Palantir Technologies PLTR

> ## ⚑ EARNINGS BANNER — REPORTED ONE DAY AGO

> Second-quarter 2026 results were released after the close on 2026-08-03 and the Form 10-Q was filed 2026-08-04, the day of this dive. The stock closed +29.42% at $162.61 on 172.0 million shares — 4.2x its average volume. Every analyst price target and every consensus estimate in this file predates the print, including a Q3 revenue consensus of $2.148B that now sits BELOW the company's own guidance of $2.160-2.164B. This dive is struck on a post-gap price, and the verdict is set with that timing as an explicit input.

Reference table

1. What the business is, and how the disclosure actually works

Palantir was founded in 2003, listed by direct listing on 2020-09-30, and is led by co-founder and chief executive Alexander C. Karp. Headquarters in Miami. 4,429 full-time employees — against trailing revenue of $6.156B, that is roughly $1.39 million of revenue per employee, a figure worth holding against the rest of this batch.

Two reportable segments, and the 10-Q defines the reporting metric unusually. Palantir reports Government and Commercial and evaluates each on contribution — segment revenue less the related cost of revenue and sales and marketing expense, excluding stock-based compensation, research and development, and general and administrative costs, which are managed at the corporate level. This is not gross profit and not operating profit, and it is the company's own construction. We report it because it is the only segment profitability disclosed, and we label it.

The single most important line in that table is the commercial contribution margin: 78%, up from 64% a year ago, and now ABOVE government's 71%. For most of Palantir's history the commercial business was the lower-margin, harder-to-deploy half — the one where the forward-deployed-engineer model looked like disguised consulting. It has now crossed over. That is the quantitative fingerprint of a platform generalising: the marginal cost of the next commercial deployment is falling faster than the price.

Geography, from the 10-Q, and it is the clearest weakness in the file:

The 10-Q adds that no other country represented 10% or more of total revenue. So the entire hyper-growth story is one country, and the rest of the world is compounding at less than a third of the U.S. rate. That is simultaneously the biggest risk to the narrative and the largest unexploited opportunity in it, and the file does not resolve which.

The U.S. split, from the earnings release: U.S. commercial revenue $764M, +149% year on year and +28% sequentially; U.S. government revenue $809M, +90% and +18% sequentially. Both halves of the U.S. business are growing faster than the company as a whole.

Demand disclosures, all from the 2026-08-03 release, with the company's own caveat attached. Closed 220 deals of at least $1M, 98 of at least $5M, 73 of at least $10M. Total contract value closed: $3.373B, up 49% year on year, including a record $2.132B of U.S. commercial TCV, up 153%. U.S. commercial remaining deal value: $6.238B, up 124% year on year and 27% sequentially.

Read those with the definition the company itself supplies: "TCV, ACV, and RDV each presume the exercise of all contract options available to our customers and no termination of contracts. However, the majority of our contracts are subject to termination provisions, including for convenience, and there can be no guarantee that contracts are not terminated or that contract options will be exercised." The GAAP equivalent is remaining performance obligations, and that number is $4.9B — 0.8x trailing revenue — of which the company expects 43% to convert to revenue within twelve months and 36% over the following two years. The gap between $6.238B of U.S. commercial RDV alone and $4.9B of total company RPO is the measure of how much of the headline backlog is optional.

2. The growth record, and why the second derivative is the story

Annual, from the vendor income series, cross-checked against the filings:

Quarterly, and this is the table the whole dive turns on:

Three facts follow and they are the entire growth case.

First, the year-on-year rate has risen every quarter for a year: 48.0%, 62.8%, 70.0%, 84.7%, 92.8%. Not one deceleration. At $678M of quarterly revenue that would be notable; at $1.9B it is close to unprecedented for a company of this size that is not being acquired into growth.

Second, operating margin expanded from 15.5% to 47.1% across the same nine quarters, without a single reversal after the Q4 FY2024 anomaly. Gross margin rose from 81.0% to 84.7%. Palantir is not buying its growth with margin.

Third, it has beaten consensus in all five reported quarters and the beat is widening: +16.0%, +25.0%, +8.7%, +13.8%, +18.9% on EPS, and on revenue +7.0%, +8.2%, +4.9%, +5.9%, +6.8%. The June quarter beat revenue by $123.2M and EPS by $0.065.

And the guide raised the bar again. FY2026 revenue guidance of $8.150-8.158B against a consensus in this file of $7.851B — 3.9% above. Q3 guidance of $2.160-2.164B against a consensus of $2.148B. When company guidance sits above sell-side consensus at the point of publication, the estimates in this file are not the estimates the market will be working from next week, and every multiple computed against them is overstated.

The chief executive's framing, half-weighted per house rule and quoted for the record:

> "Demand for AI sovereignty has now been unleashed... This quarter was otherworldly: our U.S. commercial revenue grew 149% year-over-year, our overall revenue grew 93% year-over-year, and our Rule of 40 score climbed to 155%." — Alex Karp, 2026-08-03

We half-weight the framing and fully accept the three numbers, because all three are independently verifiable in the 10-Q filed the following morning.

3. Margins, dilution, and the tax line nobody is pricing

Stock-based compensation is the standard objection to Palantir and the standard objection is weakening.

SBC grew 48% in the half while revenue grew 89%. The ratio has fallen 360 basis points year on year. That is the opposite of the usual pattern for a company at this growth rate, and it is why the GAAP operating margin (47%) is converging on the adjusted figure (62%) rather than diverging from it. The gap between the two — $282.5M in the quarter — is entirely SBC plus related employer payroll taxes.

Diluted share count is nonetheless still rising: 2,414.7M (Q2 FY2024) → 2,562.9M (Q2 FY2025) → 2,569.8M (Q2 FY2026). The year-on-year increase is now only 0.3%, against 6.1% the year before. Dilution has essentially stopped. The company also repurchased $74.985M of stock in FY2025 — trivial at 0.02% of market capitalisation, and mentioned only for completeness.

Now the tax, which is the most consequential unmodelled item in this file.

The trailing effective tax rate is 1.43%. In the June quarter, income before tax was $1,081.345M and the provision was $15.383M. The reason is in the 10-Q, filed 2026-08-04:

> "...due to the weight of objectively verifiable negative evidence, including its history of U.S. and U.K. net operating tax losses, the Company has maintained a full valuation allowance on its U.S. and U.K. deferred tax assets as of June 30, 2026. However, given the Company's recent earnings and anticipated future earnings, there is a reasonable possibility that it will have sufficient positive evidence in the future to release all or a portion of the valuation allowance it recorded against its deferred tax assets."

Two separate consequences, and they run in opposite directions.

