: Palantir’s own 2023 — growth collapsing to 17% before the AI wave re-lit it — proves this company’s growth regime has flipped twice in four years, in both directions. Bulls and bears are ultimately arguing about which chapter of that history is repeating; the fade paths simply price each answer.

The odd arithmetic of hypergrowth: at +9% after hours last night this stock was cheaper than before the print. At +23% by this morning’s open it was richer than before it. The multiple, not the quarter, is always the bet.

Weighted fair value: $149 per diluted share — 3.4% below this morning’s price, restated from yesterday’s $112. Three forces moved it, all disclosed in advance or found in the audit: the pre-registered re-weighting toward slow fade (bull 25→35%, bear 30→20%); the print itself (trailing revenue, cash, and the guide all higher); and the switch to diluted shares (which cut ~7% the other way). The honest range — $72 to $230 — remains the widest of any name we cover, and the rate sensitivity is violent: the base case alone runs $146 at a 9% discount rate to $87 at 12%. The deeper reason the range is wide: in every scenario, most of the present value sits beyond year five. Owning Palantir at any price near today’s is, unavoidably, a claim on years none of us can see — discounted through an assumption about how gracefully excellence ages.

Verdict: the market repriced through our number in one night

Yesterday morning at $125 the price sat above our framework’s weighted value of $112; by this morning’s open the market had repriced not to our restated $149 but through it, to $154. One session converted the stock from modestly rich, to modestly cheap in the after-hours, to modestly rich again — while the fair value moved only on evidence. That is not a contradiction; it is what disciplined updating looks like: the fair value moved +33% on evidence while the price moved +23% on mood. At $154 the market pays a modest premium to our weighted value — it has left the base case (growth halves every ~3 years → $120) behind and is leaning toward the bull case (the AI-sovereignty cycle persists → $230). From here, further upside requires the slow-fade world to keep confirming, print after print; the framework no longer calls this price wrong — it calls it a bet with the burden of proof attached. What would move our line again, in either direction, is pre-registered as before: the Q3 print (guide implies ~+83%; below ~$2.08B, +76%, would be the first fast-fade datum), the FY26 guide’s next revision, any quarter under +70% (breaks the breakeven path), two more quarters at 80%+ (bull weight rises again), or disclosure of net dollar retention in either direction — plus one this print put on the clock: a decision on the $9.2 billion cash pile, where a first buyback would materially change the per-share arithmetic. The bear case remains fully alive at $72 — a reminder that a wide moat, a perfect print, and a 49% drawdown in living memory can all be true of the same stock at the same time.