Conviction: Medium · Status: Research/analysis only — advisory, no trade Editorial frame: the contrarian short case (Michael Burry's publicly disclosed 13F stance) against the AIP growth case — resolved this evening by the Q2 print, in the growth case's favor

Basis note. Trend, RSI and moving-average figures are computed on the Monday 2026-08-03 settled close ($125.65). Q2 results printed after that close (5:00 PM ET); reported figures come from the company's press release and are labeled as such. The ~+12% after-hours move is an indication, not a settled price — every confirmation test in this file is written against settled closes.

---

The Story Right Now

For ten weeks this profile said the same thing in seven different ways: the tape is broken, the multiple is still large, and only the Q2 print resolves it. Tonight it resolved — emphatically, and toward the re-rate.

Q2 2026, reported after the close: revenue $1.935B, +93% year over year — against a consensus of ~$1.812B and management's own $1.799B guidance midpoint. The line this profile called the thesis carrier didn't clear its bar so much as vaporize it: U.S. commercial revenue $764M, +149% against a consensus looking for roughly +98% — while U.S. government itself accelerated to +90% ($809M), and total U.S. revenue reached $1.573B, +115%. GAAP net income printed $1.062B — a 54.9% net margin — at diluted EPS of $0.41 on the filed measure. And the guidance: full-year revenue raised to $8.150–8.158B (+82% over FY2025), U.S. commercial guidance to more than $3.424B (+134%), adjusted operating income to $4.889–4.897B, adjusted free cash flow to $4.5–4.7B. Every metric raised; every leg of the "beat with no raise" disappointment scenario refuted at once.

The market's first answer: shares reported ~+12% in after-hours trading — inside the 9–15% band the options market had priced going in, and enough, if it survives to a settled close (~$141 on the indicated move), to clear the 20-day ($128.71) and the 50-day ($130.49) in one session, leaving only the 200-day ($152.62, about 8% above the indication) of the overhead stack this profile has tracked since June.

The number that reframes the whole argument: on tonight's reported quarter, trailing-twelve-month revenue is now $6.155B, and at the settled close's $300.8B market capitalization that is 48.9x trailing sales — down from 112x at the November peak — for a company whose growth just accelerated from +81% expected to +93% delivered. The deceleration thesis that drove this stock from $208 to a death-crossed $126 spent the whole decline being priced as if the growth were fading. It wasn't. That was the entire disagreement, and it is now settled on the bulls' terms — at the fundamentals layer. The tape layer still has to do its work, which is where the discipline below comes in.

The scoreboard: what this profile predicted, and what happened

Written this morning, before the print — graded tonight. The bar was set precisely because consensus sat only +0.72% above management's own guidance midpoint, so an in-line quarter would beat guidance and still miss the street:

The one prediction this profile deliberately refused to make was direction — a 9–15% implied move on eight straight beats was called "a coin flip with a wide payoff table." The coin came up heads at +12%, and the refusal stands as written: the tradeable read was never the print, it's the base that forms after it.

What changed in the desk's own view today

Two corrections this file carries as part of its record, both made before the print:

Free cash flow never turned negative. Since May this file's bear case led with "FCF turned negative (−$181M), Rule-of-40 story breaks." That was false — a mis-derived figure that treated Treasury portfolio rolls as capital spending. Filed truth: FY2025 free cash flow +$2.10B (operating cash flow $2.13B against $33.9M of capex) and Q1 2026 +$892M. Corrected against SEC filings this afternoon (defect D037). Tonight's print then raised adjusted FCF guidance to $4.5–4.7B — the cash-quality leg of the bear case is not just corrected, it is inverted.

This morning's "news" was six weeks old. The NGC2 data-layer role and the Zeta Global partnership both surfaced today wearing today's date; both are June announcements (June 22 and 23). Neither explained the pre-print bid, and neither counts as a print-day catalyst.

The company underneath the print — what the 10-K says they're doing, and say they'll do

Source: the FY2025 10-K (filed 2026-02-17), Item 1 Business and Item 1A Risk Factors — the company's own words, on file. This is the layer the tape can't show.

