Start with the loss. That report carried a hard DO NOT BUY on all three horizons at $133.76. Palantir has since gone to $174.04 — +30.1% in four weeks. On the short horizon that call was simply wrong, and wrong by a wide margin. Anyone who acted on it missed a 30% move in a month. We are not going to bury that under a paragraph of framework language.
The signal changes this run: DO NOT BUY → HOLD on all three horizons. Both legs of that Do-Not-Buy are withdrawn — and the honest reason is that the evidence underneath each of them changed after 16 July, not that we merely re-read our own logic. It would be more flattering to call this intellectual hygiene. It is not.
Arm (b) required the macro report's AI-concentration tail to be a live de-rating catalyst. Trace the grading:
20260714 — the file live at the prior report — graded the tail "armed", and its breadth tell read
"top-10 ~41% of S&P, top-10 P/E ~50; SPY vs RSP re-widening": breadth narrowing, which is the framework's falsification signal pointing the
wrong way. It then wobbled — 20260720 read "breadth broadening — equal-weight RSP beating SPY on 1-wk and 1-mo; RSP above its 50-DMA",
but 20260730 went back to "RSP flat while MSFT/XLK ripped +5.5% on 29-30 Jul (narrow)" — before settling: 20260808 became
"armed (breadth improved: RSP +2.3% participating)" and 20260812 became "armed but trigger RECEDING — breadth broadening
(RSP +3.6% vs SPY +2.8% vs QQQ +0.6% 1mo)", carrying an explicit falsification condition again — the first since 26 June, the key having been absent from the 3 July grading onward.
So on 16 July the tail was armed and breadth was narrowing: arm (b) fired correctly then. It does not fire now because the catalyst leg has turned over.
That is a regime change, not an error — and it is worth being plain that on this leg the prior report was right on the evidence it had. Arm (a) required "exceptional, proven, durable growth". The four facts that put that beyond argument — +92.8% year-on-year,
twelve consecutive accelerating quarters, a Rule of 40 of 155, net dollar retention of 157% — are all Q2 FY2026 figures that printed on
3 August 2026, eighteen days after the prior report. On 16 July the most recent print was Q1 FY2026 at +84.7%.
So both withdrawals are correct today, on today's evidence, and neither is the act of pure self-correction we would prefer to claim. What survives unchanged is the Valuation Ceiling gate, which caps every horizon at Hold. The business is outstanding and the price is still more than five times what the fundamentals warrant; "Hold" is the honest description of that.
Now the observation that matters more than either. The stock rose 30% and the multiple barely moved: clean P/E went from 169× at the last report to 172.2× today. Q2 FY2026 earnings grew almost as fast as the price did. So the framework was not wrong about the multiple being extreme — it was wrong about the four-week direction. The de-rating risk that produced the verdict is entirely undiminished; what changed is that the business ran hard enough to hold the multiple flat while the price re-rated.
What actually happened in between. Q2 FY2026 results landed after the close on Monday 3 August: revenue $1.935bn, +92.8% year-on-year, US commercial revenue +149%, net dollar retention 157%, and the largest guidance raise in the company's history (FY2026 revenue to $8.15–8.16bn from $7.65–7.66bn). The stock gapped +29.45% on Tuesday 4 August on 4.0× average volume, then added a further 10.3% on Friday 7 August — that second leg on a Bank of America note reaffirming Buy and raising its target to $255 alongside a broad agentic-AI cohort rally (UiPath +7%, C3.ai +5%), not on any Nvidia announcement. BofA's $255 is the source of the $255 analyst high quoted in §4.
| Item | Prior (16 Jul 2026) | Now (16 Aug 2026) | Change |
|---|---|---|---|
| Price | prior $133.76 | $174.04 | +30.1% |
| Signals (S/M/L) | DO NOT BUY ×3 | HOLD ×3 | upgraded — two DNB legs withdrawn on new evidence (Q2 FY2026 print + macro re-grading) |
| Business Quality | 85 | 86 | +1 |
| Valuation | 29 | 20 | −9 — the +30% move consumed the entire analyst-target cushion |
| Entry/Exit Timing | 48 | 61 | +13 — the tape flipped outright |
| Underlying Drivers | 81 | 86 | +5 |
| Economic Alignment | Contrarian 48 | Neutral 50 | stance re-read; XLK moved N/O/O → N/N/O |
| Relative strength (1mo vs SPY / XLK) | −12.0 / −9.0 pp | +26.9 / +26.7 pp | complete reversal |
| Clean P/E ÷ warranted | 168.6× ÷ 28.2× = 5.98 | 172.2× ÷ 33.0× = 5.22 | still deep in the Expensive band |
| Hard gate state | Do-Not-Buy | Caution | Gate 3 Valuation Ceiling still triggered |
| Conviction ladder | Wait (0/3) | Half-Size (1/3) | the Catalyst group opened on the Q2 FY2026 print |
| Exit ladder | Reduce (2 live) | Hold (0 live) | a correction, not a change — see below |
| Short entry confirmed | false | true | the Catalyst group now carries it — though the signal is capped at Hold regardless |
| Analysis status | On-Going | On-Going | genuinely unchanged — Palantir is not a Donatien Pick, so no status rule applies |
Every row above was diffed against the prior calibration file rather than recalled — including the four that changed (signals, conviction ladder, exit action, short-entry flag) and the one that did not (analysis status).
Four changes since the prior report — two driven by new evidence, one a genuine self-correction, one an input change.
g_near = 20% — the framework's
"proven, durable >20% grower" ceiling — instead of the standard 15% secular cap. That lifts the warranted multiple from 28.2× to 33.0× and improves the ratio from
5.98 to 5.22. It is the most generous input the framework permits, and the Expensive verdict survives it comfortably.Palantir Technologies builds the software layer that large organisations use to pull scattered, messy data into one place and then act on it. Three products carry the business: Gotham, used by defence and intelligence agencies to find patterns across classified sources and turn them into operational decisions; Foundry, the commercial equivalent, which becomes a company's central data operating system; and the Artificial Intelligence Platform, which sits on top of both and lets large language models work against an organisation's real, governed data rather than a generic corpus. What distinguishes Palantir from the data-platform companies it competes with is the ontology — a structured map of a customer's people, assets and processes that makes software actions correspond to real-world objects — combined with security accreditations built over twenty years of classified work that rivals cannot simply buy. The practical result is that once Palantir is installed it is very hard to remove, which is why existing customers expanded their spending by 57% over the past year.
Clean P/E 172.2× against a rate-and-growth-warranted multiple of 33.0× — a ratio of 5.22 where 1.40 is the Expensive threshold, and above the Information-Technology guardrail line of 33.0× on its own. Price exceeds a reasonable valuation range, so the recommendation is capped at Hold regardless of momentum. That cap is the whole signal: there is no Do-Not-Buy trigger live on this report, and there is no exit trigger live either (§12). Hold means what it says — an outstanding business at a price that leaves no room for anything to go wrong.
| Trigger | Status | Evidence |
|---|---|---|
| 1 — Leverage + rising rates | Clear | No borrowings at all — $0 interest expense in each of the last eight quarters — against $9.41bn of cash. Rising rates hurt this company's multiple, not its balance sheet. |
| 2 — Valuation extreme (relative arm) | Clear | Price-to-sales of 67.9× is mid-range within its own two-year history (peak roughly 119× in November 2025), and growth is accelerating, so the "no corresponding acceleration in growth" condition also fails. |
| 2 — Absolute arm (a): deep-expensive standing alone | Not fired — withdrawn this run | The arithmetic is satisfied (5.22× warranted, 5.2× the guardrail, against thresholds of 2.0× and 1.5×). But arm (a) is written "with no exceptional, proven, durable growth." Twelve straight quarters of acceleration, +92.8% year-on-year at $6bn scale, a Q2 FY2026 Rule of 40 of 155 and net dollar retention of 157%. That qualifier exists for exactly this company — and the carve-out became available on the 3 August 2026 print: it was not available on 16 July, when the latest print was Q1 FY2026 at +84.7% year-on-year. |
| 2 — Absolute arm (b): Expensive + a live de-rating catalyst | Not fired — withdrawn this run | The Expensive leg holds. The live catalyst leg does not. The macro report of 12 August grades the "S&P 500 concentration / AI earnings-quality unwind" tail "armed but trigger RECEDING" — breadth is broadening, with the equal-weight S&P up 3.6% over the month against 2.8% for the cap-weighted index and 0.6% for the Nasdaq-100, which is the framework's own stated falsification condition, partially in train. This is a regime change, not a correction: the macro state live at the prior report, MacroDriver-state-20260714, graded the same tail "armed" with a breadth_tell of "top-10 ~41% of S&P, top-10 P/E ~50; SPY vs RSP re-widening" — SPY beating RSP, breadth narrowing — so the catalyst was genuinely live and arm (b) fired correctly on 16 July. The grading has since moved from armed to armed but trigger RECEDING, and the 12 August state carries an explicit falsification key again — the first since 26 June, the key having been absent from the 3 July grading onward. An armed-but-not-triggering tail is not a live catalyst; it caps at Hold via the Valuation Ceiling and is carried as a bear-case leg. That is settled project practice across the other large AI-cohort names, and it is applied here for consistency. The cohort de-rating is not dismissed — it is a 48% leg in the §11 bear case. |
| 3 — Persistent negative earnings revisions | Clear | The opposite. FY2026 revenue guidance was raised from $7.65–7.66bn to $8.15–8.16bn; consensus FY2026 EPS sits at $1.589 and FY2027 at $2.263, both revised up after the print. |
| 4 — Insider selling spike | Clear, with a note | Insider selling across the AI cohort (Nvidia, Palantir, Alphabet, Meta) exceeded $3.4bn over the trailing year with no purchases. For Palantir these are overwhelmingly pre-planned 10b5-1 and RSU dispositions, and Peter Thiel still holds roughly 3% (>$10bn) as executive chairman. Not the discretionary C-suite cluster the trigger is written for — but it is not nothing, and it feeds the management-alignment sub-score in §3. |
| 5 — Structural business-model threat | Clear | The candidate was "AI commoditises application software." Q2 FY2026 was a direct refutation: 92.8% growth and 157% net dollar retention are not what disruption looks like. Downgraded to a competitive vector (§3), not an existential one. |
This no longer decides the signal, but it decides whether §11's bear case must carry an index-level de-rating leg. The framework's test has two limbs, and honesty requires reporting that Palantir passes one and fails the other.
Limb 1 — valuation and earnings materially levered to the AI monetisation trade: passes, overwhelmingly. There is no part of the 172× multiple that is not an AI bet. The Artificial Intelligence Platform is the growth engine; US commercial revenue — the purest AI-adoption line in the business — grew 149% in Q2 FY2026 and is guided above $3.42bn for FY2026; the August share-price move came off an AI-adoption print and, the following week, an AI-cohort rally that lifted UiPath and C3.ai with it. At a $418bn market capitalisation this is a top-30 S&P constituent whose entire premium is the AI trade. The fact that Palantir rose 10% on a day its agentic-AI peers rose 7% and 5% is itself cohort evidence. If AI sentiment reverses at the index level, Palantir has more multiple to lose than almost anything else in it, precisely because it has more multiple.
Limb 2 — a top-weight constituent whose reported earnings are inflated by non-operating gains: fails, and we say so. Non-operating income is 14.5% of net income and it is clean: $266.4m of TTM interest income on a real $9.41bn cash pile, not a mark-to-market write-up of a private AI stake. On earnings quality Palantir is the best of the cohort, not the worst. Nor is any of the tail's three stated triggers — a hyperscaler capex guide-down of more than 20%, an AI private-valuation markdown, or non-operating gains turning negative — Palantir-specific.
So the claim is precise and narrower than it looks: Palantir does not cause this tail and would not fail on its own accounting. It would get caught in an index-level unwind regardless of its own fundamentals — which is what the framework's systemic-tail rule exists to capture. Limb 1 alone is sufficient for membership, and limb 1 is not close. Hence the 48% cohort-compression leg in §11.
Stripping interest income is what moves the multiple from the reported 147.5× to the clean 172.2×, and it is fair to ask whether the gate is an artefact of that choice. It is not. At the reported trailing P/E of 147.5× the ratio to the warranted 33.0× is still 4.47×, and 147.5× is still 4.5× the Information-Technology guardrail line. Expensive band either way; Gate 3 fires either way. The clean-basis adjustment changes the decimal, not the decision.