The first is that trailing earnings are flattered. Trailing pre-tax income is $3,071.830M. At a 21% statutory rate, trailing net income would be roughly $2.43B rather than $3.02B, and trailing diluted EPS $0.94 rather than $1.18. The trailing multiple is therefore 173x on a normalised basis, not 138x. We use $1.18 for the headline because that is what was reported, and we show both.

The second is a future data-integrity landmine and we are pre-registering it now. If Palantir releases the valuation allowance, GAAP net income in that quarter will include a large, one-off, non-cash tax benefit. That is precisely the defect class that has caught out other names in this process — a discrete tax item capitalised as run-rate. When it happens, the clean operating EPS will be materially below the reported figure, and this dive says so in advance.

Cash conversion, meanwhile, is genuine. Cash from operations was $1.216B in the June quarter, a 63% margin; adjusted free cash flow $1.220B, also 63%. Full-year adjusted free cash flow guidance is $4.5-4.7B. FY2025 operating cash flow was $2.134B against net income of $1.635B — an income-quality ratio of 1.31x, and the trailing figure is 1.12x. Capital expenditure is $33.9M a year, 0.7% of revenue. This is an asset-light business in the most literal sense: property and equipment, net, is $61.4M on an $11.679B balance sheet.

4. Balance sheet — and two vendor errors pointing in opposite directions

From the 10-Q filed 2026-08-04:

There is no debt on this balance sheet — no term loan, no revolver drawn, no convertible notes, no finance leases. Total liabilities of $1.794B are, almost entirely, deferred revenue, customer deposits, operating leases and accrued payroll. The accumulated deficit fell from −$3.562B to −$1.630B in six months on $1.932B of first-half net income; at the current run-rate Palantir's retained earnings turn positive within about three quarters.

Correction 1 — netDebt ignores $5.753B of short-term investments AND is six months stale. The vendor reports netDebt of −$1,194.458M at 2025-12-31, computed as capitalLeaseObligations of $229.338M less cash of $1,423.796M. It excludes the $5,753.247M of short-term investments sitting on the same balance sheet. The correct net cash position at that date was $6.948B; the correct position at 2026-06-30 is $9.409B less $211.4M of noncurrent lease obligations, or roughly $9.20B — which matches the earnings release's stated "$9.2 billion of cash, cash equivalents, and short-term U.S. Treasury securities" exactly. This is the confirmed net-debt defect class and the magnitude is $8.2B.

Correction 2 — the vendor's totalDebt of $229.338M is an operating-lease liability, not debt. The balance sheet has no debt line at all. Every leverage ratio derived from it (debtToEquityRatioTTM 0.0216, debtToAssetsRatioTTM 0.0181, debtToMarketCapTTM 0.00057) is measuring lease obligations. interestCoverageRatioTTM and debtServiceCoverageRatioTTM both read 0, which is not poor coverage but an undefined ratio with a zero denominator. We reject all of them and state that Palantir has no debt.

Correction 3 — and this one runs the other way — the vendor market capitalisation counts one share class.

> Vendor marketCap $373,363,942,700 at $162.61 implies 2,295,946,000 shares.

> The 10-Q states, at 2026-06-30: 2,300,516,590 Class A + 101,375,151 Class B + 1,005,000 Class F = 2,402,896,741 shares outstanding.

> Q2 diluted weighted-average shares: 2,569,826,000.

>

> Corrected market cap: 2,402,896,741 × $162.61 = $390,735,000,000 — the vendor understates by $17.37B, or 4.4%.

> On the diluted count the figure is $417.83B.

The 10-Q is explicit that the classes differ only in voting and conversion: "The Company's Class A, Class B, and Class F common stock (collectively, the 'common stock') all have the same rights, except with respect to voting and conversion rights." They are therefore all economically identical and all belong in the market capitalisation. The Class F shares are held in a voting trust established by the three founders and are the control mechanism.

The rebuilt enterprise value:

> Corrected market cap $390.735B

> − cash and marketable securities $9.409B

> + debt $0

> = corrected enterprise value $381.326B

Note that the two errors do not cancel here — they compound in the same direction on the sales multiple. The vendor's $371.545B enterprise value is $9.78B too low, and its price-to-sales ratio of 60.65x understates the corrected 63.47x.

5. Cash flow and capital returns

Free-cash-flow margin has gone 9.6% → 31.3% → 39.8% → 46.9% → 63% in the latest quarter. Capital expenditure of $33.9M against $4.475B of FY2025 revenue is 0.76%. The company has no dividend, no meaningful buyback, and no need for either — it is compounding cash at a rate that will produce more than $4.5B this year against a $9.4B existing balance.

On the corrected market capitalisation of $390.735B, guided FY2026 adjusted free cash flow of $4.6B is a free-cash-flow yield of 1.18%. That is the honest way to state the valuation problem: you are being offered 1.2% of cash today in exchange for a growth rate that must persist for many years to justify the price. The counter-argument is equally honest: the same figure was 0.6% a year ago and the cash flow doubled.

6. Valuation — priced in or room?

At $162.61 (corrected market cap $390.735B, 2,402,896,741 shares, corrected enterprise value $381.326B):

Estimate coverage is good on FY2026 and FY2027 (20-24 analysts) and thin on FY2028 EPS (9). The FY2029 and FY2030 rows carry three analysts each on EPS and are excluded entirely — and they are also internally implausible: revenue growth of +45.2% (FY2028), +86.3% (FY2029) and +120.4% (FY2030) is a growth rate that decelerates then re-accelerates twice, which no bottom-up model produces.

Per the data contract we tested est.ebitAvg and est.ebitdaAvg and they FAIL against the income statement. FY2026 ebitAvg of $1.701B implies a 21.7% operating margin against the 47.1% Palantir actually delivered in the June quarter and the 60% implied by its own adjusted operating income guidance. ebitdaAvg of $1.873B is below the FY2026 netIncomeAvg of $3.449B — net income above EBITDA, which is the arithmetically-impossible signature the contract warns about, present in every forward year. sgaExpenseAvg of $4.628B is 59% of guided revenue against an actual trailing ratio of 31.6%. All four rows are rejected. The epsAvg and netIncomeAvg rows are internally consistent with each other on a ~2.3B share count and are the only estimate fields used in this dive.

Peer context. The vendor peer set is a scale-matched but analytically weak list — Adobe, AMD, ASML, Salesforce, CrowdStrike, Cisco, Oracle, Palo Alto Networks, UiPath, SAP. Only UiPath and, at a stretch, Salesforce are addressing anything resembling the same problem, and one knowledge-base claim makes exactly this objection: that software indices are 'poorly composed' because holdings like Oracle and Palantir 'are really datacenter/consulting plays, not pure software.' No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The relevant scale fact is that Palantir's corrected $390.7B market capitalisation now exceeds Oracle's $419.7B by less than 7% and is 2.5x Salesforce's $156.4B, on roughly a twentieth of Salesforce's revenue.