What they've built. Four platforms plus the thing that makes them one product: Foundry (the data-operations platform — data management, logic authoring, analytics, workflow), AIP (the generative-AI layer — secure connectivity to third-party LLMs, an agent-building toolchain, and an evaluations framework for governing AI workflows in production), Apollo (continuous delivery — run the stack in virtually any environment, cloud to classified edge), and Gotham (the defense/intel product, "operations centers to the tactical edge"). The heart is the Ontology — mapping an organization's data, logic, and actions into one decision representation. The strategic claim embedded in that architecture: AIP isn't a chatbot bolted onto a database, it's LLMs wired into an operational model of the enterprise with access controls that propagate from source data to shared analyses. That is the moat argument in one sentence, and it's why the +149% U.S. commercial print reads as architecture winning, not marketing.

How they sell. The 10-K names the go-to-market that produced the commercial acceleration: AIP bootcamps — "deliver real workflows on actual customer data in days" — plus a Developer Tier (limited free-ish access to Foundry/AIP) to seed adoption, and the embed-with-the-customer forward-deployed model. The filing's own framing of demand: recent crises taught customers that months-long software implementations "are not a viable option" — they buy what works in days. Q2's numbers are this motion compounding: 954 customers at 2025-12-31 (the catalyst table below already tracks the 1,000 crossing).

What they say they'll do next. Three stated growth vectors, verbatim intent from the filing: (1) "capture an even greater share of U.S. federal government spending on software systems" — the June NGC2 data-layer award is that sentence turning into backlog, and Q2's +90% government growth says the intent is executing; (2) industry operating systems via partnerships — airline, space, shipbuilding, insurance, healthcare, telecommunications, automotive, security — distribution through partners rather than a giant direct sales force; (3) channel/cloud alliances and joint ventures (Palantir Japan KK is the filed example) to reach markets they won't build alone.

What the filing warns about — the risks the desk actually prices. Top of the 10-K's own list: sustaining revenue growth (the exact thing Q2 just answered, for one quarter); long, unpredictable sales cycles and expensive sales efforts; customer concentration — the top three customers were 16% of FY2025 revenue (improving from 17%, but a single renewal still moves a quarter); realizing full contract deal-value; expense growth against profitability; quarterly lumpiness and seasonality; execution risk on new AI technology; and — a filed risk factor worth savoring given the social layer below — that "news or social media coverage or other external scrutiny of Palantir or our leadership" may harm the business. Palantir is the rare company that files its own discourse as a risk factor.

Price Data

Settled basis: 2026-08-14 close (momentum has cooled off the RSI 73 peak but the confirmation sequence still holds; pre-print row preserved below for reference).

Monday settled +2.1% on 1.69x average volume (57.7M shares vs a 34.1M average) — the heaviest tape in weeks, into the print. Since then: 08-04 settled +29.45% ($162.66) on the print itself, and the tape has continued to run through 08-07 — RSI 73.0 is now the hottest reading in this file's history, and the confirmation sequence this profile set as its own bar (settled closes above SMA20/SMA50, then a higher low above the July range) has been cleared on every count. The remaining overhead from the pre-print technical map is the 200-day (SMA200 $152.28, now +13.0% below current price — meaning price has also cleared it) and the June bounce high (~$132.38, also cleared). Only the 52-week high ($207.52) and the pre-de-rate range remain above.

2026-08-08 tape note: PLTR $172.01 (+5.72% vs $162.66 on 2026-08-04), RSI 73.0 (up from the 08-04 settled reaction, itself already elevated). 7D +39.78%, 30D +30.09%, 3M +24.83%. From the 52-week high, the gap is -23.66% (narrowed from -39.45% pre-print). The desk's news pull (fetched 2026-08-08) carries no fresh company-specific catalyst beyond continued post-print coverage (WSJ quote page, recap pieces) — this reads as the tape continuing to price in the Q2 beat and raised guidance already documented above, not a new event. No new fundamentals re-researched this cycle; the 10-Q flagged as pending in the Catalysts table below has not yet been checked against the massive/edgar caches this cycle — treat the reported figures above as press-release basis until confirmed. Per the full-scan summaries (2026-08-07 close). 2026-08-14 tape note: PLTR $174.04 (+1.18% vs $172.01), RSI 68.3 (cooled from 73.0, off the hottest read in this file's history) — still well above SMA20 (+19.6%, narrowed from +27.4%) and SMA50 (+28.9%, roughly flat vs +29.7%) and above SMA200 (+14.4%), though SMA50 still sits below SMA200 so the engine continues to tag a death cross. 7D +1.18% (moderated sharply from +39.78%), 30D +30.11% (roughly flat vs +30.09%), 3M +29.89% (accelerated from +24.83%). From the 52-week high, the gap is -13.74% (narrowed from -23.66%). No fresh company-specific catalyst on record for this window beyond the already-logged 08-03 print; the 10-Q flagged as pending in the Catalysts table below has still not been reconciled against the filings caches this cycle — treat the reported Q2 figures as press-release basis until confirmed. No new fundamentals re-researched this cycle. Per the full-scan summaries (2026-08-14 close).