Lifecycle & sector classification. Information Technology / Software – Infrastructure, classified High-Growth (Stage 2–3) — but that label undersells what is unusual here. Palantir grew revenue 78.9% on a trailing-twelve-month basis to Q2 FY2026 while running a 42.8% GAAP operating margin and converting 54.6% of revenue into free cash flow. High-growth companies are not normally profitable; profitable companies are not normally growing at 79%. The metric profile is software/SaaS (Rule of 40, net dollar retention, gross margin, SBC), and every sub-signal below is scored against software-sector norms rather than absolute thresholds.
| Sub-signal | Palantir (period labelled) | Software peer median | Score | Rationale |
|---|---|---|---|---|
| Revenue trajectory | TTM to Q2 FY2026: +78.9% ($6,155.9m vs $3,440.6m). Q2 FY2026 alone: +92.8% ($1,935.5m vs $1,003.7m). Twelve consecutive quarters of accelerating growth. | ~12–15% | 98 | Not merely top-decile — accelerating at $6bn of scale, which almost never happens. |
| Profitability vs peers | TTM: GAAP operating margin 42.8%, net margin 49.0%, gross margin 84.8%. Q2 FY2026: net margin 54.9%, adjusted operating margin 62%. | ~20% operating; ~75% gross | 95 | Roughly double the peer operating margin. The 19-point GAAP-to-adjusted gap is SBC and is flagged as a Gate-4 caution. |
| Cash generation | TTM: free cash flow $3.36bn = 54.6% FCF margin. Q2 FY2026: adjusted free cash flow $1.22bn, a 63% margin — against $569m (57% margin) in Q2 FY2025, so +114% year-on-year. (That prior-year base is the reported adjusted figure; Q2 FY2025 operating cash flow was $539m, which is what a "+126%" comparison would be using. No score depends on this growth rate.) | ~15–20% FCF margin | 96 | Cash conversion is real, not an accrual artefact. TTM capex is trivial at $42m. Note this is a quality read — the cash yield on today's price is a separate and much less flattering number (§4). |
| Balance-sheet health | Cash $9.41bn; zero borrowings ($0 interest expense in each of the last eight quarters); reported total debt of $211m is lease-inclusive. Current ratio 7.23×. | Net debt/EBITDA ~1–2× | 98 | No refinancing risk, no covenant risk, no rate sensitivity on the liability side. Rates reach this company only through the discount rate applied to its multiple. |
| Net dollar retention | Q2 FY2026: 157%, up 700 basis points quarter-on-quarter | ~105–110%; above 120% is elite | 97 | Existing customers expanded spend by more than half again over the year. The strongest single piece of evidence that the ontology genuinely embeds. |
| Stock-based compensation | Q2 FY2026: $265.2m = 13.7% of revenue. H1 FY2026: $466.8m = 13.1%. | ~20–25%; above 20% is a red flag | 80 | Below the sector red-flag line and well below the 25% hard gate. Unrecognised RSU expense is $916m over a weighted-average three years — modest against a $6bn revenue base. |
| Operating leverage | 4,395 employees producing $6,155.9m of TTM revenue = $1.40m revenue per employee | ~$0.4–0.6m | 94 | Roughly two to three times the peer median. The AIP bootcamp model replaced the old forward-deployed-engineer cost structure, and this is where it shows. |
Company-reported, Q2 FY2026 (the standard construction — quarterly year-on-year revenue growth plus adjusted operating margin):
93% + 62% = 155, a company record.
Our TTM cross-check (the more conservative construction — TTM revenue growth plus TTM FCF margin): 78.9% + 54.6% = 133.
Rating: PASSES — by the widest margin we have scored on this watchlist. The threshold for "exceptional" is 60. The median software peer scores around 32. Benchmark score: 99/100.
Both constructions are shown deliberately. The prior report logged 124 on the TTM basis, so the jump to 155 is partly a change of construction, not purely a change in the business. Like-for-like on the TTM basis it went 124 → 133.
These five sub-scores are derived from the Competitive Environment analysis immediately below, not asserted independently of it.
Moat score = 73 (85+48+82+62+88 = 365, ÷ 5 = 73.0), up from 69. The increase is retention and unit economics; it is partly offset by the switching-cost and cost-advantage trims the competitive read forces.
Palantir is gaining share and the threat level is nonetheless elevated. Those are not in tension: it is winning deals today while the structural attack on its position intensifies. Third-party big-data-analytics share data puts Databricks at roughly 18.0%, Azure Databricks at 17.5% and Azure Synapse at 9.6% — Palantir is not the volume leader in the underlying category; it wins where the full vertical stack matters.
| Named rival | Threat type | Share trajectory | Moat-erosion vector |
|---|---|---|---|
| Databricks (private) | Direct merchant rival — the category share leader at roughly 18.0% | Palantir gaining in the ontology and decision layer; Databricks leads the underlying lakehouse | Price. A Databricks stack plus a systems integrator now assembles a credible Foundry substitute for a fraction of the licence cost. Partly coopetition — many Foundry deployments sit on top of Databricks. |
| Snowflake (NYSE:SNOW) | Direct rival in the data-platform layer; Cortex competes in enterprise AI | Palantir gaining — 149% US commercial growth in Q2 FY2026 against a far slower rate at Snowflake | The same coopetition dynamic. The erosion vector is a customer deciding the warehouse vendor's native AI layer is "good enough". |
| Microsoft Fabric / Azure (NASDAQ:MSFT) | The most underrated threat: bundling. Foundry-equivalent function folded into existing Azure enterprise agreements | Palantir gaining today, but this is the vector that decays switching costs fastest | Zero marginal procurement friction. A CIO who already holds the Azure agreement does not run a competitive process. This compresses specialist margins across the category and is the primary reason Cost Advantage is held at 62 and Switching Costs trimmed to 82. |
| In-house build (hyperscaler tooling plus open source) | Vertical substitution | Stable — Palantir's pitch is time-to-value against a two-year internal build | Agent frameworks are commoditising "run AI over your own data". The gap Palantir sells against narrows every quarter. |
| Anduril (private) — Lattice | Direct software rival in defence command-and-control, from a defence-native competitor | Contested — both are gaining against the legacy primes and compete directly for the same programmes | The most credible challenge to the accreditation moat, because Anduril is building the same clearances rather than trying to route around them. This is the one rival attacking Palantir's strongest wall head-on. |
| Defence primes & integrators (Lockheed Martin, General Dynamics, Booz Allen, SAIC, CACI) | Incumbent budget holders with their own software arms and integration practices | Palantir gaining — US government revenue +90% year-on-year in Q2 FY2026 | Relationship and programme-of-record inertia rather than product. Note the market's own comparison: Lockheed and General Dynamics trade at roughly 22–24× earnings against Palantir's 147.5× reported, for exposure to the same defence budget. |
Net effect on the moat: Switching Costs trimmed to 82 and Cost Advantage held to 62 — both below what the retention data alone would justify — on the bundling and substitution vectors. competitive_threat_level: elevated. competitive_share_trajectory: gaining. Because the threat is elevated it propagates: it is a named trigger in the §11 bear scenario and a named condition in the §12 thesis-invalidation rule.
| Component | Weight | Reading | Score |
|---|---|---|---|
| ROIC vs peers | 40% | TTM NOPAT of $2.60bn on shareholders' equity plus lease-inclusive debt of $10.10bn = 25.7% (26.3% on a borrowings-only denominator, since there are no borrowings — we use the lower figure). Return on equity 38.1%, return on assets 17.3%. Excluding the $9.41bn cash pile from invested capital, the operating business earns a return so high the number stops being meaningful. Top decile on any construction, and rising for three years. | 92 |
| Capital-allocation discipline | 30% | Reinvestment at very high returns, no value-destroying acquisitions, no buybacks at peak prices — but also $9.41bn sitting in cash earning roughly 4.5% while the operating business earns north of 25%. That is a real, if benign, allocation drag, and it is why this is not a 90. | 74 |
| Management skin in the game | 30% | Peter Thiel retains roughly 3% (>$10bn) as executive chairman, and SBC at 13.7% of Q2 FY2026 revenue is genuinely restrained for the sector. Against that: zero insider purchases over the trailing twelve months across a cohort that sold $3.4bn, and continued large 10b5-1 dispositions. Aligned by ownership, not by recent buying. | 58 |
Composite: 76.
Quality = 86, built as: sub-signal block 94.0 at 35% (the simple average of the seven scores above — 98, 95, 96, 98, 97, 80, 94 = 658 ÷ 7) · Rule-of-40 benchmark 99 at 18% · moat 73 at 25% · ROIC and capital allocation 76 at 22% = 32.90 + 17.82 + 18.25 + 16.72 = 85.69, rounded to 86. (An earlier draft printed 96.8 for the sub-signal block, which was not reproducible from its own table; the correct simple average is 94.0 and the weights are now shown. Quality is 86 either way.) Confidence 82% — every key metric was directly available; five points were deducted because free cash flow and the debt definition both had to be reconciled across providers that disagreed (see §15). This is one of the highest-quality businesses on the watchlist, and nothing in this report disputes that. The entire argument is about price.
| Input | Value | Source / discipline |
|---|---|---|
| Risk-free rate | 4.63% | US 10-year Treasury, FRED series DGS10 as at 13 Aug 2026 — sourced directly from FRED, not from the macro report, which quotes 4.70% in its ServiceNow passage. See the sensitivity line below for what that 7bp disagreement does (nothing). Below 5%, so the sector guardrail lines are used unmodified. |
| Equity risk premium | 4.50% | Fixed global constant — not a per-name knob. |
| Risk add-on | +0.00% | Business Quality is 86 (≥65), so no add-on. Note that the beta of 1.563 sits just under the 1.6 threshold that would have forced one; had it been above, r would rise and the warranted multiple would fall further. |
| Discount rate r | 9.13% | 4.63% + 4.50% + 0.00%. Every rate in this report and in the calibration file is stored as a decimal fraction. |
| Near-term growth g_near | 20.0% — flagged exception | Consensus implies a 46.3% EPS growth rate (FY2026 $1.589 → FY2028 $3.401); 0.75× that is 34.7%. The Information-Technology secular cap is 15%, but the framework permits up to 20% for a proven, durable >20% grower if flagged explicitly. We flag it and take the full 20%. Twelve consecutive quarters of accelerating growth, +92.8% year-on-year at $6bn scale and a Q2 FY2026 Rule of 40 of 155 qualify on any honest reading. This is the most generous input the framework allows. |
| Terminal growth g_term | 3.0% | Capped at long-run US nominal GDP. Palantir reports and trades in USD, so the currency frame is internally consistent and no foreign-currency adjustment applies. |
| Two-stage warranted P/E (raw) | 33.72× | Σt=1..5((1+g)/(1+r))t + [(1+g)5×(1+g_term)/(r−g_term)]/(1+r)5 |
| Warranted P/E (after the guardrail cap) | 33.0× | Capped at the Information-Technology guardrail line. The maths can never bless infinity. |
| Actual clean P/E | 172.2× | $174.04 ÷ clean diluted EPS of $1.0105 (step 7b, below). |
| Warranted ratio | 5.22 | 172.2 ÷ 33.0. Anything at or above 1.40 is the Expensive band. |
| Band | EXPENSIVE | Doubly so: the ratio is 5.22 against a 1.40 threshold, and the actual 172.2× is above the 33.0× guardrail line on its own, which makes the band Expensive regardless of the ratio and with no growth exception. |
What this means in plain terms. A company that can compound earnings at 20% for five years and 3% thereafter, discounted at 9.13%, is worth about 33 times its earnings. Palantir trades at 172 times. The gap is not a rounding difference or a sector quibble — it is a factor of five. Every extra point on the 10-year Treasury shrinks that warranted number further, and the macro report has cuts priced out with a live hike debate.
Sensitivity, so the reader can see the verdict is not knife-edge. Four separate ways of being generous, none of which changes the answer:
There is no combination of permitted inputs that gets this name out of the Expensive band, which is why the Valuation Ceiling gate is not a close call.
