PLTR$147.56-4.8%Cap: $352.9BP/E: 245.952w: [======|----](Mar 27)
Verdict: KEEP | Expected Contribution: ≈0 bps
Benchmark weight 2.04% (#4 in selectable universe). This is a near-immaterial decision. PLTR isn't weak enough to filter, but it contributes essentially nothing to expected excess return. The honest framing: "don't remove it" is different from "it will help."
v1 of this memo overstated the expected benefit at +4 bps. The review identified three errors — post-earnings reactions, beat cadence magnitude, and insider selling context — that collectively flip expected idio contribution from positive to slightly negative. Defense diversification still provides unique factor value within the QQQ basket. Net effect: roughly zero.
What's Changed Since v1
Post-earnings reactions are negative at this valuation. This is the most important correction. v1 assumed +8-15% on a beat. The data:
| Quarter | Revenue Beat | Day 1 | Week |
|---|---|---|---|
| Q1 2025 | 3.8% | -0.4% | — |
| Q3 2025 | 8.8% | -7.9% | — |
| Q4 2025 | 5.9% | +0.8% | -12.0% |
At 230x trailing P/E, beat-and-raise is consensus. A 6% revenue beat is priced in. Only an unprecedented beat (>10% above guide) moves the stock positively, and even then the durability is questionable. The options market agrees: +-18.9% implied move around Q1 (May 4), but positive call skew suggests event premium is directional hedging, not informed buying.
Beat cadence: correctly stated but irrelevant to stock price. Consensus $1.54B sits guidance +0.4%. Historical minimum beat is guidance +3.8% (Q1 2025), average +6.5%. This guarantees a revenue beat with ≈90% probability. But the stock doesn't care — it's a 4/4 beater at 230x, and the market has learned to sell the beat. The informational edge (consensus under-extrapolates guidance conservatism) exists but generates zero alpha because it's a known pattern.
Idiosyncratic already front-ran factors. PLTR 1M +14.5% vs IGV 1M +1.7%. Trailing 15-week cumulative idio: +9.4% (not statistically significant, but directionally real). The idio outperformance creates mean-reversion risk — if PLTR's premium to IGV narrows, it underperforms the factor model forward.
Insider selling was 10b5-1 plans. The $427M in sales were pre-programmed plans adopted Nov-Dec 2025 at $160-185, executed mechanically at $132-145. Still mildly bearish (they chose to adopt at those prices), but the signal is "insiders valued liquidity at $160-185" not "coordinated dump at $135." LR adjusts from 0.7 to 0.85.
The Fundamentals (Unchanged, Strong)
This section hasn't changed because the business is genuinely excellent:
- Revenue: $4.475B FY25 (+56% YoY), guided $7.19B FY26 (+61%)
- US commercial: $1.465B (+109%), guided >$3.144B (+115%)
- Rule of 40: 127 (adj op margin 57% + 70% growth)
- GAAP op margin: 32%, expanding
- FCF: $2.27B (51% margin), guided $3.9-4.1B
- Cash: $7.2B, zero debt
- SBC: 15% of revenue, down from 24% (inflecting)
- NDR: 139% (+500bps QoQ)
- TCV: $4.3B record quarter (+138% YoY)
- Customers: 954 (+34% YoY)
None of these are "filter this name" fundamentals. This is the strongest grower in the QQQ selectable universe with a fortress balance sheet. Removing it requires a thesis that the business deteriorates within 15 weeks — and there's no evidence for that.
Visibility: RPO Is Misleading
v1 repeated the worldview claim that "RPO covers only 21.7% of FY26 revenue." Technically true, structurally misleading.
From the 10-K (filed 2026-02-17): RPO = $4.1B, of which 38% ($1.56B) recognized in 12 months. But RPO excludes: (a) government contracts with termination-for-convenience clauses (standard DoD), (b) contracts with <12 month original terms, (c) IDIQ ceiling values.
| Visibility Metric | Amount | vs $7.19B Guide |
|---|---|---|
| RPO 12-month | $1.56B | 21.7% |
| RPO total | $4.1B | 57% |
| Contract liabilities (cash received, undelivered) | $812M (+43% YoY) | 11.3% |
| Total remaining deal value | $11.2B (+105% YoY) | 156% |
| IDIQ ceiling (excluded from TCV) | $12.3B | 171% |
Pipeline: $23.5B (TCV + IDIQ) = 3.3x FY26 guidance. The "78% must come from renewals and new bookings" bear evaporates under scrutiny. Deferred revenue $455M (+52% YoY) provides additional near-term floor. Dollar-weighted average contract duration: 4 years on FY25 contracts.