Technical Levels — now a confirmation checklist

Fundamentals

Valuation on the new quarter (settled close, 2,393.9M basic shares, $300.8B market cap): 48.9x trailing-twelve-month sales ($6.155B TTM through Q2) and 108.3x trailing GAAP earnings ($1.16 TTM diluted EPS) — down from 57.6x and 141.2x this morning on the strength of the quarter alone, before any price change settles. On the indicated after-hours price those multiples run ~12% higher. Reported figures are press-release numbers; the 10-Q reconciliation follows when it files.

Basis caution: the widely-quoted $0.35 consensus EPS was adjusted (non-GAAP); the $0.41 reported here is filed-basis GAAP. Different measures — the beat is real on both, but don't mix the series.

The earnings record — eight filed quarters, and what they compound into

Series basis: Massive filed quarterly statements (calendar quarters, GAAP); growth rates and margins computed by script on that series. Tiny divergences from the 10-Q's own stated rates (e.g., +84.7% here vs +83.3% filed for Q1 2026) are year-ago-figure rounding between sources, not a disagreement. The Q2 row is press-release basis, pending the 10-Q.

What the series says, in three lines. The growth is accelerating, not just large: five consecutive quarters of rising year-over-year growth — +48% → +63% → +70% → +85% → +93% — which is the exact opposite of the deceleration the de-rate priced. The operating leverage is violent: operating income went from $11M (1.3% of revenue) in Q4 2024 to $754M (46.2%) in Q1 2026 — six quarters — while gross margin climbed 78.9% → 86.8%; net margin compounded 9% → 54%, and diluted EPS went $0.03 → $0.41 in seven quarters. The cash follows the accrual: quarterly operating cash flow nearly tripled year over year ($310M → $899M, Q1'25 → Q1'26), so the margin expansion is collected, not just booked. This is the table that explains why a 48.9x sales multiple has buyers: the model at this scale still improves on every line every quarter, and the market is pricing the trajectory, not the level.

Positioning into the print (settlement data through 2026-07-15): short interest 77.9M shares — 3.25% of basic shares, 1.83 days to cover — up from ~69–71M in May. Real but modest squeeze fuel: enough to amplify the after-hours pop, nowhere near a crowded-short structure. The +12% indication is mostly longs repricing, not shorts trapped.

Bull / Bear — rewritten on the print

Bull (now the base case at the fundamentals layer): the AIP commercial flywheel accelerated — +149% U.S. commercial against a +98% bar — while government re-accelerated to +90%; a 54.9% GAAP net margin; guidance raised on every line to +82% full-year growth; $8B of liquidity, no long-term debt, capital-light ($34M FY2025 capex); and the multiple compressed from 112x to 48.9x trailing sales while growth sped up. The de-rate was a liquidity event and a narrative error, not a fundamental read.

Bear (what honestly remains): 48.9x trailing sales and ~108x trailing earnings still price sustained hyper-growth — the argument is thinner but not gone; a +12% after-hours move can fade by the open (July's whole range was built by a forced seller whose buyer, by the desk's own whiplash report, caps rallies with redistribution flow); the settled tape is still death-crossed below its 200-day until it isn't; and one quarter, however loud, is one quarter.

Thesis killers, updated: the after-hours move fully round-trips and the range floor (~$118) breaks anyway — that would say the market no longer pays for even accelerating growth, a regime statement bigger than this name; commercial growth halving toward +75% with guidance held next quarter; a Databricks listing above ~$300B resetting the comp.