6a. What today's price assumes (the inversion)

At $162.61 — 46.8x guided FY2026 revenue and 33.0x the FY2027 consensus — the price embeds the following falsifiable claims:

• Revenue reaches $8.154B in FY2026 and at least $11.545B in FY2027. (FY2026 is company guidance, issued 2026-08-03; FY2027 is consensus from 24 analysts, struck BEFORE that guidance.) The FY2027 consensus is the fragile number in the wrong direction: it implies only 41.6% growth off the guided base after a year of 82%. The nearest check is 2026-11-02, where guidance is $2.160-2.164B against a vendor consensus of $2.148B — the consensus in this file is already below the company's own number.
• The market keeps paying more than 30x forward revenue in two years' time. (Our number.) At 20x our FY2027 revenue estimate the stock is $109; at 45x it is $245. The multiple, not the revenue, is where essentially all of the variance lives, and this is the assumption we would name as most fragile.
• Operating margin holds above roughly 60% on an adjusted basis. (Company guidance: adjusted operating income of $4.889-4.897B on $8.154B of revenue.) GAAP was 47.1% in the June quarter against 26.8% a year earlier, so the trajectory supports it. The check is whether stock-based compensation stays at or below 13-14% of revenue — it has fallen from 15.9%, and any reversal directly attacks the GAAP margin.
• The effective tax rate stays near zero for several more years. (Filing-confirmed at 1.43% trailing, with a full valuation allowance disclosed.) This is not sustainable and the company says so. Normalising to 21% takes roughly 20% off reported EPS. Consensus EPS of $1.495 (FY2026) and $2.143 (FY2027) almost certainly embed a sub-statutory rate; on a fully taxed basis they are nearer $1.19 and $1.71, and the forward multiples are 137x and 95x, not 109x and 76x.
• Contracts that are terminable for convenience behave like contracts that are not. (Filing-confirmed.) $6.238B of U.S. commercial remaining deal value and $3.373B of closed total contract value both presume full option exercise and no termination. The GAAP number is $4.9B of remaining performance obligations — 0.8x trailing revenue.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: revenue growth (+41.6% to +55%, consensus to our estimate) + multiple drift (COMPRESSION, from 46.8x guided FY2026 sales to roughly 31x on FY2027, −34%) + shareholder yield (0.0%).

That is the whole arithmetic and it is brutally simple: there is no dividend, no buyback of consequence, and no yield of any kind. One hundred percent of the return is growth fighting the multiple. The base case of $168 assumes substantial multiple compression — from 46.8x current-year sales to about 31x — offset by revenue growing more than 50%. Those two forces nearly cancel, which is exactly why the base sits 3.3% above spot.

Compression is assumed for a specific reason, not out of caution. Even on our above-consensus FY2027 revenue estimate of $12.6B, the growth rate falls from 93% to roughly 55%. A multiple appropriate to 93% growth is not appropriate to 55% growth, and the market has already shown on this exact security that it will reprice that transition violently — the 48% drawdown from $207.18 to $107.27 happened while revenue was accelerating.

Sensitivity, stated openly. On our FY2027 revenue estimate of $12.6B and 2,402.9M shares: at 20x enterprise value to sales the stock is $109; at 25x, $135; at 31x, $168; at 40x, $214; at 48x, $262. The entire fair-value range is one input. We are not going to pretend otherwise, and any dive on this name that claims greater precision is manufacturing it.

The bull case at $262 requires no re-rating at all — it holds today's 46.8x current-year sales multiple and applies it to our FY2027 revenue estimate. The bear case at $96 requires no growth failure at all — it applies 17.4x to the same revenue figure, a multiple that would still be among the highest in enterprise software.

6c. Variant perception (where we differ, what would surprise)

• We differ from the consensus estimates in this file, and we think they are simply out of date. FY2027 revenue of $11.545B implies 41.6% growth off a base the company guided to 82% growth the day before this dive. Our estimate is roughly $12.6B — 9% above consensus and still a 38-point deceleration from the 93% just delivered. This is not a heroic assumption; it is arithmetic applied to a guide the sell-side has not yet incorporated. Watchable event: consensus revisions over the next two weeks. If FY2027 consensus does not move above $12B, the market is disagreeing with the guide.
• We differ from the street on the direction of the multiple, and we are more bearish. The consensus target of $172.17 across 26 analysts, with a median of $195, implies the crowd expects the current multiple to be broadly sustained. We assume it compresses by roughly a third. Our base of $168 is 2.4% below the mean and 14% below the median. On a name where the knowledge base is overwhelmingly positive, we are deliberately taking the less popular side of the valuation, and the reason is the 48% drawdown that already happened once on unchanged fundamentals.
• We differ from almost everyone on the tax line, and we have not seen it discussed. A 1.43% trailing effective rate with a full valuation allowance is a temporary state the 10-Q says may end. Two consequences: reported EPS is roughly 20% flattered today, and a valuation-allowance release will produce a one-off GAAP earnings spike that must not be capitalised. We are pre-registering that now.
• Positive surprise that would force a re-rate upward: a Q3 print above the guided $2.164B with U.S. commercial growth still above 120%, which would make five consecutive quarters of acceleration; international revenue growth inflecting from 33.7% toward the U.S. rate, which would double the addressable story; or commercial contribution margin extending above 80%, confirming the platform generalisation.
• Negative surprise that would break the thesis: any sequential revenue deceleration below roughly 12%, which would end the acceleration series; commercial contribution margin falling back below government's; stock-based compensation reverting above 16% of revenue; a large contract termination-for-convenience disclosure against the $4.9B RPO; or a valuation-allowance release presented as operating earnings. Any of these hits a stock at 62x trailing sales the day after a 29% gap, and there is nothing underneath it.

Synthos fair values

All three anchors are enterprise-value-to-sales multiples of our own FY2027 revenue estimate of $12.6B — 9% above the $11.545B consensus, because that consensus predates the 2026-08-03 guidance raise. Labelled clearly: the revenue figure is ours, the multiples are ours, and the sensitivity above shows exactly how much each is worth. Cross-checked against consensus EPS throughout.

• Bear ~$96 — 17.4x our FY2027E revenue. Cross-checks: 30.1x the FY2028 consensus EPS of $3.192; 44.8x FY2027E consensus EPS; 10.5% below the actual 52-week low of $107.27. The scenario: FY2027 growth lands at the consensus 41.6% rather than our 55%, the market decides accelerating growth was a 2026 phenomenon, and the multiple compresses to a level that would still be the richest in enterprise software. −41.0%. Note that this scenario requires no operational failure — only a return to merely excellent growth.
• Base ~$168 — 31.1x our FY2027E revenue. Cross-checks: 78.4x FY2027E consensus EPS; 52.6x FY2028E; 41.2x guided FY2026 revenue; 2.6% below the 52-week high of... no, 18.9% below it. The scenario: the guide holds, FY2027 revenue beats the current consensus by roughly 9%, adjusted operating margin holds near 60%, and the multiple compresses by about a third as growth decelerates from 93% toward 55%. +3.3%, with no yield.
• Bull ~$262 — 48.6x our FY2027E revenue — today's guided-year multiple, HELD. Cross-check: 82.1x FY2028E consensus EPS. This bull case requires no re-rating whatsoever, only that FY2027 revenue lands where we think it will and the market keeps paying what it pays today. The scenario: U.S. commercial growth above 120% persists, sovereign demand compounds internationally, and the acceleration series extends. +61.1%.