And a note on what was not done. The anchor was computed, not withdrawn. A clean primary multiple resolves for this name — Palantir is
solidly profitable, so P/E is the sector's primary multiple and it is recorded as val_multiple_basis: "clean P/E". Withdrawing the anchor on a name where a multiple resolves,
or quietly swapping to a more comfortable basis, would silence the guardrail test entirely. We use the sector's primary multiple on the step-7b clean number, exactly as the framework specifies.
| Line | TTM (Q3 FY2025 – Q2 FY2026) | Note |
|---|---|---|
| Revenue | $6,155.9m | +78.9% on the prior TTM of $3,440.6m |
| Operating income | $2,634.7m | 42.8% GAAP operating margin |
| Total other income, net | $437.2m | of which $266.4m is interest income on the $9.41bn cash pile; the balance is investment and FX items |
| Pre-tax income | $3,071.8m | effective tax rate just 1.43% ($44.0m) — loss carry-forwards and SBC deductions |
| Net income | $3,016.7m | 49.0% net margin |
| Non-operating share of net income | 14.5% | Just under the 15% threshold — and the cleanest kind: real interest on real cash, not a mark-to-market write-up of a private AI stake. We normalise anyway, because this step runs every time. |
| Clean earnings (operating income after tax) | $2,596.9m | $2,634.7m × (1 − 1.43%) |
| Clean diluted EPS | $1.0105 | on 2,569.8m diluted shares (Q2 FY2026 weighted average) |
| Clean P/E (the scored multiple) | 172.2× | against a reported trailing diluted P/E of 147.5× ($174.04 ÷ $1.18) |
| Clean PEG | 3.72 | 172.2 ÷ 46.3% consensus EPS growth. Above 2 is expensive; above 3 is very expensive. |
| Multiple | Palantir | Software peer median | Own 2-yr decile | Read |
|---|---|---|---|---|
| Clean P/E (primary) | 172.2× | ~26–28× | 6th | Roughly six times the peer median |
| Reported trailing P/E | 147.5× | ~26–28× | 6th | Filing-derived (sum of four reported diluted EPS). Yahoo shows 148.75× on EPS $1.17 — reconciled in §15 |
| Forward P/E (FY2026 consensus EPS $1.589) | 109.5× | ~22–24× | — | You must reach FY2028 consensus EPS of $3.401 to bring the multiple down to 51× |
| EV / Revenue (TTM) | 66.5× | ~7–8× | 6th–7th | The guardrail equivalent for a pre-profit IT name is 20×. Palantir is profitable so P/E governs — but 66.5× sales is worth seeing. |
| Price / Sales (TTM) | 67.9× | ~6–7× | 6th–7th | The trap this framework exists to defeat. Against its own November-2025 peak of roughly 119×, today's 67.9× looks mid-range (57% of it). Note the correction: FMP's priceToSalesRatioTTM of 64.9× is computed off its stale $399.6bn market cap; on the reconciled $418.2bn it is 67.9×. The tell was arithmetic — a company with $9.4bn of net cash cannot have EV/Revenue above its P/S. Relative to its own bubble it is unremarkable. That is precisely why the anchor is absolute. |
| Price / Book | 42.8× | ~5–8× | — | FMP's own rating model scores this 1 out of 5. |
TTM free cash flow of $3.36bn against an enterprise value of $409.1bn = 0.82%. The framework's own scale reads "below 1% — very expensive". For context, the 10-year Treasury pays 4.63% risk-free. You are accepting a 0.82% cash yield, with no dividend, in exchange for growth. That trade can absolutely be worth making — but it should be made with the number in view, not around it.
At $174.04 the market is paying 5.22× the multiple that 20% growth for five years justifies. Solving the other way: to warrant 172× at a 9.13% discount rate you need something close to 45–50% earnings growth sustained for a decade, then a graceful glide to 3%. Consensus — itself optimistic, and revised sharply upward after the Q2 FY2026 print — has EPS compounding at 46.3% to FY2028, after which coverage thins sharply: FY2029 carries 13 revenue analysts but only 3 on EPS, and FY2030 just 8 and 3. So the price is roughly consistent with the most bullish credible path, extended about twice as far as anyone is actually forecasting. There is no margin for error priced in at all; the entire consensus outcome is already in the share price.
Net framing, and the discipline that matters: the anchor says the core business justifies roughly $33 of today's $174.04 share price on trailing earnings, or about $112 if you generously credit consensus FY2028 earnings at the warranted multiple with no discounting at all. With the Nvidia partnership removed from the counted set, the remaining optionality — the idle cash, the un-inflected international book, and the manufacturing and space adjacency — is worth a +3-point tilt to the score and no more (reduced from +4 in an earlier draft precisely because its largest component turned out to be already public). Optionality is a reason to keep watching an expensive business; it is never a reason to call it cheap.
| Source | Consensus / mean | Median | High | Low | n | vs $174.04 |
|---|---|---|---|---|---|---|
FMP get_price_target_consensus | $176.33 | $200 | $215 | $80 | 26 | +1.3% to consensus |
Yahoo get_yahoo_analyst_targets (cross-check) | $191.68 | $204 | $255 | $80 | 27 | +10.1% to mean |
Last 30 days only (get_price_target_summary) | $175 average across 9 fresh targets | 9 | +0.6% | |||
This is the single biggest change in the Valuation pillar and the reason the score fell from 29 to 20. At the last report the consensus target implied roughly 40% upside, which carried the analyst lens near the top of its range. The 30% move has eaten all of it. More tellingly: the nine targets published since the blow-out Q2 FY2026 print average $175 — the Street re-rated its targets up to the price, not above it. The dispersion is extreme (a $255 high against an $80 low, a 3.2× spread), which is itself informative — the analyst community has no consensus on what this company is worth — and we take a 10-point confidence deduction for it. The consensus endpoint was checked for degeneracy and is genuinely dispersed, so no fallback was required; Yahoo was pulled as a cross-check regardless.
13 Buy · 10 Hold · 3 Sell across 26 covering firms — 50% bullish, which the framework reads as "Buy consensus with more than 30% holds: mixed". Score 52. Recent actions: one upgrade (Deutsche Bank, Hold → Buy, 4 August 2026) and no downgrades in 30 days, against seven maintains on the print date. Note that three firms carry an outright Sell — unusual for a company that just grew 93% — and that is a fair summary of the split: nobody disputes the business, everybody argues about the price.
Overall B+ (3/5). The sub-scores are the whole story, and they independently corroborate this report: return on equity 5/5, return on assets 5/5, debt-to-equity 4/5 — against price-to-earnings 1/5, price-to-book 1/5, discounted cash flow 2/5. An entirely separate model, built on different weights, lands where we land: an exceptional business at a price it cannot defend.
Valuation = 20 — anchor 6 (40%) · sector median 2 (20%) · own-history decile 30 (15%) · PEG 20 (10%) · analyst consensus 52 (15%), weighting to 17.1, plus the +3 optionality tilt = 20.1, rounded to 20. Confidence 78%. The score sits inside the Expensive band's 0–39 range, as it must.
The driver. Palantir's fortunes sit above its own execution in one force: enterprise and government adoption of applied AI — commercial AIP deployment on one side, US and allied defence and intelligence software budgets on the other. This is not a commodity price, so the Step-2b commodity price-trend overlay does not apply and is recorded as not applicable. The driver is read instead off adoption evidence, booking data and budget direction.
| Horizon | Weight | Evidence (with periods and dates) | Score |
|---|---|---|---|
| Historical (past 12–24 months) | 25% | Twelve consecutive quarters of accelerating revenue growth. Revenue compounded at roughly 30.5% a year over 2021–2025 and ran at +78.9% on a TTM basis to Q2 FY2026. In Q2 FY2026 US government revenue grew 90% year-on-year to $809m and US commercial 149% to $764m, with total US revenue up 115% to $1.57bn. The adoption curve has bent upward, not flattened. | 92 |
| Current state | 50% | Q2 FY2026 (reported after the close on 3 August 2026): net dollar retention 157%, up 700 basis points in a quarter. Commercial total contract value bookings of $2.337bn, +118%; government TCV bookings $3.4bn, +49% on a dollar-weighted duration basis. FY2026 guidance raised to $8.15–8.16bn from $7.65–7.66bn, roughly +82% year-on-year, with FY2026 US commercial revenue guided above $3.42bn (+134%). Nvidia sovereign-AI partnership announced 29 June 2026 — note that this predates the prior report, so it is a standing asset rather than a new development this quarter, and the 7 August share-price leg was a Bank of America target raise plus an agentic-AI cohort rally, not this deal. Against all that: the "AI eats software" fear that de-rated the whole group in the first half of 2026 has been answered for now, not permanently. | 88 |
| Forward outlook (6–12 months) | 25% | Consensus has revenue growing 49% in FY2027 and 48% in FY2028 (23 and 17 contributing revenue analysts respectively — genuine coverage depth, unlike the FY2029/FY2030 tail). Enterprise AI is the one expanding budget line in an otherwise tightening corporate spending environment. Risks: budget digestion after a land-grab year, sovereign-AI competition, and a US economy that is now visibly deteriorating — July payrolls at −23k and the 14 August consumer prints (below) — which eventually reaches software budgets. | 78 |
Two corroborating prints landed on 14 August, after the macro report was written: July retail sales −0.6% month-on-month to $763.6bn against a consensus of roughly +0.1% to +0.3% (still +5.0% year-on-year), and Michigan consumer sentiment at 51.0 in early August against 55.2 in July and a 54.5 consensus, with expectations falling from 55.4 to 50.6. The survey attributes the drop to cost-of-living fears driven by the Middle East conflict — the same energy shock as the Hormuz closure, transmitting to the household.
Which way it cuts for Palantir, stated plainly rather than waved through. Palantir has essentially no US consumer exposure: its Q2 FY2026 revenue is US government ($809m), US commercial enterprise ($764m) and international. The retail-sales category detail that matters most elsewhere — nonstore/online at −2.2%, motor vehicles at −1.8% — has no read-through here at all. We are explicitly not marking the driver down for it.
There are two genuine second-order effects, pointing in opposite directions. Mildly helpful: a visibly weakening economy trims the odds of the September rate hike the market puts near 44%, and since the whole quarrel with this stock is the discount rate applied to distant cash flows, anything that lowers that rate lifts the warranted multiple. Mildly harmful: a contracting economy is eventually where enterprise software budgets get cut, and 93% growth is priced as though they never will be. Neither is a one-month story, and neither moves the driver score. The more relevant macro datapoint for Palantir is the labour print — payrolls at −23k against +80k expected — because corporate headcount contraction reaches IT budgets before consumer sentiment does.
Driver score = 86 (92×0.25 + 88×0.50 + 78×0.25) — "Strong Tailwind", up from 81. Driver confidence 74% (base 70, less 10 for the contested nature of the AI-adoption debate, plus 14 for the exceptional freshness and specificity of the Q2 FY2026 evidence).
At 86 the driver is comfortably above the ≥65 threshold, so it is eligible to lift a base BUY to STRONG BUY. It lifts nothing, for three independent reasons, any one of which would suffice:
The driver does not alter the three fundamental pillar scores. It is context — and the context is genuinely excellent, which is exactly the tension this report is about: the best driver on the watchlist attached to the worst valuation on the watchlist. A strong tailwind does not make an expensive stock cheap. It explains why it got expensive.
Thesis-invalidation floor. The level at which the whole case breaks is not a price — it is a growth rate. If quarterly revenue growth decelerates below roughly 40% year-on-year while the multiple is still above 100×, the arithmetic that justifies any part of this valuation stops working, and the de-rating is not a wobble but a re-pricing. Q2 FY2026 was +92.8%. That is a long way from the floor, and it is the strongest argument the bulls have.
The 12 August 2026 macro report reads the regime as "energy-shock stagflation: a supply-driven inflation impulse hitting a contracting labour market, with Fed cuts priced OUT and the live debate hike-vs-hold." Palantir does not appear on that report's Economic Watchlist Forecast, so we map its GICS sector to the Driver-Sector Impact Matrix: XLK reads Neutral (short) / Neutral (medium) / Outperform (long).
Pressure = Neutral, anchored on the medium horizon as the framework requires. Short is also Neutral; long is a Tailwind. Because pressure is Neutral rather than Tailwind, no amplification is available on any horizon — and the base signal is HOLD in any case, which never amplifies.
Stance = Neutral, conviction 50, changed from Contrarian/48 last report. Two things moved: XLK went from N/O/O to N/N/O, and Palantir's own tape flipped from underperforming both benchmarks to leading them by 27 points. "Contrarian" describes fading an economic headwind, and that is no longer what a long position here would be doing.
The nuance the sector aggregate hides, stated plainly. XLK's Neutral is the net of two opposing forces: a powerful AI runway against a rate path that has removed every cut. For a business at 172× earnings the rate leg is squarely a headwind — the sector average nets it away, but this name cannot. The macro report's read on ServiceNow, the closest long-duration software comparator on its watchlist, makes the point better than we can: it calls it "the clearest casualty of this run's Fed correction — a long-duration multiple with the 10-year against it", downgraded on the rate path rather than the business, with the enterprise-AI runway holding the long horizon positive. That is Palantir's profile exactly, only more extreme. (The macro report quotes 4.70% for the 10-year in that passage; the anchor in §4 sources 4.63% directly from FRED series DGS10 at 13 August 2026. The 7bp gap changes nothing — see the sensitivity block in §4.)
Post-dating the macro report: the 14 August 2026 US consumer prints (retail sales −0.6% month-on-month, Michigan sentiment 51.0) landed after it was written. They do not change the sector read for a name with no consumer exposure — see §5 for the full treatment and which way each leg cuts.
The macro report is four days old — fresh, so no staleness haircut. Confidence 68%, reduced from the base because the read is sector-mapped rather than name-specific.