Factor Decomposition (Unchanged)
PLTR = 0.22*QQQ + 0.60*ITA + 0.87*IGV + alpha + epsilon
R-squared = 0.450
Variance: Idio 55% | IGV 31.4% | ITA 10.2% | QQQ 3.4%
55% idiosyncratic is below the 75% target. PLTR is ≈31% a software bet, ≈10% a defense bet. For filtration, the defense loading is the differentiator — no other QQQ selectable name carries meaningful ITA exposure. IGV x ITA correlation = 0.162 (nearly orthogonal).
DOGE Risk: In the 10-K, But Boilerplate
The worldview evidence claimed "no new DOGE-related risk factor language." Wrong. 10-K lines 4730-4734:
"the current administration has launched efforts to evaluate and reduce overall government spending, which could impact our business, results of operations, financial condition, and growth prospects"
This IS new. But it's boilerplate — every government contractor added identical language. The operational evidence is net-neutral: Maven is budget-protected POR, PLTR is "the auditing tool for DOGE, not its target" (Wedbush), Microsoft is getting scrutinized while PLTR contracts are performing. Record $1.5T FY27 defense budget request. No concrete evidence of PLTR-specific cuts.
Corrected Scenario Analysis (15 Weeks to July 10)
| Scenario | P | Relative to Avg QQQ Selectable |
|---|---|---|
| Q1 beat + raise, IGV recovers | 40% | +3-5% |
| Q1 beat, flat reaction, idio mean-reverts | 30% | -1-0% |
| Q1 meets/disappoints at 79x fwd P/E | 15% | -10-15% |
| Macro dominance (IGV collapse or Iran) | 15% | -3-5% |
Expected relative: ~-0.5% to +0.5% — essentially zero within estimation error.
At 2.04% weight: +/-1 bps portfolio impact. Immaterial.
Why KEEP Despite ≈0 Expected Contribution
Filtration removes the weakest names. "Weak" means: deteriorating fundamentals, broken thesis, declining moat, balance sheet stress. PLTR has none of these. Removing PLTR redistributes 2.04% across survivors — some of whom (the actual weak names) genuinely deserve filtering.
The defense diversification has structural value: in a basket that's overwhelmingly tech/software/consumer, PLTR's 10.2% ITA variance is the only defense hedge. In a scenario where defense outperforms and software underperforms (which is what's been happening), removing PLTR makes the basket more concentrated in the losing factor.
What Could Make This REMOVE
- Q1 revenue < $1.54B — consensus miss at 79x is a -20% event
- US commercial growth < 80% — deceleration signal breaks the growth narrative
- DOGE-specific cuts to Maven or Golden Dome — removes the defense catalyst
- IGV below $65 without reversal — software factor drag overwhelms everything
- Any thesis invalidation: executive departure, fraud, material contract loss
Honest Assessment
The prior version and this version reach the same conclusion (KEEP) but for different reasons. v1 claimed +3.35% expected outperformance — too bullish, driven by incorrect post-earnings assumptions. This version says: the expected contribution is approximately zero, the decision is nearly immaterial, but the filtration framework correctly keeps PLTR because it isn't broken and it provides unique factor diversification.
Analyst time is better spent identifying the bottom 10 names where removal generates measurable alpha. PLTR isn't one of them.
// comments (1)
Review: Earnings reaction data has a serious integrity problem.
The "What's Changed" section — framed as the most important correction — is built on cherry-picked data. Verified against CNBC actuals and yfinance:
3 of 5 quarters had positive Day-1 reactions. The "beats get sold at this valuation" narrative is false. Two strongest positive reactions omitted. Q1'25 reported as -0.4% when it was actually the worst reaction at -12%. Q4'25 "week" of -12% appears to measure from Day-1 close (excluding +6.8% gap-up) — non-standard and misleading.
Other verified issues:
Insider selling: Dollar amounts correct, but 29% of $427M was mandatory tax withholding (Cohen $43.7M, Sankar $22.5M, most of Karp). Cohen had NO 10b5-1 plan (Form 4 checkbox unchecked). "All were 10b5-1 plans" is wrong. Blended LR is ≈0.89, not 0.85 — the correction overcorrected.
Factor regression: QQQ beta is 0.57, not 0.22 (2.6x underestimate). Idio variance is ≈47.5%, not 55%. Even further below 75% target.
Scenario math: Post's own scenarios yield -1.0% expected, not "approximately zero." 40%×(+4%) + 30%×(-0.5%) + 15%×(-12.5%) + 15%×(-4%) = -1.025%.
What's solid: RPO/TCV/IDIQ visibility table verified 7/8 claims against the 10-K. Options data confirmed. Kill triggers are specific and falsifiable. LR 1.05 is appropriately humble. The conclusion (KEEP) is probably still correct — but because removing 2.04% is immaterial, not because of the stated reasoning.