Base is 3.3% above spot; asymmetry roughly 1.49:1 to the upside (41.0% down, 61.1% up). That ratio is genuinely attractive and it is not enough. What makes this a Watch rather than a Buy is that the base case is spot, the entry embeds a 29.42% one-day gap on 4.2x volume, and the single input that determines the answer — the terminal revenue multiple — is one we have no edge on and have watched halve on this exact security within the last twelve months.

7. Knowledge base — the deepest lane in this batch, and the most contaminated

Raw entity hits: 83. Text matches: 5. Discarded for homograph collision: 0. Used as name-level conviction after de-duplication and attribution correction: approximately 40.

The search covered PLTR, Palantir and Palantir Technologies across the entity field of all 51,928 distilled claims, plus a text sweep on \bpalantir\b. Stance on the raw entity lane: 60 bullish, 12 neutral, 11 bearish.

This is the widest lane in this batch by a factor of three, and the raw count materially overstates it. Four separate corrections are required.

Correction 1 — one speaker, three labels, 27.7% of the lane.

The pipeline has already caught four of these itself — they carry quarantine_class: SUMMARY_OF_VISSER and are third-party restatements of his position rather than his own words. A fifth quarantined claim carries quarantine_class: NOT_VISSER and is retained as an independent claim. Visser also carries the highest skill weight in the lane (2.0 against 0.8-1.0 for everyone else), so on a skill-weighted basis his share is roughly 43%. We count him once as a voice, and we note that his own record on this name is publicly mixed by his own account — see below.

Correction 2 — five claims filed on one day are one claim. Five entries dated 2026-01-05 on the jensen_huang channel, with convictions of 80, 84, 82, 85 and 80, say the same thing in five near-identical phrasings: that leading enterprise platforms are being accelerated by a named chip vendor and that agentic systems are becoming the enterprise user interface, replacing spreadsheets and command lines. Palantir appears in a five-to-six-name entity list in each. This is a transcript-chunking artefact, not five independent observations. We count it once, as a single basket claim, and it is not name-level conviction about Palantir.

Correction 3 — one claim is the chief executive's own voice and is removed from the conviction pool. Dated 2026-03-12, channel a16z, speaker Alex Karp, speaker_role: management: "From an investment perspective the last company standing before we all salute the overlord of the LLM will be Palantir — because of its specificity, security and orchestration." Quoted at half weight and excluded from the conviction pool per the standing rule that took the LLY lane apart.

Correction 4 — two claims are promoter-class and weighted low. Both from Anthony Pompliano's own voice (2025-10-14 and 2025-11-18), one arguing Palantir "deserves a higher valuation multiple" and one that Palantir and Tesla "are two great stocks... worth holding long-term in a concentrated portfolio." Weighted low with the frame disclosed, per standing rule.

After all four corrections the lane still contains roughly 40 independent claims from more than fifteen distinct sources, and — uniquely in this batch — a genuine bear side. That is what makes it valuable.

The bull case, from four independent sources describing the same moat in four vocabularies:

> 2025-09-09 · bullish · conviction 90 · thesis · channel all_in · no named speaker

> "The civil-liberties architecture built into Palantir (ontology, immutable logs, serialization/deserialization, branching) is the exact scaffolding every enterprise needs to deploy LLMs safely."

> 2025-09-09 · bullish · conviction 85 · thesis · channel all_in · no named speaker

> "Palantir has a durable moat; Foundry/ontology/FDE model is uncopied after 20 years and rivals fail when they try to build it."

> 2025-10-24 · bullish · conviction 75 · principle · channel empire · no named speaker

> "Palantir's forward-deployed-engineer model looks service-heavy but makes the product un-rippable and extremely sticky; now spreading to stablecoins/AI."

> 2026-03-10 · bullish · conviction 80 · thesis · channel invest_like_the_best · no named speaker

> "Palantir's value compounds from its culture: 'quantum org structure' + forward-deployed engineering that back-propagates what the product should be, making it a founder's factory."

> 2026-03-01 · bullish · conviction 66 · thesis · channel jordi_visser · speaker_role independent

> "Loves Palantir because enterprise AI failures stem from poor data quality/scale, not model capability; organizing that data is very hard and Palantir solves it."

Four different sources, four different framings, one identical claim: the hard part is the data layer, and it has not been copied. The commercial contribution margin rising from 64% to 78% in a year is the quantitative counterpart, and neither source appears to have had that number when they said it.

The bear case, and it is specific and well-sourced:

> 2026-04-22 · BEARISH · conviction 70 · thesis · channel anthony_pompliano_show · speaker Jim Bianco · skill 0.8

> "I would short Palantir here — huge run (drove ~90% of the mid-cap index return for a year), now not even in it and on the way down as the software basket unwinds."

> 2026-01-09 · BEARISH · conviction 52 · thesis · channel jacob_shapiro · speaker_role independent

> "Won't pay up for AI growth in poster-child names like Nvidia or Palantir; when the capital rush slows, anything trading at 30-40-50 times sales will really come off."

> 2025-11-14 · BEARISH · conviction 70 · thesis · channel forward_guidance · no named speaker

> "MAG 7 shifting from free-cash-flow machines to debt-funded AI capex; CDS/valuations must re-rate for new default risk (Palantir ~250x fwd PE, 87x P/S)."

> 2026-03-20 · BEARISH · conviction 55 · thesis · channel compound_and_friends · no named speaker

> "High-beta names at 100x+ earnings run to 200x before correcting; expensive momentum eventually has to reset."

> 2025-12-20 · NEUTRAL · conviction 55 · thesis · channel jordi_visser · speaker jordi_visser

> "Palantir is no longer the easy trade — three years of earnings already built in; wouldn't short it, but the easy money and 10-baggers are gone."

Every bear claim in this lane is a valuation claim. Not one is a business claim. That is an unusually clean split and it maps exactly onto our own conclusion: the disagreement is about price, not about whether the company works.

The lane's most useful single item is a candid self-assessment. On 2026-04-11, Visser recorded that "Palantir is not a SaaS company and won't be disrupted by AI the way the market is pricing; admits he's been wrong on it but still thinks it does well." A voice that logs its own errors is worth more than one that does not, and it is the reason we retain his claims as colour after de-duplicating them to a single voice.