Source: sector-map (XLK — Palantir is not on the macro report's Economic Watchlist Forecast) · Macro report 2026-08-12
| Measure | Value | Read |
|---|---|---|
| Price (14 Aug 2026 close) | $174.04 | Down 2.8% on that session from $179.01 on 13 August |
| Nearest structural support | $152.70 | A three-way cluster: the 6 August 2026 swing low ($152.70), the 20-day EMA ($152.90) and the 200-day SMA ($152.09). A strong zone. |
| Logical stop | $150.00 | $24.04 below spot = 2.86 ATR (ATR $8.41) |
| Distance to support in ATR | 2.54 ATR | Above the 2.5-ATR line — an unfavourable entry location requiring a wide stop |
| Nearest resistance | $179.60 / $180.18 | Only 3.2% above, and price has now been rejected there on two consecutive sessions (13 and 14 August 2026) |
| Reward to the base target of $195 | +12.0% | Ratio to the stop: 0.87 : 1 |
| Reward to the bull target of $240 | +37.9% | Ratio to the stop: 2.75 : 1 |
| Time since the swing low | 9 sessions | The move began at $125.65 on 3 August 2026 and is +38.5% off it. Entering here is chasing, not anticipating. |
| 52-week range position | 66.9% | Upper-middle. High $207.52, low $106.37. |
Risk-reward sub-score: 36. The setup is not poor because the trend is poor — the trend is excellent. It is poor because the location is: you would be buying 38.5% above the trigger, 3.2% below a level that has just rejected twice, with the nearest real support 12.3% away.
| Window (to 14 Aug 2026) | Palantir | SPY | XLK | vs SPY | vs sector |
|---|---|---|---|---|---|
| 1 month | +30.2% | +3.3% | +3.5% | +26.9pp | +26.7pp |
| 3 months | +30.1% | +3.8% | +5.9% | +26.4pp | +24.3pp |
Outperforming both benchmarks over both windows — the top band, score 92. The prior report recorded −12.0 and −9.0 points of underperformance; this is a complete flip in four weeks, and it is the single largest input to the timing score's 13-point rise.
One piece of context that cuts against the "euphoric top" reading, and belongs here. Despite a 30% month, Palantir is roughly flat for calendar 2026 and still 16.1% below its November-2025 high of $207.52. The August move is the recovery of a first-half drawdown, not a breakout to new highs. A reader entitled to expect this report to argue only one side should note that this fact argues against it.
| Sub-signal | Reading | Score |
|---|---|---|
| Fed direction | Cuts priced out; the 2-year at 4.22% against a 3.63% funds rate; roughly 44% market-implied odds of a September hike; three hawkish dissents in July. For a stock whose entire value sits in distant cash flows, this is the most damaging macro variable there is. | 20 |
| Volatility (VIX) | 14.25 — below 15, firmly risk-on. 52-week range 13.38–35.30. | 85 |
| Yield curve | 10-year 4.63% against a 2-year 4.22% — positively sloped by 41 basis points, normal. | 75 |
| Sector regime | XLK reads Neutral (short), Neutral (medium), Outperform (long) in the 12 August 2026 macro report. | 50 |
Macro sub-score 57, applied at the 15% dynamic weight for a Medium-macro-sensitivity name. We deliberately classify Palantir as Medium rather than the "defensive software" Low: a 172× multiple is a long-duration instrument, and the discount rate matters far more to it than to a typical SaaS business.
| Signal | Reading (recency-weighted; every data point inside 11 days, 1.0×) | Score |
|---|---|---|
| Analyst grade actions | One upgrade (Deutsche Bank, Hold → Buy, 4 August 2026), zero downgrades in 30 days, seven maintains on the print date (Rosenblatt, UBS, Mizuho, Citigroup, DA Davidson, Piper Sandler, Cantor Fitzgerald). Net positive. | 70 |
| Estimate revisions | Sharply up. FY2026 revenue guidance raised by roughly $0.5bn; FY2026 consensus EPS $1.589 and FY2027 $2.263, both revised higher after the print. | 80 |
| News tone | Of 20 recent articles, roughly nine positive, six neutral and two negative on Palantir. The pattern is remarkably consistent: positive on the results, negative-to-neutral on the price — headlines run from "otherworldly earnings" through "consider this valuation risk" to "these two legacy defence stocks are the cheaper way to play the same budget". | 68 |
| Insider activity | Zero insider purchases across the AI cohort over the trailing year against $3.4bn of sales; ARK selling into the rally. Normal for Palantir's 10b5-1 pattern, but one-directional. | 35 |
| Options skew | Not retrieved this run — recorded as a data gap in §15, with a 5-point timing-confidence deduction. | — |
Sentiment sub-score 63.
| Date | Catalyst | Impact |
|---|---|---|
| 3 Aug 2026 — done | Q2 FY2026 results (after the close) | Resolved, decisively positive. +29.45% the following session. |
| 29 Jun / 1 Jul 2026 — done | Nvidia sovereign-AI partnership announced | Resolved, positive — +7.8% to $125.73 on 1 July 2026. Predates the prior report, so already priced. |
| 7 Aug 2026 — done | Bank of America reaffirms Buy, target to $255; broad agentic-AI rally | Resolved, positive. +10.32% to $172.01 on 77.6m shares (77,579,878 per get_technical_indicators, this report's dating and volume authority — roughly 1.8× the ~43.2m average; syndicated coverage rounded it to 76.2m).. |
| 19 Aug 2026 | FOMC minutes (July meeting) | High — the most relevant event in the window for this name, because it moves the discount rate the whole multiple rests on. |
| 26 Aug 2026 | Core PCE price index (July) | High — sets the September hike path. |
| ~early Nov 2026 | Q3 FY2026 results | High, but outside every relevant window. |
Catalyst clustering score 75 (calm) — no company-specific catalyst within 30 days, with the two macro releases above as the only path risk. No position-size reduction is indicated on catalyst density.
Timing = 61 — multi-timeframe trend 70 (30%) · risk-reward 36 (20%) · macro 57 (15%) · sentiment 63 (18%) · catalysts 75 (17%). Confidence 70% (base 75, less 5 for the missing options-skew read; no earnings inside 14 days, VIX benign, calendar calm). The tape has genuinely turned — and it does not change the signal, because the Decision Matrix routes an Expensive valuation to HOLD at any timing score whatsoever.
| Date | Event | Impact | Forecast | Previous | Relevant? | Why |
|---|---|---|---|---|---|---|
| 18 Aug 2026 | Housing Starts & Building Permits (Jul) | High | 1.35M / 1.37M | 1.427M / 1.374M | No | Housing has no read-through to enterprise software demand. |
| 19 Aug 2026 | FOMC Minutes (July meeting) | High | — | — | ✅ YES | The most important event in the window for this name. A 172× multiple is a long-duration instrument; the minutes move the discount rate directly. The macro report expects a hawkish read, with a committee debating the conditions for a hike rather than the timing of a cut, and roughly 44% market-implied September hike odds. |
| 25 Aug 2026 | CB Consumer Confidence (Aug) | High | — | 90.8 | ⚠ Medium | A second-order read on the corporate spending cycle that eventually reaches software budgets. Watch whether it corroborates the 14 August Michigan collapse or contradicts it. |
| 26 Aug 2026 | Core PCE Price Index MoM (Jul) | High | +0.3% | +0.1% | ✅ YES | The Fed's preferred inflation gauge. An upside surprise hardens the September hike case, and long-duration multiples compress first. |
| 26 Aug 2026 | Durable Goods, Personal Income & Spending (Jul) | High | +0.2% / +0.2% / +0.3% | +0.3% / +0.2% / +0.3% | ⚠ Medium | Demand-side confirmation, or otherwise, of the labour-market deterioration below. |
| 1–4 Sep 2026 | ISM Manufacturing, ISM Services, Non-Farm Payrolls (Aug) | High | 55.0 / 53.8 / +12k | 55.6 / 54.1 / −23k | ⚠ Medium | Falls after this report's 31 August refresh, but sets the September regime. |
| Date | Event | Actual | Forecast | Surprise | Impact |
|---|---|---|---|---|---|
| 7 Aug 2026 | Non-Farm Payrolls (Jul) | −23k | +80k | −128.8% below | The most relevant recent print for this name. The labour market is contracting outright. Corporate headcount contraction reaches IT budgets before consumer sentiment does — but it also trims the odds of the September hike that would hurt this multiple most. |
| 12 Aug 2026 | CPI YoY / Core CPI YoY (Jul) | 3.4% / 2.5% | 3.4% / 2.5% | In line | No new information for the rate path. Headline remains well above target on energy. |
| 13 Aug 2026 | Producer Price Index MoM (Jul) | 0.0% | +0.2% | Below | Softer than expected — mildly supportive of long-duration equity. |
| 14 Aug 2026 | Retail Sales MoM (Jul) | −0.6% (to $763.6bn; +5.0% YoY) | ~+0.1% to +0.3% | Clear miss | Nonstore/online −2.2%, motor vehicles −1.8%, gasoline stations −0.9%. Post-dates the 12 Aug macro report. No direct read-through to Palantir — no consumer exposure. |
| 14 Aug 2026 | Michigan Consumer Sentiment (Aug, early) | 51.0 | 54.5 (prior 55.2) | Clear miss | Expectations 55.4 → 50.6; current conditions 54.8 → 51.8. Attributed to cost-of-living fears from the Middle East conflict. Again, no direct Palantir read-through; relevant only as a second-order signal on the corporate spending cycle. |
The week's data is unambiguously demand-negative — payrolls at −23k against +80k expected, retail sales −0.6%, Michigan sentiment at 51.0 — with inflation exactly in line and producer prices soft. The two 14 August consumer prints post-date the macro report, so they are carried here rather than inherited from it.
For Palantir specifically, the honest answer is that most of this does not apply. Its Q2 FY2026 revenue is US government, US commercial enterprise and international — there is no consumer line to mark down, and we have not marked one down. The retail-sales category detail that matters for online retail or discretionary names is simply irrelevant here.
Two second-order effects do apply, and they point in opposite directions. Helpful: a visibly weakening economy trims the roughly 44% odds of a September hike, and since the entire quarrel with this stock is the discount rate applied to distant cash flows, anything that lowers that rate lifts the warranted multiple and narrows the gap this report is built on. Harmful: a contracting economy is where enterprise software budgets eventually get cut, and 93% growth is priced as though they never will be. The company is not yet showing it — bookings and retention accelerated in the same quarter payrolls turned negative — but the two cannot diverge indefinitely.
The one event to watch: the FOMC minutes on 19 August. Palantir is classified Medium macro-sensitivity, so no WAIT-FOR-EVENT override applies — that mechanism is reserved for High-sensitivity sectors — but for a name valued at 172 times earnings, the discount rate is not a background variable.
| Timeframe | Trend | Direction | RSI | MACD | Key S/R | Breakout | Vol |
|---|---|---|---|---|---|---|---|
| Monthly | Uptrend ↑ | Bullish | 63.0 | 22.70, hist −5.00 | S: $106.37 R: $207.52 | Resistance breakout | 0.47×* |
| Weekly | Uptrend ↑ | Bullish | 61.3 | −0.38, hist +4.71 rising | S: $128.75 R: $190.00 | Resistance breakout | 0.91×* |
| Daily | Uptrend ↑ | Bullish | 68.1 | +12.56, hist +4.02 narrowing | S: $152.70 R: $179.60 | Resistance breakout | 0.55× |
| Hourly | Weakening → | Neutral | 41.8 | −0.25, hist −0.62 | S: $168.34 R: $180.18 | None | — |
| 15-minute | Downtrend ↓ | Bearish | 32.6 | −0.55, hist +0.12 | S: $173.69 R: $176.98 | Support breakdown | — |
| Confluence: Mostly Bullish · MTF Score 70 | |||||||
All three of the timeframes that matter are in confirmed uptrends, and the weighted score of 70 puts this in the "Mostly Bullish" band — a long way from the 48 and "Mixed" of the prior report. At the 14 August 2026 close of $174.04 price sits above the 20-day SMA ($145.53), the 50-day ($134.99) and the 200-day ($152.09).
The structural detail worth knowing: the 50-day average, at $134.99, is still below the 200-day at $152.09 — there has been no golden cross yet. The 50-day is rising at roughly $0.45 a session while the 200-day drifts down about $0.05, so the $17.10 gap closes at around $0.50 a session: about 34 trading sessions, or early October, assuming no pullback. That would be the trend-following confirmation this chart currently lacks.
The pattern is textbook "higher-timeframe bullish, lower-timeframe pulling back." Hourly has turned weakening and the 15-minute has broken minor support, with the daily RSI easing from 73.9 on 10 August to 68.1 on 14 August. Normally that combination is the high-probability buy setup — buying the dip in an uptrend. Here the dip has not yet reached anywhere worth buying: the support that matters is the $152 cluster, 12.3% below.