What the street and the socials are saying

Print-night read, 2026-08-04 morning pulls. News: the desk's news pass. Social: hosted search saved to the capture corpus. Both are sentiment context, not price truth.

The press is unanimous on the fact, split on the meaning. Reuters and the WSJ carry the raise straight ("lifts guidance as revenue nearly doubles"; "strong demand from US government, commercial clients"); CNBC leads with the +149% commercial number; Invezz frames the quarter as a Rule of 40 score of 155 — a figure so far outside SaaS norms it functions as the bull case's one-line summary. No major outlet in the pull carried a bear rebuttal beyond valuation — after a −40% drawdown, the skeptics' argument printed and lost.

The social layer is exuberant retail plus ambient noise, not insight. The hosted sweep surfaced earnings-night YouTube streams walking the beat line by line, and Hacker News running hot ("PLTR is the most de-risked 10x opportunity on the market" — conviction as a meme, the kind of phrasing that historically clusters near local tops, worth logging precisely because it's unfalsifiable). The recurring class-action/"securities fraud investigation" press releases in the sweep are law-firm ambulance-chasing boilerplate that follows every large-cap drawdown — standing noise on this name, not signal. Net read: retail sentiment has already flipped fully bullish on the print, which is worth knowing as positioning context — the after-hours buyer is not early. No desk-followed X account had a substantive PLTR take in the capture window; that absence is itself the note.

Catalysts

The Verdict

HOLD / research-only, Medium conviction — the argument is settled; the tape's confirmation is not. This file's ten-week refrain — "only the Q2 print resolves it" — is discharged: it resolved toward the re-rate, at the fundamentals layer, about as decisively as a single quarter can. What has not happened yet is anything on the settled tape: as of tonight's close PLTR remains death-crossed, below all three moving averages, in a collapse regime tag. The desk's doctrine is unchanged by good news — the entry is the base, not the knife and not the gap: a settled close above the 50-day, then a higher low above the July range, converts tonight's answer into a position-grade structure. Chasing the after-hours print buys the gap. The thing worth being early on was the print; the thing worth being patient on is everything after it.

Research Log

---

Post-print edition, 2026-08-03 evening · tape figures on the 2026-08-03 settled close ($125.65) · reported Q2 figures from the company press release, 10-Q pending · after-hours move is an indication, not a close · research/analysis only, no trade

Sources

• Q2 2026 reported figures and guidance: Business Wire — company press release, 2026-08-03, CNBC, StreetInsider. After-hours move as reported by those outlets.
• Tape data (price/RSI/SMA/returns), settled 2026-08-03 close: ai-infra; 52-week range: info (fetched 2026-08-03 post-close).
• Filed financials (Q1 2026 and FY2025), reconciled provider-vs-filing: income-statements, cash-flow-statements, key-financials, and SEC XBRL company facts via the desk's own tooling (10-K filed 2026-02-17; 10-Q filed 2026-05-05).
• Earnings-record series (eight quarters) and positioning: income-statements + cash-flow-statements.json (refreshed 2026-08-04; growth/margin rates script-computed on the series) and short-interest (settlements through 2026-07-15).
• Pre-print consensus, guidance midpoint, segment splits and implied-move quotes: EBC earnings preview, TradingKey preview, TipRanks implied move.
• Date verification for the June announcements (NGC2, Zeta): Business Wire 2026-06-22, Business Wire 2026-06-23.
• Company narrative, strategy, and risk factors: FY2025 10-K filed 2026-02-17 (period 2025-12-31), Item 1 Business and Item 1A Risk Factors — cached at 10k-business and 10k-risk_factors.json; customer count (954), top-three concentration (16%/17%), and all quoted strategy language from that filing. Q2 8-K (filed 2026-08-03): 8K-2026-08-03.html.
• Print-night news read: latest (fetched 2026-08-04 — Reuters, WSJ, CNBC, Invezz items as characterized).
• Social read: hosted social sweep saved to 2026-08-04-social-search-pltr (Hacker News + YouTube clusters; class-action PR noise noted, not treated as signal).
• Prior-cycle history: this file's dated research log. The pre-print edition of this profile (with the full consensus setup and the D037 correction detail) is preserved in the repo history.