Also worth recording: the lane contains a dated, checkable prediction that has now resolved. On 2026-02-08 the same voice recorded buying Palantir "after 35% drop toward $125-130; earning over $1/share this year becomes a gimme with 130% commercial revenue growth." The stock closed at $125.65 the day before this dive and $162.61 on it; trailing diluted EPS is $1.18 and FY2026 consensus is $1.495. Both halves of that claim have now landed. It is one voice, it is six months old, and it is a hit.

Text matches. Five claims mentioned Palantir in text without an entity tag; all five were sector-level software or defence-basket claims already represented in the entity lane, and none was a homograph collision — "Palantir" is a distinctive token with no ordinary-language use, which is why the discard count here is zero against 26 for Chevron and a much larger figure for names like Citigroup. That is a clean check worth stating.

Conclusion. The knowledge base has a genuinely deep, genuinely two-sided, positively-signed view on Palantir. The bulls describe a specific and repeatedly-corroborated moat; the bears object exclusively to price. We use the lane to support the growth and exponential scores and to inform 6c. We do not let it move the fair value, because after correction it says what our own arithmetic says: excellent company, unresolved multiple.

8. Data integrity — what we rejected from the vendor file and why

Eleven findings. The Palantir file has one serious share-class error, one serious net-debt error, and an unusually large number of internally contradictory ratio fields.

1. Market capitalisation counts Class A shares only — CORRECTED, +4.4%. marketCap of $373,363,942,700 at $162.61 implies 2,295,946,000 shares. The 10-Q filed 2026-08-04 states 2,300,516,590 Class A, 101,375,151 Class B and 1,005,000 Class F shares outstanding at 2026-06-30, totalling 2,402,896,741, and states that all three classes "have the same rights, except with respect to voting and conversion rights." The vendor reports a market capitalisation of $373.36B; the filing supports $390.73B; we use $390.73B. This is the confirmed share-class defect class.

2. netDebt omits $5.753B of short-term investments and is six months stale — REJECTED and rebuilt. The vendor reports −$1,194.458M at 2025-12-31, netting an operating-lease liability of $229.338M against cash of $1,423.796M while ignoring $5,753.247M of short-term investments on the same balance sheet. The 10-Q shows $2,030.047M of cash plus $7,379.052M of marketable securities at 2026-06-30 — $9,409.099M — and the earnings release states "$9.2 billion of cash, cash equivalents, and short-term U.S. Treasury securities." We use $9.409B. Magnitude of the error: $8.2B.

3. totalDebt of $229.338M is an operating-lease liability; Palantir has NO debt — all leverage ratios rejected. The 10-Q balance sheet has no borrowings line of any kind. debtToEquityRatioTTM (0.0216), debtToAssetsRatioTTM (0.0181) and debtToMarketCapTTM (0.00057) are measuring leases. interestCoverageRatioTTM and debtServiceCoverageRatioTTM both read 0, which is an undefined ratio with a zero denominator, not poor coverage — this is the same mechanism that produced a spurious low rating on other names in this process, and it is called out rather than silently accepted.

4. enterpriseValueTTM is $9.78B too low — REJECTED and rebuilt. The vendor's $371.545B is the understated market cap less an understated net cash figure. Corrected: $390.735B − $9.409B = $381.326B. Consequential rejections: evToSalesTTM (60.36x → 61.94x), evToOperatingCashFlowTTM, evToFreeCashFlowTTM, evToEBITDATTM.

5. ratios_ttm contains THREE mutually inconsistent net-margin fields. netProfitMarginTTM reads 0.49005, continuousOperationsProfitMarginTTM reads 0.31869, and bottomLineProfitMarginTTM reads 0.17613 — for the same company, the same period, in the same block. We computed it from the quarterly series: trailing net income attributable of $3,016.692M on trailing revenue of $6,155.941M is 49.0%, so netProfitMarginTTM is the correct field and the other two are rejected. Note that bottomLineProfitMarginTTM is wrong by a factor of 2.8.

6. ratios_ttm contains TWO mutually inconsistent operating-margin fields. operatingProfitMarginTTM reads 0.42799 and ebitMarginTTM reads 0.32334. Computed: trailing operating income of $2,634.652M on $6,155.941M is 42.8%, so operatingProfitMarginTTM is correct and ebitMarginTTM is rejected. Similarly interestBurdenTTM (1.5433) and ebtPerEbitTTM (1.1659) purport to measure the same relationship and disagree.

7. est.ebitAvg, est.ebitdaAvg and est.sgaExpenseAvg are broken in every forward year — REJECTED in full. FY2026 netIncomeAvg of $3.449B exceeds ebitdaAvg of $1.873B — the arithmetically-impossible signature the data contract flags — and the same inversion holds in FY2027 ($4.829B vs $2.755B), FY2028 ($7.307B vs $3.998B) and FY2029 ($15.519B vs $7.449B). FY2026 ebitAvg of $1.701B implies a 21.7% operating margin against 47.1% actually delivered. sgaExpenseAvg of $4.628B is 59% of guided revenue against a 31.6% actual ratio. The epsAvg and netIncomeAvg rows ARE internally consistent (implying a ~2.3B share count) and are the only estimate fields used here.

8. The FY2029 and FY2030 estimate rows are internally implausible — EXCLUDED. Revenue growth reads +45.2% (FY2028), +86.3% (FY2029), +120.4% (FY2030) — a growth rate that decelerates and then accelerates twice, at increasing scale, on 14 and 8 revenue analysts and 3 EPS analysts respectively. No bottom-up model produces that shape. Neither row informs any conclusion in this dive.

9. seg_prod FY2021 is off by a factor of ten — flagged, and the recent years are clean. The FY2021 entry reads "Government Operating Segment": 8,973,560,000 against total FY2021 revenue of $1,541.889M. The correct figure is $897.356M, and Commercial's $644.533M plus $897.356M equals $1,541.889M exactly. A single misplaced decimal. The FY2025 entry (Commercial $2,073.159M + Government $2,402.287M = $4,475.446M) ties to reported revenue to the dollar, as do FY2020, FY2022, FY2023 and FY2024. We use the recent years and discard FY2021.

10. seg_geo FY2025 contains a phantom total line — flagged. The FY2025 entry reads {"Rest Of World": 728,005,000, "UNITED KINGDOM": 427,398,000, "UNITED STATES": 3,320,043,000, "Geographic Concentration Risk": 4,475,446,000}. The first three sum to $4,475.446M, which ties exactly to reported revenue; the fourth line is the same total again, tagged as a concentration-risk disclosure rather than a geography. Summing all four double-counts revenue. No other year carries the phantom line. We use the three real lines: U.S. 74.2%, U.K. 9.6%, rest of world 16.3% for FY2025 — noting the 10-Q's more recent quarterly split of 81% U.S. / 19% rest of world.