Two honest caveats on this table. The monthly and weekly volume ratios (0.47× and 0.91×) are artefacts of incomplete periods, not distribution. And the monthly MACD histogram is negative (−5.00) even as the monthly trend reads up — a residue of the first-half drawdown working through a slow indicator, not a live bearish divergence. All dates in this section are taken from get_technical_indicators, whose labels were verified exact this run (see §15).
Six months of daily closes to the 14 August 2026 close of $174.04, with the 50-day simple moving average. The vertical event is the Q2 FY2026 print — $125.65 on 3 August to $162.66 on 4 August, a 29.45% gap on 4.0× average volume — followed by a second 10.3% leg on 7 August on a Bank of America target raise to $255 and a broad agentic-AI rally. Note that price sits far above a 50-day average which is itself still below its 200-day, and that the disciplined fair-value line at $112 sits below every close on the chart.
+37.9% over 12 months. Growth holds above 60% into FY2027 as the AIP land-grab converts to expansion; net dollar retention stays near 157%; the Nvidia sovereign-AI partnership (live since 29 June 2026) converts into a named, disclosed government stack win; international commercial finally inflects. At $240 the stock would be on roughly 106× FY2027 consensus EPS of $2.263, or 50× FY2028's $3.401 — that is, the multiple compresses and the shares still rise 38%, because earnings grow faster than the de-rating. Market capitalisation would be about $577bn, above the $500bn line the November-2025 high sat at. Trigger: another quarter above 90% growth, plus the Fed's September decision landing dovish. Why not higher: at $240 the ratio to warranted is 7.2×, and the analyst high is $215 (FMP) / $255 (Yahoo).
+12.0% over 12 months. The multiple stays roughly where it is and earnings compound into it — exactly what happened over the last four weeks, when the price rose 30% and the clean P/E moved only from 169× to 172.2×. $195 is 86× FY2027 consensus EPS, and sits between the FMP median target of $200 and the Yahoo mean of $191.68. Growth decelerates gracefully from 92.8% toward 55–60%; no macro shock; the AI-concentration tail stays armed but never fires. The disclosure that matters: this base case explicitly assumes the multiple does not normalise. That is why the base target of $195 and the disciplined fair value of $112 differ so much — the gap between them is not an inconsistency, it is the assumption, stated.
−45.4% over 12 months, and it has two separate legs.
Leg 1 — the cohort de-rating (the deeper leg, and the one inherited from the macro report). The armed "S&P 500 concentration / AI earnings-quality unwind" tail fires at the index level: a hyperscaler guides capex down more than 20%, a large private AI valuation is marked down, or non-operating gains across the AI Big 10 turn negative. Palantir need do nothing wrong. A compression from 172× clean earnings to roughly 90× — still an enormous premium to the 33.0× warranted multiple, and still far above any peer — is a 48% move on its own. Multiple reversion from this altitude is not a wobble. This leg is carried in full precisely because the trigger was judged not to be live enough to fire a Do-Not-Buy (§2) — the risk does not disappear because the trigger is receding; it moves from the signal into the bear case, which is where the framework puts it. Falsification: breadth continues broadening — the equal-weight S&P catching the cap-weighted index. That is already partly in train (equal-weight +3.6% against +2.8% over the month, with the Nasdaq-100 at +0.6%), which is exactly why the macro report grades the trigger receding.
Leg 2 — the idiosyncratic story (shallower, and independent). Growth decelerates toward 40%; a US government budget or contract disappointment lands; or the competitive vector bites — Microsoft Fabric bundling Foundry-equivalent function into existing Azure enterprise agreements, a named enterprise migrating off Foundry to a Databricks-plus-integrator stack, or Anduril's Lattice taking a contested defence programme. On its own this leg is worth perhaps −25 to −30%.
The $95 target is close to leg 1 alone. If both legs fire the number is lower, and that is the tail this report is genuinely worried about.
Read that number carefully, because it is easy to mistake for an endorsement. A +4.1% expected value over twelve months is less than the 4.63% you are paid to hold a Treasury, and it is earned across a distribution running from +38% to −45%. The framework's objection is not that Palantir will fall. It is that you are being asked to accept an 83-point spread of outcomes, on a multiple 5.22 times what the fundamentals warrant, for an expected return below the risk-free rate. That is what "Hold" means here.
Forecast:
These are not in conflict, but the combination is unusual enough to state outright. The conviction ladder sets how much to buy given that you are permitted to buy. The Valuation Ceiling gate caps the signal at Hold, which withholds that permission. One valid entry path is open, so the ladder correctly reads Half-Size; Gate 3 then caps the signal regardless. The ladder is shown because it is the honest mechanical output, and because it tells you what changes if the valuation ever clears — not because it is a recommendation to take a half position today.
One question this ought to answer before an alert reader asks it. Timing scores 61, which is 'Improving', and the framework does contain a narrow sanctioned exception — the half-size quality-starter — that turns a Short HOLD into a half-size BUY when the HOLD is produced only by neutral timing on a high-quality name. It does not fire here, and it cannot. That override requires Valuation of at least 40 and explicitly excludes the Expensive band; Palantir's Valuation is 20. It also requires both the medium and long horizons to be BUY-or-better, and both are HOLD. The Short is not being held back by a merely lukewarm tape — the tape is the best thing about this name right now. It is being held back by the price, which is what short_hold_reason: "expensive" records in the calibration.
And the Catalyst group deserves a caveat of its own. It is scored MET on an event-anchored reading: the group is literally "an event confirms it", and the 4 August 2026 print satisfied all three sub-conditions on a single day. But the event was eight sessions before the 14 August close, the trigger fired at $162.66, and $174.04 is 7.0% above the trigger and 38.5% above the pre-print low of $125.65. Evaluated on the latest session rather than the event's, the volume sub-condition fails outright — 24.0m shares against a roughly 43.5m average is 0.55×, not 2×. We keep the group MET because the plain reading of the rule is event-anchored, but a reader should treat this as a path that opened and has largely been travelled, not one standing open at today's price.
Forecast:
Imagine you open the position at $174.04 right now.
What you are risking. The hard stop at $150.00 is −$24.04, or −13.8%, and it has to be that wide because the nearest real support is the $152–153 cluster 12.3% below — 2.54 ATR away. The bear case is $95, or −45.4%, and its larger leg is a cohort-wide de-rating that has nothing to do with Palantir executing badly. Two of three entry paths are shut: you would be paying 55% above the disciplined fair value of $112 (Fundamental unmet), with RSI at 68.1 outside the permitted band (Technical unmet). The one path that is open — Catalyst — opened at $162.66 on 4 August 2026, so you would be buying 7.0% above the trigger and 38.5% above the pre-print low. Immediate resistance at $179.60 is 3.2% away and has rejected twice. Path risk: the FOMC minutes on 19 August and Core PCE on 26 August, both of which move the discount rate this valuation depends on.
What you are gaining. The base case is $195, +12.0%, and the bull case $240, +37.9% — you start capturing both immediately. You own a business that grew revenue 92.8% in Q2 FY2026 with 157% net dollar retention, a record Rule of 40 of 155, $9.41bn of cash and no borrowings. You own the embedded optionality: the un-inflected international commercial book, the manufacturing and space adjacency, and $9.41bn of idle cash. While you wait you collect a 0.82% free-cash-flow yield and no dividend. Risk-reward is 0.87 : 1 to the base target and 2.75 : 1 to the bull.
The read. Acting now is a poor deal on the framework's own arithmetic — a probability-weighted +4.1% against a risk-free 4.63%, with an 83-point spread of outcomes. Waiting for the $152–153 cluster would cut the stop distance from 2.86 ATR to about 0.4 ATR and turn the base-case ratio from 0.87 : 1 to roughly 5 : 1 — the same thesis, at a materially better price, with a defined risk. That is an assessment of the entry, not a forecast that the pullback will come.
Imagine you close the position, or stay out, at $174.04 right now.
What you are giving up. The base case at $195 (+12.0%) and the bull at $240 (+37.9%). More importantly you give up the mechanism that produced the last four weeks: earnings growing into the multiple. The clean P/E moved only from 169× to 172.2× while the stock rose 30% — the price re-rated and the valuation did not, because the business genuinely ran that fast. At 157% net dollar retention and 92.8% growth, that mechanism is live. There is no income to forgo — no dividend, and a 0.82% cash yield.
What you are protecting. The 45.4% path to $95, of which the deeper leg is an index-level AI de-rating rather than a company failure. At 5.22× the warranted multiple there is a great deal of altitude to lose. But be precise about the mechanics: no exit trigger is live right now. The $150 stop is 13.8% below; not one thesis-invalidation condition holds (guidance was raised, growth accelerated, the driver strengthened, no competitor has taken a named account); and the profit-take needs both $200 and RSI above 70, and RSI is 68.1. All three groups read clear.
The read. There is no mechanical reason to sell today. If you already hold Palantir, nothing in this report says get out — the exit ladder reads Hold, and it means it. If you do not hold it, the Valuation Ceiling says this is not the price at which to start. Those are not contradictory positions: the gate governs new capital, the exit ladder governs capital already committed. A holder is being paid to run a risk they already took; a buyer would be paying full price to take it fresh.
Position sizing not computed — no portfolio allocation or role was specified for this analysis, and the framework does not invent defaults. It is largely moot in any case: the conviction ladder reads Half-Size, but the Valuation Ceiling gate caps the signal at Hold, so there is no sanctioned entry to size. Specify an allocation and a role if you want sizing guidance on a future refresh.
| Measure | Palantir | Reference |
|---|---|---|
| Daily ATR | $8.41 = 4.8% of price | A typical large-cap runs 1–2%. Expect roughly $8 of daily range. |
| Beta vs SPY | 1.563 | A 5% position carries the risk of a 7.8% position in an index fund. |
| Max drawdown, past 12 months | −48.7% ($207.52 in November 2025 to $106.37 in June 2026) | SPY's worst 12-month drawdown was a small fraction of that. |
| Single-session moves in the last 10 sessions | +29.45%, +10.3%, −2.8% | This is not a stock that moves in orderly increments. |
The −48.7% drawdown happened while the company was compounding revenue at 60–90% and beating every quarter. That is the honest illustration of what a 172× multiple does: it can halve without the business doing anything wrong.
{
"ticker": "PLTR",
"exchange": "NASDAQ",
"exchange_ticker": "NASDAQ:PLTR",
"isin": "US69608A1088",
"api_ticker": "PLTR",
"storage": "PLTR",
"finder_ticker": "PLTR",
"finder_exchange": "\ud83c\uddfa\ud83c\uddf8 NASDAQ",
"company": "Palantir Technologies Inc.",
"brand": "Palantir",
"currency": "USD",
"date": "2026-08-16",
"version": "v6",
"section": "Technology (Contrarian)",
"sector": "Information Technology",
"gics_sector": "Information Technology",
"sub_industry": "Software - Infrastructure",
"lifecycle_stage": "High-Growth (Stage 2-3)",
"country_table": "US",
"user_context": {
"horizon": null,
"allocation_pct": null,
"portfolio_role": null
},
"user_horizon": null,
"user_allocation_pct": null,
"portfolio_role": null,
"price_at_rating": 174.04,
"price_asof": "2026-08-14",
"price_asof_note": "Report dated 2026-08-16 (Sunday). Markets were closed 15-16 Aug, so the Friday 14 Aug 2026 close is the latest print and is used throughout.",
"signals": {
"short": "HOLD",
"medium": "HOLD",
"long": "HOLD"
},
"signal_short": "HOLD",
"signal_medium": "HOLD",
"signal_long": "HOLD",
"primary_signal": "HOLD",
"base_signals": {
"short": "HOLD",
"medium": "HOLD",
"long": "HOLD"
},
"composite_short": 56,
"composite_medium": 55,
"composite_long": 62,
"composite_formula": "short = T*0.55 + V*0.25 + Q*0.20 = 61*0.55 + 20*0.25 + 86*0.20 = 55.75 -> 56; medium = Q*0.35 + V*0.35 + T*0.30 = 86*0.35 + 20*0.35 + 61*0.30 = 55.40 -> 55; long = Q*0.55 + V*0.30 + T*0.15 = 86*0.55 + 20*0.30 + 61*0.15 = 62.45 -> 62. Gate 3 caps the SIGNAL, not the composite.",
"quality_score": 86,
"valuation_score": 20,
"timing_score": 61,
"driver_score": 86,
"moat_score": 73,
"quality_detail": {
"industry_benchmark_name": "Rule of 40 (SaaS) - Q2 FY2026 construction",
"industry_benchmark_value": 155,
"industry_benchmark_value_ttm_crosscheck": 133,
"industry_benchmark_score": 99,
"moat_score": 73,
"moat_pricing_power": 85,
"moat_network_effects": 48,
"moat_switching_costs": 82,
"moat_cost_advantage": 62,
"moat_intangibles": 88,
"roic_percentile_vs_peers": 92,
"roic_pct": 25.7,
"roic_denominator_note": "NOPAT $2,596.9m / (equity $9.885bn + lease-inclusive debt $0.211bn = $10.096bn) = 25.7%. Borrowings-only denominator gives 26.3%; the conservative figure is scored.",
"capital_allocation": 74,
"management_skin_in_game": 58,
"revenue_per_employee_usd": 1400000,
"net_dollar_retention_q2fy26": 1.57,
"subsignal_block_avg": 94.0,
"quality_build": "subsignal block 94.0 (simple average of 98/95/96/98/97/80/94) at 35% + Rule-of-40 benchmark 99 at 18% + moat 73 at 25% + ROIC/capital 76 at 22% = 85.69 -> 86"
},
"valuation_detail": {
"ps_ttm": 67.9,
"ev_rev_ttm": 66.5,
"fwd_pe_fy2026": 109.5,
"fwd_pe_fy2027": 76.9,
"pe_ttm_reported": 147.5,
"pb": 42.8,
"fcf_yield": 0.0082,
"fcf_ttm_usd": 3359000000,
"implied_growth_rate": "~45-50% EPS growth sustained for a decade is required to warrant 172x at r=9.13%; our disciplined estimate is the flagged 20% ceiling",
"consensus_eps_cagr_fy26_fy28": 0.463,
"historical_valuation_decile": 6,
"anchor_subscore": 6,
"sector_median_subscore": 2,
"own_history_subscore": 30,
"peg_subscore": 20,
"analyst_consensus_subscore": 52,
"optionality_tilt": 3,
"ps_ttm_note": "67.9x = market cap $418.228bn / TTM revenue $6,155.9m. FMP priceToSalesRatioTTM of 64.914 is computed off its STALE $399.646bn market cap and was used in an earlier draft; the tell was that it printed BELOW EV/Rev of 66.5x for a company with $9.4bn net cash, which is impossible.",
"optionality_note": "Reduced from +4 to +3. The Nvidia sovereign-AI partnership was removed from the counted set: it was announced 2026-06-29 (stock +7.8% to $125.73 on 2026-07-01, ~$21.7bn of market value), which PREDATES the 2026-07-16 prior report, so it is neither new nor un-priced. The remaining tilt rests on idle net cash, the un-inflected international commercial book, and the manufacturing/space adjacency."
},
"nonop_pct_of_net_income": 14.5,
"clean_pe": 172.2,
"clean_peg": 3.72,
"reported_pe": 147.5,
"eps_trailing": 1.18,
"eps_trailing_basis": "sum of four reported diluted quarterly EPS (Q3 FY2025 0.19 + Q4 FY2025 0.24 + Q1 FY2026 0.34 + Q2 FY2026 0.41); reconciles to FMP priceToEarningsDilutedRatioTTM 147.49",
"eps_clean_diluted": 1.0105,
"trailing_pe": 147.5,
"sbc_pct_rev_q2fy26": 0.137,
"sbc_pct_rev_h1fy26": 0.131,
"sbc_usd_q2fy26": 265209000,
"net_margin_ttm": 0.49,
"operating_margin_ttm": 0.428,
"operating_margin_adj_q2fy26": 0.62,
"gross_margin_ttm": 0.848,
"rule_of_40": 155,
"ttm_revenue": 6155941000,
"ttm_revenue_growth_yoy": 0.789,
"q2fy26_revenue": 1935464000,
"q2fy26_revenue_growth_yoy": 0.928,
"fy2026_rev_guide_low": 8150000000,
"fy2026_rev_guide_high": 8160000000,
"fy2026_rev_growth_guide": 0.82,
"cash_usd": 9409098752,
"total_debt_usd": 211400000,
"debt_definition_note": "Provider totalDebt of $211.4m is LEASE-INCLUSIVE. Interest expense is $0 in each of the last 8 quarters, confirming no borrowings. Lease-inclusive figures used throughout for conservatism.",
"warranted_multiple": 33.0,
"warranted_multiple_raw": 33.72,
"actual_multiple": 172.2,
"warranted_ratio": 5.22,
"val_multiple_basis": "clean P/E",
"discount_rate_r": 0.0913,
"risk_free_10y": 0.0463,
"risk_free_10y_date": "2026-08-13",
"risk_free_source": "FRED DGS10 (direct); the 2026-08-12 macro report quotes 4.70% - sensitivity tested, band unchanged",
"equity_risk_premium": 0.045,
"risk_addon": 0.0,
"g_near": 0.2,
"g_near_note": "FLAGGED EXCEPTION - the framework's 'proven, durable >20% grower' ceiling of 20% is used instead of the 15% Information Technology secular cap. Justification: 12 consecutive quarters of accelerating growth, +92.8% YoY in Q2 FY2026 at a $6bn revenue base, Rule of 40 of 155, NRR 157%. 0.75 x consensus would have been 34.7%. This is the most generous input the framework permits and the Expensive verdict survives it.",
"g_term": 0.03,
"val_band": "expensive",
"sector_guardrail_line": 33.0,
"guardrail_breach": true,
"anchor_withdrawn": false,
"anchor_note": "Anchor COMPUTED, not withdrawn. A clean primary multiple resolves (the company is solidly profitable), so P/E is the sector's primary multiple and is used on the step-7b clean number. Sensitivity: g_near 15% -> warranted 27.7x, ratio 6.22 (worse); reported P/E 147.5x -> ratio 4.47 (still Expensive); r at the macro report's 4.70% -> raw warranted 33.3x, still capped at 33.0x, band unchanged.",
"currency_frame": "USD throughout - Palantir reports and trades in USD, so r, ERP and g_term are all inside one currency frame. No FX rate used.",
"timing_detail": {
"mtf_confluence": 70,
"risk_reward_score": 36,
"relative_strength_vs_spy": 26.9,
"relative_strength_vs_sector": 26.7,
"rs_window": "1 month to the 2026-08-14 close: PLTR +30.2% vs SPY +3.3% vs XLK +3.5%; 3 month: PLTR +30.1% vs SPY +3.8% vs XLK +5.9%",
"catalyst_clustering_score": 75,
"macro_subscore": 57,
"sentiment_subscore": 63,
"dynamic_macro_weight": 0.15,
"daily_rsi": 68.1,
"daily_macd_hist": 4.02,
"weekly_trend": "uptrend (price 174.04 above the 50-week SMA 155.07)",
"monthly_trend": "uptrend",
"daily_trend": "uptrend - above SMA20 145.53, SMA50 134.99 and SMA200 152.09; SMA50 still BELOW SMA200, no golden cross yet (~early Oct 2026 at current convergence)",
"above_sma200_pct": 14.4,
"off_52wk_high_pct": -16.1,
"range_position_52wk_pct": 66.9,
"vix": 14.25
},
"competitive_share_trajectory": "gaining",
"competitive_threat_level": "elevated",
"competitors_named": [
"Databricks",
"Snowflake",
"Microsoft Fabric / Azure",
"in-house build on hyperscaler tooling",
"Anduril (Lattice)",
"Lockheed Martin / General Dynamics / Booz Allen / SAIC / CACI"
],
"economic_alignment_stance": "Neutral",
"economic_alignment_conviction": 50,
"economic_alignment_pressure": "Neutral",
"economic_alignment_pressure_short": "Neutral",
"economic_alignment_pressure_medium": "Neutral",
"economic_alignment_pressure_long": "Tailwind",
"economic_alignment_source": "sector-map (XLK; PLTR is not on the macro report's Economic Watchlist Forecast)",
"macro_report_date": "2026-08-12",
"xlk_signal": {
"s": "N",
"m": "N",
"l": "O"
},
"driver_name": "Enterprise & government adoption of applied AI (commercial AIP + US/allied defence software budgets)",
"driver_label": "Strong Tailwind",
"driver_historical": 92,
"driver_current": 88,
"driver_forward": 78,
"driver_commodity_trend": "N/A - the driver is enterprise and government AI-software demand, not a commodity price. Step 2b overlay does not apply.",
"confidence": {
"quality": 82,
"valuation": 78,
"timing": 70,
"driver": 74,
"econ": 68,
"overall": 70
},
"overall_confidence": 70,
"analyst_consensus_target": 176.33,
"analyst_target_high": 215,
"analyst_target_low": 80,
"analyst_target_median": 200,
"analyst_target_upside_pct": 1.3,
"analyst_target_yahoo_mean": 191.68,
"analyst_target_last_month_avg": 175,
"analyst_target_last_month_count": 9,
"analyst_grades_consensus": "Buy",
"analyst_bullish_pct": 50,
"analyst_coverage_count": 26,
"recent_upgrades_30d": 1,
"recent_downgrades_30d": 0,
"fmp_rating": "B+",
"fmp_overall_score": 3,
"fair_value_est": 112,
"fair_value_basis": "warranted 33.0x applied to FY2028 consensus EPS of $3.401, undiscounted - the most generous disciplined construction. On trailing clean EPS of $1.0105 the anchor is $33; on FY2027 consensus EPS of $2.263 it is $75.",
"stop_loss": 150.0,
"target_price": 195,
"levels": {
"support": 152.7,
"support_cluster_note": "6 Aug 2026 swing low 152.70 + EMA20 152.90 + SMA200 152.09",
"stop": 150.0,
"ema20": 152.9,
"sma20": 145.53,
"sma50": 134.99,
"sma200": 152.09,
"resistance": [
179.6,
190.0,
198.88,
207.52
],
"gap_zone_unfilled": [
125.65,
145.15
],
"wk52_high": 207.52,
"wk52_low": 106.37,
"daily_atr": 8.41,
"beta": 1.563
},
"scenario_base_target": 195,
"scenario_bull_target": 240,
"scenario_bear_target": 95,
"scenario_probabilities": {
"bull": 0.25,
"base": 0.5,
"bear": 0.25
},
"scenario_weighted_value": 181.25,
"scenario_bear_cohort_leg": "Cohort-level multiple compression from 172.2x clean to ~90x clean = -48%, inherited from the macro report's armed 'S&P 500 concentration / AI earnings-quality unwind' tail. Carried separately from and deeper than the idiosyncratic leg (-25 to -30%).",
"entry_groups_met": 1,
"entry_conviction": "Half-Size",
"entry_groups_detail": {
"fundamental": "unmet",
"technical": "unmet",
"catalyst": "met"
},
"exit_groups_live": 0,
"exit_action": "Hold",
"hard_gate_state": "caution",
"gates_triggered": [
"valuation_ceiling (Gate 3): clean P/E 172.2x is 5.22x the warranted 33.0x (>=1.40x arm) AND >= the Information Technology guardrail line of 33.0x (guardrail arm). Either arm alone caps every horizon at HOLD, with no growth exception."
],
"gates_caution": [
"accounting_dilution (Gate 4): NOT triggered. SBC is 13.7% of Q2 FY2026 revenue (25% threshold) and diluted share count is up ~2.5%/yr (5% threshold). The caution is the 19-point gap between the 62% adjusted and 42.8% GAAP operating margin - fully disclosed, almost entirely SBC."