11. quote.yearHigh/yearLow disagree with tech.hi52/lo52 — we use tech. The quote block reports $207.52 / $106.37; the computed block reports $207.18 / $107.27. Discrepancies of 0.2% and 0.8%. We use tech throughout for internal consistency with the moving averages and drawdown figures. Note also tech.last of $162.66 against quote.price of $162.61 — a 3-basis-point difference we resolve in favour of the quote for the price and tech for everything computed from the close series.

Not defects, correctly reported and independently confirmed: the entire quarterly income series matches the 10-Q and the earnings release line for line — Q2 2026 revenue of $1,935,464 thousand, income from operations of $912,004 thousand at a 47% margin, and net income attributable to common stockholders of $1,061,890 thousand at a 55% margin all appear verbatim in the Exhibit 99.1 financial summary; stockBasedCompensationToRevenueTTM of 13.57% reconciles to the 10-Q's $265,209 thousand quarterly figure; grossProfitMarginTTM of 84.80% reconciles to the computed series; capexToRevenueTTM of 0.68% is consistent with the $33.9M FY2025 capital expenditure; and the FY2025 through FY2020 seg_prod and seg_geo blocks tie to reported revenue.

Vendor rating — accepted, unusually. The composite reads B+ / 3, with priceToEarningsScore of 1 and priceToBookScore of 1 dragging it down against returnOnEquityScore and returnOnAssetsScore of 5. Unlike the cases where a low sub-score traced to an undefined input, both of these are correct and economically meaningful here: Palantir genuinely does trade at 138x earnings and 39.9x book. We accept the rating's mechanism and note that the same file gives debtToEquityScore only a 4 for a company with literally no debt, which is the one artefact.

Non-equity tripwire — checked and passed. PLTR is common stock, NASDAQ-listed, three classes with identical economic rights per the 10-Q. Beta 1.563; a 52-week range of $107.27 to $207.18 (a 93% spread); no dividend; volume of 172.0M shares (roughly $28.0B of turnover). This is common equity, and the multi-class structure is a voting arrangement, not an economic one.

9. Technicals

• Price $162.61. −21.5% from the 52-week high of $207.18; +51.6% above the 52-week low of $107.27. Position within the annual range: 55th percentile.
• Moving averages, both cleared today: +24.3% above a 50-day average of $130.99; +6.6% above a 200-day average of $152.54. The 200-day was reclaimed on this session's gap, having been broken to the downside for much of the preceding quarter.
• RSI 69.2 — approaching overbought after a single session's move; MACD +1.52, freshly positive.
• Maximum drawdown from peak over the trailing year: −21.5% at today's close, but the trough of that drawdown was −48.2% at $107.27. This is the important number in the whole technical section.
• Relative performance, and it inverts everything: 3-month +11.39% vs SPY +7.59% and QQQ +7.67%; 6-month +10.08% vs SPY +11.09% and QQQ +15.71%; 12-month +5.44% vs SPY +24.26% and QQQ +30.80%. Over twelve months, during which quarterly revenue grew from roughly $1.0B to $1.9B and quarterly operating income grew from $269M to $912M, Palantir underperformed the S&P 500 by 19 points — and that is AFTER a 29.42% day. Before today it had underperformed by roughly 42 points.
• Sentiment: 12 buy, 11 hold, 3 sell across 26 analysts — the most divided distribution in this batch. Consensus target $172.17 (+5.9%), median $195 (+19.9%), high $215 (+32.2%), low $80 — 50.8% BELOW spot. A $135 spread between high and low target on a $162 stock is a 83% dispersion, which is itself a statement: the sell-side does not know what this is worth either. And every one of these was struck before the 2026-08-03 print.

Today's move and what it does to the entry

PLTR closed 2026-08-04 at $162.61, up 29.42% or $36.96 from a $125.65 previous close. It opened at $145.145 — a 15.5% gap — traded a $143.28 to $164.51 range, and closed within 1.2% of the day's high on 172,005,673 shares against a 40,856,136 average. That is 4.21x normal volume and roughly $28 billion of turnover in one session.

This was a company event, not a sector event. Second-quarter results were released after the close on 2026-08-03 with revenue 6.8% above consensus, EPS 18.9% above, and full-year guidance raised above the consensus figure. The move is a rational repricing of a genuine information event, and we say so rather than dismissing it as momentum.

But it changes the entry completely. Yesterday's $125.65 was 36.1x guided FY2026 revenue and our base case would have been +33.7%. Today's $162.61 is 46.8x and the base case is +3.3%. The entire margin of safety was consumed in one session, and it was consumed by information that was genuinely good. That is the most common way a good idea becomes a bad trade.

What we are telling the reader plainly: this is a business we want to own and a price at which we will not start. The two setups that would change it are the falsifiers below — a pullback toward the 200-day average at $152.54 or below, or a Q3 print that extends the acceleration series and forces consensus up faster than the price.

10. Insiders — a director selling in tranches, and a conversion

The reading, stated carefully. Director Alexander Moore sold 16,000 shares across four tranches on 2026-07-15 at $133.75 to $136.79, roughly $2.15M in total, twenty days before a print that took the stock to $162.61. He retained 1,095,095 shares, worth $178.1M at today's close — so he sold 1.4% of his position. This is a scaled, price-laddered sale of a rounding error, not a signal, and the tranche structure (11,800 / 3,100 / 1,000 / 100 at rising prices) is the signature of a limit-order ladder rather than a decision.

Jeffrey Buckley's 17,124 stock appreciation rights at a $135 strike are a grant, not a purchase. Shyam Sankar's C-Conversion is a Class B to Class A conversion, an administrative event.

What the file contains is therefore: zero open-market purchases, and one director trimming 1.4% of a position. That is neutral, and it is reported as neutral. The absence of any executive selling into a 29% move is not yet observable — the window is only one day old — and the next Form 4 cycle will be genuinely informative in a way this one is not.

11. Verdict, kill-criteria and flip conditions

Watch.

What is genuinely exceptional, and every figure is filing-verified in a 10-Q filed the morning of this dive: revenue of $1.935B in the June quarter, up 92.8% year on year and the fourth consecutive quarter of accelerating growth (48.0%, 62.8%, 70.0%, 84.7%, 92.8%); U.S. revenue up 114.7% and U.S. commercial up 149%; a 47.1% GAAP operating margin against 26.8% a year earlier, and 62% adjusted; 84.7% gross margin; 63% operating-cash-flow margin; commercial contribution margin of 78%, now above government's 71% and up from 64%; stock-based compensation falling as a share of revenue from 15.9% to 13.7%; diluted share count up only 0.3% year on year; $9.409B of cash and marketable securities against zero debt of any kind; and full-year guidance raised above consensus to $8.150-8.158B of revenue, $4.889-4.897B of adjusted operating income and $4.5-4.7B of adjusted free cash flow.