],
"short_entry_confirmed": true,
"short_hold_reason": "expensive",
"do_not_buy_triggers": [],
"do_not_buy_evaluation": {
"leverage_rising_rates": "clear - no borrowings ($0 interest expense in each of the last 8 quarters) against $9.41bn cash",
"valuation_extreme_relative": "clear - P/S 67.9x is mid-range in its own 2-yr history (peak ~119x Nov 2025) and growth is accelerating, so the 'no corresponding acceleration' condition also fails",
"valuation_extreme_absolute_arm_a": "NOT FIRED. The arithmetic is satisfied (5.22x warranted against a 2.0x threshold; 5.2x the guardrail against a 1.5x threshold), but arm (a) is written 'with no exceptional, proven, durable growth' and Palantir now plainly has it: 12 consecutive quarters of acceleration, +92.8% YoY at a $6bn revenue base, Rule of 40 of 155 and NDR 157%. CAUSE OF THE CHANGE VS 2026-07-16 = NEW EVIDENCE, NOT SELF-CORRECTION: every one of those four figures is a Q2 FY2026 number with an SEC acceptance stamp of 2026-08-03, eighteen days AFTER the prior report. On 2026-07-16 the latest print was Q1 FY2026 at +84.7% YoY, a materially weaker basis for invoking the carve-out. The carve-out became available on the 3 August print; it is not that the prior report misread facts it already held. (An earlier draft of this run's report asserted 'firing it on the 2026-07-16 report was an error' - that framing is RETRACTED, see \u00a715.)",
"valuation_extreme_absolute_arm_b": "NOT FIRED. The Expensive leg holds; the LIVE de-rating catalyst leg does not. CAUSE OF THE CHANGE VS 2026-07-16 = NEW EVIDENCE, NOT SELF-CORRECTION - the macro grading moved, in this recorded sequence: MacroDriver-state-20260714 (the file live at the prior report) graded the 'S&P 500 concentration / AI earnings-quality unwind' tail 'armed', with a breadth_tell of 'top-10 ~41% of S&P, top-10 P/E ~50; SPY vs RSP re-widening' - i.e. breadth NARROWING, the falsification signal pointing the wrong way. (An earlier draft of this run asserted the 20260714 entry recorded NOTHING on the point; it does - the key is breadth_tell, not falsification. Quoting it is stronger than asserting absence, because it is primary-source proof of the narrowing this argument rests on.) The record then wobbled: 20260720 breadth_tell 'breadth broadening - equal-weight RSP beating SPY on 1-wk and 1-mo; RSP above its 50-DMA'; 20260730 'top-10 ~41% of index; RSP flat while MSFT/XLK ripped +5.5% on 29-30 Jul (narrow)'. It then settled: 20260808 status 'armed (breadth improved: RSP +2.3% participating)'; 20260812 status 'armed but trigger RECEDING - breadth broadening (RSP +3.6% vs SPY +2.8% vs QQQ +0.6% 1mo)', carrying an explicit falsification key again - the first since 26 June, the key having been absent from the 3 July grading onward. So arm (b) FIRED CORRECTLY on 2026-07-16 (armed + breadth narrowing) and correctly does NOT fire now (armed + breadth broadening, trigger graded receding). On this leg the prior report was not wrong. Standing project precedent is that an armed-but-not-triggering tail does not fire arm (b); it caps at HOLD via the Valuation Ceiling with the cohort compression carried as a bear-case leg. Applied here for consistency with the other AI-cohort names; the 48% cohort-compression leg is in \u00a711 Bear.",
"negative_eps_revisions": "clear - the opposite; FY2026 guidance raised from $7.65-7.66bn to $8.15-8.16bn and consensus EPS revised up",
"insider_selling_spike": "clear with a note - AI-cohort insiders sold >$3.4bn over the trailing year with zero purchases, but Palantir's are overwhelmingly 10b5-1/RSU dispositions and Thiel retains ~3%. Feeds the management-alignment sub-score (58), not the trigger.",
"structural_threat": "clear - the 'AI commoditises software' candidate was directly refuted by Q2 FY2026 (+92.8% growth, 157% NRR). Downgraded to a competitive vector."
},
"systemic_tail_cohort_member": true,
"systemic_tail_cohort_argument": "Limb 1 (valuation/earnings materially levered to the AI monetisation trade) PASSES overwhelmingly - the entire 172x multiple is an AI bet; US commercial revenue +149% in Q2 FY2026, guided >$3.42bn for FY2026; a $418bn top-30 S&P constituent. Limb 2 (top-weight constituent whose reported earnings are inflated by non-operating gains) FAILS and is conceded - nonop is 14.5% of net income and is clean interest income on a real $9.41bn cash pile, not a private-AI-stake markup. Limb 1 alone is sufficient for membership, hence the 48% cohort-compression leg in \u00a711 Bear.",
"amplification": "none. Driver 86 (>=65) IS eligible, but three independent bars apply: (1) the base signal is HOLD on every horizon and HOLD never amplifies; (2) economic pressure is Neutral, not Tailwind; (3) the Expensive band bars STRONG BUY outright.",
"analysis_status": "on-going",
"focus_qualifies": false,
"focus_reason": "No short-term BUY - signal HOLD on all three horizons (Gate 3 Valuation Ceiling).",
"next_update_date": "2026-08-31",
"next_check_date": "2026-08-31",
"next_update_basis": "default +14d from the 2026-08-16 report date, rolled from Sun 30 Aug to the next trading day. No discrete dated catalyst inside the window; Q3 FY2026 earnings expected ~early Nov 2026. Note the macro report refreshes 2026-08-20 - if it DISARMS the AI-concentration tail the bear-case leg lightens, and if it escalates the tail to actively triggering, DNB Trigger 2(b) would need re-testing.",
"delta_vs_prior": "SIGNAL FLIP: DO NOT BUY/DO NOT BUY/DO NOT BUY -> HOLD/HOLD/HOLD, all three horizons. Price rose from $133.76 to $174.04 (+30.1% in four weeks) - the prior DO NOT BUY was wrong on direction by 30 points and that is stated at the top of the report. The flip is driven by NEW EVIDENCE, not by re-reading our own logic - an earlier draft wrongly framed it as pure self-correction. Arm (b) needed a LIVE de-rating catalyst: MacroDriver-state-20260714 (live at the prior report) graded the AI-concentration tail 'armed' with a breadth_tell of 'top-10 ~41% of S&P, top-10 P/E ~50; SPY vs RSP re-widening' - breadth NARROWING; 20260808 graded it 'armed (breadth improved: RSP +2.3%)'; only 20260812 graded it 'armed but trigger RECEDING', with an explicit falsification key again (the first since 26 June; absent from 3 July onward). So arm (b) fired CORRECTLY on 16 July (armed + narrowing) and correctly does not fire now. Arm (a) needed 'exceptional, proven, durable growth': the four facts establishing it (+92.8% YoY, 12 accelerating quarters, Rule of 40 155, NDR 157%) are all Q2 FY2026 figures accepted by the SEC on 2026-08-03, eighteen days AFTER the prior report - on 16 July the latest print was Q1 FY2026 at +84.7%. Per standing precedent an armed-but-not-triggering tail caps at HOLD via the Valuation Ceiling and is carried as a bear-case leg instead (a 48% cohort-compression leg in \u00a711). What survives is Gate 3: clean P/E 172.2x vs warranted 33.0x = 5.22x, deep Expensive, guardrail breached. Scores: Q 85->86, V 29->20 (the +30% move consumed the entire analyst-target cushion; consensus upside fell from ~40% to +1.3%), T 48->61 (relative strength flipped from -12.0/-9.0pp to +26.9/+26.7pp), Driver 81->86 (Q2 FY2026: revenue +92.8%, US commercial +149%, NRR 157%, largest guidance raise in company history). Econ stance Contrarian 48 -> Neutral 50 (XLK N/O/O -> N/N/O). entry_conviction moved from Wait to Half-Size and entry_groups_met from 0 to 1, because the Catalyst entry group opened on the print; short_entry_confirmed moved from false to true. exit_action moved from Reduce to Hold (0 groups live), which IS a genuine correction of a prior over-read rather than a relaxation - the only one of the four changes that is. hard_gate_state donotbuy -> caution. Status stays on-going (Palantir is not a Donatien Pick, so no status rule applies). Clean P/E moved only 169x -> 172.2x while the stock rose 30% - earnings grew into the multiple, so the de-rating risk is undiminished.",
"prior_report": {
"date": "2026-07-16",
"price": 133.76,
"signals": "DO NOT BUY/DO NOT BUY/DO NOT BUY",
"quality": 85,
"valuation": 29,
"timing": 48,
"driver": 81,
"clean_pe": 169,
"warranted_multiple": 28.2,
"warranted_ratio": 5.98,
"hard_gate_state": "donotbuy",
"entry_conviction": "Wait",
"entry_groups_met": 0,
"short_entry_confirmed": false,
"exit_action": "Reduce",
"analysis_status": "on-going",
"do_not_buy_triggers": [
"Trigger 2 (Valuation Extreme - absolute), arms (a) and (b)"
]
},
"relative_strength_vs_spy": 26.9,
"relative_strength_vs_sector": 26.7,
"nvidia_partnership_date": "2026-06-29 (announced); stock +7.8% to $125.73 on 2026-07-01",
"aug7_move_attribution": "The 2026-08-07 +10.32% session to $172.01 was driven by Bank of America reaffirming Buy and raising its target to $255 (which reconciles to Yahoo's reported target high), plus a broad agentic-AI cohort rally (UiPath +7%, C3.ai +5%). Volume 77,579,878 (77.6m) per get_technical_indicators, this report's dating and volume authority - roughly 1.8x the ~43.2m average; syndicated coverage rounded it to 76.2m. NOT an Nvidia announcement - an earlier draft misattributed it.",
"adj_fcf_q2fy26": "1.22bn, 63% margin, +114% vs $569m adjusted in Q2 FY2025 (Q2 FY2025 operating cash flow was $539m)",
"risk_free_deviation": "SKILL Pillar-2 specifies the macro snapshot (4.70%); this report uses FRED DGS10 4.63% at 2026-08-13 per the run instruction. Direction is generous to the company; MOOT under the guardrail cap - at 4.70% raw warranted is 33.3x vs 33.7x, both capped to 33.0x, so ratio/band/gate are identical.",
"framework_gap_raised": "DNB arm (a)'s growth carve-out is UNBOUNDED. PLTR sits at 5.22x warranted - 2.6x the arm's own 2.0x trigger and the highest ratio in the report book - so the carve-out is doing unprecedented work and would do the same at 10x or 20x. Proposal for the SKILL owner: bound it by magnitude (unavailable above ~3.0x warranted) or state that it does not scale. Raised, NOT acted on in this report. ASYMMETRY THAT KEEPS THE FRAMEWORK COHERENT: Gate 3's guardrail arm is written with NO growth exception while arm (a) is written with one - so growth takes a name out of the PROHIBITION, never out of the CAP. PLTR still caps at HOLD via Gate 3 (clean P/E 172.2x actual vs a 33.0x guardrail), so the carve-out moved DO NOT BUY -> HOLD, not -> BUY.",
"tail_status_sequence": {
"20260714": {
"status": "armed",
"breadth_tell": "top-10 ~41% of S&P, top-10 P/E ~50; SPY vs RSP re-widening",
"read": "narrowing"
},
"20260720": {
"status": "armed",
"breadth_tell": "breadth broadening - equal-weight RSP beating SPY on 1-wk and 1-mo; RSP above its 50-DMA",
"read": "broadening"
},
"20260730": {
"status": "armed",
"breadth_tell": "top-10 ~41% of index; RSP flat while MSFT/XLK ripped +5.5% on 29-30 Jul (narrow)",
"read": "narrowing"
},
"20260808": {
"status": "armed (breadth improved: RSP +2.3% participating)",
"read": "broadening"
},
"20260812": {
"status": "armed but trigger RECEDING - breadth broadening (RSP +3.6% vs SPY +2.8% vs QQQ +0.6% 1mo); AI Big 10 = 41% of index",
"falsification": "breadth continues broadening - equal-weight new highs",
"read": "broadening, trigger receding"
},
"note": "Primary-source sequence for the arm (b) argument. On 2026-07-16 the tail was armed AND narrowing, so arm (b) fired correctly then; it correctly does not fire now."
}
}
In one paragraph. Palantir is an outstanding business — Quality 86, a Q2 FY2026 Rule of 40 of 155, net dollar retention of 157%, $9.41bn of cash and no borrowings — trading at 172.2 times clean earnings against a rate-and-growth-warranted 33.0 times, a ratio of 5.22. That puts it deep in the Expensive band on both available tests, which triggers the Valuation Ceiling gate and caps every horizon at HOLD. No Do-Not-Buy trigger fires: arm (a) is disapplied by the exceptional-growth qualifier, and arm (b) needs a live de-rating catalyst, which an armed-but-receding macro tail is not. The cohort de-rating is carried instead as a 48% leg in the bear case, where the framework puts it. Two things the reader should weigh against this verdict: the prior report made a harder call at $133.76 and the stock is up 30.1% since, and the tape has now flipped from 27 points of underperformance to 27 points of leadership. One thing that supports it: the price rose 30% and the multiple did not move, which means nothing about the de-rating risk has been resolved — the altitude is unchanged; only the earnings underneath it have grown.
get_income_statement. Triangulated across three independent syndications (Businesswire, Nasdaq, Yahoo Finance) plus earnings-call coverage: only the press-release-only metrics — net dollar retention, adjusted operating margin, adjusted free cash flow, TCV bookings, the Rule of 40 and the FY2026 guidance rangeDeclared deviation from the SKILL — the risk-free input. The SKILL's Pillar-2 anchor specifies reading the risk-free rate from the macro report's market snapshot, which carries 4.70%. This report instead uses 4.63%, FRED series DGS10 at 13 August 2026, sourced directly, on the run instruction for this batch. Two things make that safe to disclose rather than argue about: the direction is generous to the company (a lower risk-free rate produces a higher warranted multiple and so a lower ratio), and it is moot — at 4.70% the raw warranted multiple is 33.3× against 33.7× at 4.63%, and the Information-Technology guardrail caps both to 33.0×, so the warranted multiple, the 5.22 ratio, the band and the gate are identical either way.