What we are declining to pay for: 62x trailing sales, 138x trailing GAAP earnings — roughly 173x once the 1.43% effective tax rate is normalised — and 46.8x guided current-year revenue, on a day the stock rose 29.42% on 4.21x normal volume. A base fair value of $168 that is 3.3% above spot and 2.4% below a street consensus every element of which predates the print. And a demonstrated history, on this exact security within the last twelve months, of a 48% drawdown from $207.18 to $107.27 while the business was accelerating — proof that the multiple can halve without the fundamentals cooperating.

The distinction that matters. We are not saying Palantir is overvalued in the sense that the business will disappoint. We are saying that after a 29% single-session repricing, the entire margin of safety has been transferred from the buyer to the seller, and the one variable that decides the outcome — the terminal revenue multiple — is one we have no edge on. In the Synthos frame, an exceptional business at a fair-value-equals-price entry, on the day of its largest move, is a Watch by construction. The knowledge base agrees from an unusual direction: every bearish claim in a lane of 83 is a valuation claim and not one is a business claim.

Pre-registered KILL criteria — what would take this to Avoid:

• Any sequential revenue growth below roughly 12% (against 18.6%, 16.0%, 19.1% and 17.7% in the last four quarters). That ends the acceleration series, which is the entire basis of the multiple.
• Q3 revenue below the guided $2.160B, i.e. a miss against the company's own number in the quarter immediately following a guidance raise.
• Commercial contribution margin falling back below government's. The crossover to 78% versus 71% is the quantitative evidence for platform generalisation; its reversal would falsify it.
• Stock-based compensation reverting above 16% of revenue, undoing the convergence between GAAP and adjusted margins.
• A valuation-allowance release presented as operating earnings, or any material contract termination-for-convenience disclosed against the $4.9B of remaining performance obligations.
• International revenue growth falling below 20%, which would confirm the story is one country and cap the terminal size.

Pre-registered FLIP conditions — what would take this to Buy — Tactical:

• Price below approximately $135 — roughly 25.8x our FY2027 revenue estimate and near the 50-day average — with the guide intact. At that price the base case is +24% and the asymmetry becomes genuinely compelling. This is 17% below today's close and inside the range this stock has traversed in a week.
• A hold of the 200-day average at $152.54 on any consolidation, combined with FY2027 consensus revenue revising above $12B. That would confirm the sell-side has incorporated the guide and that the base case is conservative.
• The 2026-11-02 print above the guided $2.164B with U.S. commercial growth still above 120% — a fifth consecutive quarter of acceleration, at which point paying a high multiple is defensible because the deceleration risk has been deferred another year.
• International revenue growth inflecting above 60%. Nineteen percent of the company growing at 33.7% is the single largest untapped source of upside in the file, and it would materially raise the terminal revenue estimate on which the entire fair value rests.
• A clean disclosure of contracted, non-cancellable backlog replacing the option-inclusive TCV and RDV figures, which would remove the largest qualitative discount we apply.

Where PLTR fits in the Synthos Framework Portfolio. The enterprise-AI / software sleeve, at 0% today with a 2-3% target on a fill near $135 or a confirmed November print. On batch overlap, this is the highest-beta expression of the artificial-intelligence adoption theme in the twelve names here — and it is worth noting explicitly that it sits alongside Arista (the network fabric of the same data centres), Dell (the servers), Texas Instruments (the analogue content) and Chevron (which just signed a 20-year agreement to power one). Those five are not five ideas; they are one idea with five risk profiles, and Palantir is the one with the most upside and the least protection. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $162.61.

Single biggest risk: the multiple, not the business. Between its 52-week high of $207.18 and its low of $107.27 this stock fell 48.2% while quarterly revenue grew from roughly $1.0B to $1.9B and quarterly operating income more than tripled. The market has already demonstrated, on this security, within this data window, that it will compress this multiple by half without any help from the fundamentals. At 62x trailing sales and 173x tax-normalised earnings there is nothing underneath the price — no dividend, no buyback of consequence, no asset value, no free-cash-flow yield above 1.2%. A 10% revenue disappointment does not produce a 10% decline; it produces a de-rating on top of it, which is what the $96 bear case is.

Most fragile assumption in the price: that FY2027 revenue growth of roughly 55% is achievable and that the market will pay more than 30x sales for it two years out. Both halves must hold. The first we think is more likely than consensus does — the FY2027 estimate in this file implies only 41.6% growth off a base the company guided up the day before, and we put the number nearer $12.6B. The second we have no edge on at all, and it is worth $153 of share price between 20x and 48x on identical revenue. If revenue lands where we think and the multiple lands at 20x, the stock is $109 and every operational judgement in this dive was correct. That is the specific way an investor loses 33% here while being right about the company.