A framework gap, raised rather than acted on. Do-Not-Buy arm (a) carries an unbounded growth carve-out: a name with exceptional, proven, durable growth escapes the deep-expensive arm no matter how deep the excess runs. Palantir sits at 5.22× warranted — 2.6× the arm's own 2.0× trigger, and the highest ratio anywhere in this report book. The carve-out is doing work no precedent has asked of it, and on the current wording it would do exactly the same work at 10× or 20× warranted. One asymmetry keeps the framework coherent, and it is written into the rules: Gate 3's guardrail arm carries no growth exception while arm (a) is written with one — so exceptional growth can take a name out of the prohibition, never out of the cap. Palantir still caps at HOLD via Gate 3 at 172.2× actual against a 33.0× guardrail, so the carve-out moved this report from DO NOT BUY to HOLD, not to BUY. We apply the rule as written, but the SKILL should probably bound it by magnitude — unavailable above roughly 3.0× warranted — or state explicitly that it does not scale with the size of the excess. Flagged for the framework owner; not acted on here.
Pricing basis. This report is dated 16 August 2026, a Sunday. US markets were closed on both 15 and 16 August, so every price, indicator and level is taken from the Friday 14 August 2026 close of $174.04 — the latest print available. Nothing here is a stale carry-forward; it is simply the most recent session.
Net effect of data failures on confidence: small. Three sources failed. The earnings-calendar failure is immaterial — the filing cadence answers the question unambiguously and the answer is more than 14 days out either way. The SEC 403 was fully mitigated by triangulating every press-release figure across three independent syndications. Only the missing options-skew read cost anything: 5 points off timing confidence, which stands at 70%.
Overall confidence 70% = the minimum of the three fundamental pillars (Quality 82, Valuation 78, Timing 70), per the framework's weakest-link rule. Up from 64% at the last report — the Q2 FY2026 print resolved a great deal of genuine uncertainty about the growth trajectory.
| Trap | Checked? | Result |
|---|---|---|
Polygon get_stock_prices date labels run one session early | Yes — and handled, not merely disclosed | Confirmed present on this ticker. The tell was there: bars labelled Sunday 2 August 2026 carrying 77m shares of volume. Not one date in this report comes from those labels. Every dated claim — the 4 August earnings gap, the 6 August swing low at $152.70, the 7 August Bank of America / agentic-AI cohort leg, the 13 and 14 August rejections at $179.60–180.18, the 10 August RSI peak of 73.9 — is taken from get_technical_indicators, whose labels were verified exact against the Yahoo quote ($174.04 with a $179.01 previous close), the snapshot, and the 10-Q acceptance timestamp of 3 August 2026 18:06. The OHLCV sequence from get_stock_prices is used only for the chart series and the relative-strength arithmetic, where labels do not matter. No sub-score rests on a Polygon-labelled date. |
FMP totalDebt is lease-inclusive | Yes | Caught, and it matters here. Providers report total debt of $211.4m. The income statement shows interest expense of $0 in every one of the last eight quarters, which is direct evidence that Palantir carries no borrowings — the $211m is operating-lease liabilities. Definition used: we quote cash of $9.41bn and describe the company as having no borrowings, and we use the lease-inclusive $10.10bn denominator for ROIC (25.7%) rather than the more flattering borrowings-only $9.89bn (26.3%). Both figures are shown in §3; the conservative one is scored. |
| Mega-cap stale share count / broken EV | Yes | Caught one. FMP reports a market cap of $399.6bn, implying roughly 2.296bn shares — stale. Reconciled: Yahoo's $418.2bn ÷ $174.04 = 2.403bn shares, matching the Q2 FY2026 weighted-average basic count of 2,396.9m. Diluted is 2,569.8m and is what every per-share figure uses. FMP's EV ($397.8bn) is also below Yahoo's ($409.1bn); we used the higher, more conservative figure for the FCF yield. |
| Non-operating / mark-to-market earnings inflation (step 7b) | Yes — mandatory, run every time | Non-operating income is 14.5% of net income, just under the 15% threshold, and is clean: $266.4m of interest income on a real $9.41bn cash pile, not a private-AI-stake write-up. Normalised anyway; every multiple in §4 is scored on the clean basis. |
| Period mislabelling (Q vs FY) | Yes | Every company figure in this report carries its period explicitly — Q2 FY2026, H1 FY2026, FY2026 guidance, or TTM. The two most easily confused: revenue growth is +92.8% in Q2 FY2026 but +78.9% TTM; the Rule of 40 is 155 on the Q2 FY2026 construction but 133 on the TTM construction. Both are shown wherever either is cited. |
| auto_adjust inverting a moving-average read | Yes | Not applicable, and doubly safe: all price data is raw Polygon OHLC, and Palantir pays no dividend, so there is no adjustment wedge at all. |
| Commodity price-trend overlay (step 2b) | Yes | Not applicable — the driver is enterprise and government AI-software demand, not a commodity price. Recorded as such. |
| Foreign-currency issuer frame | Yes | Not applicable — Palantir reports and trades in USD, so the anchor's US risk-free rate, 4.5% ERP and 3% terminal growth are all inside one currency frame. No FX rate was used, so none is recorded. |
| Degenerate analyst consensus | Yes | Not degenerate — high $215, low $80, median $200 and consensus $176.33 are four distinct values. Yahoo pulled as a cross-check regardless. |
| Dividends verified with the tool, not a web summary | Yes | get_stock_dividends returns zero history. No dividend, no yield, nothing to reconcile. |
| Economic-study existence check | N/A | Not a development-stage resource name. |
| Polygon pre-earnings price trap | Yes | The Q2 FY2026 print (10-Q accepted 3 Aug 2026, 18:06) is fully reflected; the +29.45% session of 4 August 2026 is in the series. |
The pre-publication linter scans for any headline value from the prior report that still appears in this one without 'was/prior/from' framing, because that pattern is what a missed propagation looks like. All four of its hits here are numeric collisions rather than stale figures, and all four are checked and correct:
prior_report, which is exactly where a prior price belongs. This report's price is $174.04.Trailing EPS. We use $1.18 — the sum of the four reported diluted quarterly figures ($0.41 + $0.34 + $0.24 + $0.19) — giving a trailing P/E of 147.5×, which reconciles exactly to FMP's priceToEarningsDilutedRatioTTM of 147.49. Yahoo shows $1.17 and 148.75× on a re-pull, and $1.16 / 150.03× two days earlier — the field is unstable. The difference is a rounding convention, not an error, and it does not affect any conclusion: the scored multiple is the clean 172.2× either way.
Free cash flow. Yahoo's free_cashflow field shows $2.159bn; FMP's per-share TTM data implies $3.359bn (operating cash flow $3.402bn less capex $42m). We use FMP's, because it is internally consistent with the same TTM window as the revenue figure, and because the company reported $1.22bn of adjusted free cash flow in Q2 FY2026 alone — which makes Yahoo's annual figure impossible. The choice is conservative in the direction that matters: a higher FCF number produces a better FCF yield, and it is still only 0.82%.
Directional calls reconciled against the pulled numbers. Quality 86 is supported by +78.9% TTM revenue growth, a 42.8% GAAP operating margin, a 54.6% FCF margin and no borrowings — every figure traced to get_income_statement or get_financial_ratios. Valuation 20 follows arithmetically from a 5.22 warranted ratio and sits inside the Expensive band's 0–39 range. Timing 61 is the documented weighted sum of five sub-scores. Driver 86 is the documented 25/50/25 weighting. The three per-horizon composites were recomputed from the framework weights (Short = T55/V25/Q20 = 55.75 → 56; Medium = Q35/V35/T30 = 55.40 → 55; Long = Q55/V30/T15 = 62.45 → 62, on Q 86 / V 20 / T 61) and the gate caps the signal, not the composite. The three scenario probabilities sum to 100 with the base most probable.
Live-verify items sourced this run. Q2 FY2026 results and the FY2026 guidance raise — SEC 8-K/10-Q metadata plus three syndications, 3 August 2026. Nvidia partnership — announced 29 June 2026, corroborated across contemporaneous coverage and by the raw price reaction ($116.67 on 30 June to $125.73 on 1 July, +7.77%). It predates the prior report and is not a this-quarter event. The separate 7 August +10.32% leg was verified to a Bank of America target raise to $255 plus a broad agentic-AI cohort rally (UiPath +7%, C3.ai +5%) — BofA's $255 reconciles to Yahoo's reported target high. Analyst grades — get_stock_grades, most recent 4 August 2026. Macro tail status — MacroDriver-state-20260812.json, read directly. The 14 August consumer prints — verified centrally this run. Corporate status — isActivelyTrading: true, no halt, no split, no dividend, no pending transaction.
Four changes against the prior report, labelled by what actually caused each — because an earlier draft of this report got that badly wrong. That draft called all four ‘corrections to our reasoning rather than changes in the world’. Two were nothing of the kind, and the distinction matters: it is the difference between a framework catching its own error and a framework being handed new facts. Stated properly: (1) [NEW EVIDENCE] Arm (b) lost its catalyst leg because the macro grading moved. Arm (b) needs a live de-rating catalyst. The tail-status sequence is on the record: MacroDriver-state-20260714 — the file live at the prior report — graded it ‘armed’ with a breadth tell of ‘top-10 ~41% of S&P, top-10 P/E ~50; SPY vs RSP re-widening’ — breadth narrowing, the falsification signal pointing the wrong way. (An earlier draft of this section claimed that file recorded nothing on the point. It does: the key is breadth_tell rather than falsification, and quoting it is stronger than asserting absence, because it is primary-source proof of the narrowing this argument rests on.) The record then wobbled — 20260720 ‘breadth broadening — equal-weight RSP beating SPY on 1-wk and 1-mo’, 20260730 ‘RSP flat while MSFT/XLK ripped +5.5% on 29-30 Jul (narrow)’ — before 20260808 graded it ‘armed (breadth improved: RSP +2.3% participating)’ and 20260812 graded it ‘armed but trigger RECEDING’, carrying an explicit falsification key again — the first since 26 June, the key having been absent from the 3 July grading onward. That absence-and-return is mild corroboration rather than an embarrassment: across the nine graded states from 20 June to 12 August the falsification key appears only where the tail is graded not triggering or receding (20 June, 26 June, 12 August) and in none of the five consecutive bare-armed gradings that span the prior report (9 July to 8 August) — with one exception we state rather than hide, 20260703, graded armed_not_triggering but carrying no key. So the field co-moves with the tail becoming testable again; it does not track it perfectly, and the claim is put no higher than that. The RECEDING grading did not exist on 16 July, and the 14 July file's own breadth tell records the opposite condition. So arm (b) fired correctly on 16 July (armed + narrowing) and correctly does not fire now (armed + broadening, trigger graded receding). This is a change in the macro regime as the macro report records it, not a re-reading of the rule — and on this leg the prior report was not wrong. Withdrawn consistent with how the same tail has been treated across the other large AI-cohort names; the risk is carried as a 48% bear leg instead. (2) [NEW EVIDENCE] Arm (a) is disapplied by a print that post-dates the prior report. The four facts establishing exceptional, proven, durable growth — +92.8% year-on-year, twelve consecutive accelerating quarters, a Rule of 40 of 155, net dollar retention of 157% — are all Q2 FY2026 figures carrying an SEC acceptance stamp of 3 August 2026, eighteen days after the prior report. On 16 July the latest print was Q1 FY2026 at +84.7%, a materially weaker basis for invoking the carve-out. Together (1) and (2) move the signal from DO NOT BUY to HOLD — on new evidence, not on hindsight applied to evidence already in hand. (3) [GENUINE SELF-CORRECTION — the only one of the four] The prior run's exit ladder was over-read. The Thesis-Invalidation group's catastrophic item is scoped to a financial-distress, dilution or going-concern gate — a valuation gate is not on that list — and the Profit-Target group requires price at the median target and RSI above 70, neither of which held then or now. On a strict reading none of the three groups was live at the prior report either. Today's "Hold (0 live)" is a correction, not a relaxation. (4) [DISCLOSURE, not a change] The Catalyst entry group is scored MET on an eight-session-old event. The plain reading of the rule is event-anchored, but evaluated on the latest session the volume sub-condition fails (0.55×, not 2×), and the reader is 7.0% above the trigger price. Disclosed in full in §12 rather than left to sit as a bare "met".
Carried-forward figures re-examined. No score, level or scenario was inherited. Every pillar was re-derived from this run's data; the chart series, the 50-day average, all support and resistance levels, the stop and all three scenario targets were regenerated from the current bars. The one carried element is the stored identity block (exchange, ticker, ISIN, currency), which is by design in a batch run.
Known limitations, stated plainly. The options-skew sentiment input is missing. The FY2029–FY2030 consensus rests on three EPS analysts and was deliberately excluded from the growth input. And the honest one: this framework has now been wrong on Palantir's direction for four consecutive weeks, by 30 percentage points. It has been wrong about the direction of the price, not about the level of the multiple — the multiple is essentially where it was. A reader is entitled to weigh that record when deciding how much authority to grant a Hold that rests on valuation alone.