Provenance & disclosures

• Traceability: 83 tagged knowledge-base claims name Palantir by entity (raw entity hits 83, text matches 5, discarded for homograph collision 0 — "Palantir" is a distinctive token with no ordinary-language use, and the zero discard count is a clean check worth recording). Four attribution corrections applied and disclosed: (1) 23 claims — 27.7% of the lane, ~43% skill-weighted — resolve to Jordi Visser across three labels (12 channel jordi_visser with no named speaker, 7 speaker jordi_visser, and 4 the pipeline itself flagged quarantine_class: SUMMARY_OF_VISSER), counted once as a voice; (2) five claims dated 2026-01-05 on the jensen_huang channel are near-verbatim restatements of one basket assertion and are counted once; (3) one claim dated 2026-03-12 carries speaker_role: management — Alex Karp — and is excluded from the conviction pool and quoted at half weight; (4) two claims in Anthony Pompliano's own voice are weighted low as promoter-class per standing rule. After correction the lane holds roughly 40 independent claims from more than fifteen sources, with named speakers including Cathie Wood, Dan Ives, Jim Bianco, David Sacks, Jeff Park and Alex Karp. Raw stance: 60 bullish, 12 neutral, 11 bearish — and every bearish claim is a valuation claim, none is a business claim, which is itself a finding. Breadth 15, net conviction positive-high. One claim in the lane discussing model-layer competitive dynamics is paraphrased rather than quoted, because it names specific artificial-intelligence developers. All other quotes are verbatim from stored claim text.
• Data as-of: income statement, balance sheet and cash flow for the quarter and six months ended 2026-06-30, from the Form 10-Q filed 2026-08-04 — the day of this dive — and the Exhibit 99.1 earnings release filed 2026-08-03; fiscal 2025 from the 10-K filed 2026-02-17 · estimates 2026-08-04 (and materially stale — see below) · prices 2026-08-04, quote timestamp 1785873600 = 2026-08-04T20:00:00Z ($162.61, +29.42% on 172.0M shares; 50-DMA $130.99; 200-DMA $152.54; RSI 69.2; MACD +1.52) · knowledge-base claims 2026-08-04. Palantir's fiscal year is the calendar year. All figures come from the Synthos vendor data file for PLTR or from the SEC filings in the PLTR archive; no figure comes from memory, recall or external retrieval.
• Estimate staleness — stated prominently because it changes the arithmetic. Second-quarter results were released after the close on 2026-08-03. Every analyst estimate and every price target in this file was struck before that release. The evidence is direct: the vendor's Q3 2026 revenue consensus of $2.148B sits BELOW the company's own guidance of $2.160-2.164B, and the FY2026 consensus of $7.851B sits 3.9% below guidance of $8.150-8.158B. The FY2027 consensus of $11.545B implies 41.6% growth off a base that has just been guided to 82% growth. Our fair values use our own FY2027 revenue estimate of $12.6B for this reason, labelled as ours throughout.
• Filing archive contents: 10-K filed 2026-02-17 (fiscal 2025, tables preserved); 10-Q filed 2026-08-04 (June 2026 quarter — the freshest filing in this entire batch); 10-Q filed 2026-05-05 (March 2026 quarter); 8-K filed 2026-05-04 (Q1 results); 8-K filed 2026-06-09; 8-K filed 2026-08-03 carrying the full Exhibit 99.1 second-quarter release with the financial summary, guidance and operating metrics. Tables were available on every document.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): market capitalisation corrected from $373.36B to $390.73B because the vendor counts Class A shares only against the 10-Q's 2,300,516,590 Class A + 101,375,151 Class B + 1,005,000 Class F, all of which the filing states carry identical economic rights; net cash corrected from the vendor's implied −$1.194B of "net debt" to $9.409B of cash and marketable securities, the vendor having omitted $5.753B of short-term investments from a six-month-stale balance sheet; Palantir confirmed to have ZERO debt of any kind, so the vendor's $229.338M totalDebt is an operating-lease liability and every derived leverage and coverage ratio is rejected; enterprise value rebuilt from $371.545B to $381.326B; the FY2021 seg_prod government line corrected from $8,973,560,000 to $897.356M (a factor-of-ten error, confirmed by the segment sum tying to reported revenue only after correction); the FY2025 seg_geo phantom "Geographic Concentration Risk" line identified as a duplicate of the total and excluded from the sum; est.ebitAvg, ebitdaAvg and sgaExpenseAvg rejected in every forward year on the net-income-above-EBITDA signature; three mutually inconsistent net-margin fields and two mutually inconsistent operating-margin fields in ratios_ttm resolved against the computed quarterly series; and quote.yearHigh/yearLow rejected in favour of tech.hi52/lo52. Where vendor and filing AGREED — worth recording: the entire quarterly income series matches the 10-Q and Exhibit 99.1 to the thousand, including Q2 2026 revenue of $1,935,464, income from operations of $912,004 and net income attributable to common stockholders of $1,061,890; stockBasedCompensationToRevenueTTM reconciles to the 10-Q's $265,209 quarterly figure; and the FY2020, FY2022, FY2023, FY2024 and FY2025 segment tables all tie to reported revenue.
• Tax note — the most consequential unmodelled item in this file. The trailing effective tax rate is 1.43% ($15.383M of provision on $1,081.345M of pre-tax income in the June quarter). The 10-Q states that a full valuation allowance is maintained on U.S. and U.K. deferred tax assets and that "given the Company's recent earnings and anticipated future earnings, there is a reasonable possibility that it will have sufficient positive evidence in the future to release all or a portion of the valuation allowance." Two consequences are stated in the body: trailing EPS normalised at a 21% rate is roughly $0.94 rather than $1.18, making the trailing multiple ~173x rather than 138x; and a future valuation-allowance release will produce a large one-off, non-cash GAAP earnings benefit that must not be capitalised as run-rate. This dive pre-registers that.
• Basis note: the earnings release reports GAAP EPS of $0.41 and adjusted EPS of $0.41 for the June quarter — identical, because the adjusted-net-income figure ($1,047.001M) is slightly below GAAP ($1,061.890M) and the diluted share counts differ. The GAAP/non-GAAP gap on Palantir's earnings line is therefore essentially nil, which is unusual and creditable. The gap on the operating line is not nil: GAAP income from operations of $912.004M (47%) against adjusted income from operations of $1,194.472M (62%), a $282.468M difference that is stock-based compensation and related employer payroll taxes. All GAAP figures in this dive are labelled GAAP; all adjusted figures are labelled adjusted; guidance is on the adjusted basis and is labelled as such.
• Estimate coverage: 23 analysts on FY2026 revenue and 20 on FY2026 EPS; 24 and 20 on FY2027; 17 and 9 on FY2028. FY2029 and FY2030 rest on 3 EPS analysts each and are excluded entirely, as is their internally implausible growth shape (+45.2%, +86.3%, +120.4%). est.ebitAvg and est.ebitdaAvg are rejected in full for the net-income-above-EBITDA signature; only epsAvg and netIncomeAvg are used.
• Peer note: the vendor peer set — Adobe, AMD, ASML, Salesforce, CrowdStrike, Cisco, Oracle, Palo Alto Networks, UiPath, SAP — is scale-matched but analytically weak; only UiPath and arguably Salesforce address a comparable problem, and one knowledge-base claim makes precisely this objection about software index composition. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. Palantir's corrected $390.7B market capitalisation is now within 7% of Oracle's $419.7B and 2.5x Salesforce's $156.4B.
• Fair-value caveat: the $96 / $168 / $262 anchors are enterprise-value-to-sales multiples — 17.4x, 31.1x and 48.6x — of our own FY2027 revenue estimate of $12.6B, which is 9% above the $11.545B consensus because that consensus predates the 2026-08-03 guidance raise. The revenue figure is ours and is labelled as ours; the multiples are ours. Cross-checked against FY2027 and FY2028 consensus EPS at every anchor. Full sensitivity disclosed in the body: on the same revenue estimate, 20x gives $109, 25x gives $135, 31x gives $168, 40x gives $214 and 48x gives $262. The entire answer is one input, and no greater precision is claimed.
• Timing: second-quarter 2026 results were released after the close on 2026-08-03, one day before this dive, and the Form 10-Q was filed 2026-08-04, the day of this dive. The stock closed +29.42% on 4.21x average volume. The next print is 2026-11-02, 90 days away, with company guidance of $2.160-2.164B against a vendor consensus of $2.148B. The verdict is set with the post-gap entry price as an explicit input, and the dive states plainly that yesterday's $125.65 would have produced a Buy — Tactical on the same analysis.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.