Phase A — Understand the business
Lens 1 · Company Overview
Palantir builds and deploys software platforms that act as "the central operating system" for its customers' data. Incorporated in Delaware May 6, 2003; fiscal year ends Dec 31; HQ Denver, CO; co-founders Alex Karp (CEO) and Peter Thiel (Chairman).
What it actually sells (three revenue mechanics, per the 10-Q "Components of Results of Operations"):
- Palantir Cloud — hosted SaaS subscriptions with stand-ready operations-and-maintenance (O&M); recognized ratably over the term.
- On-Premises Software — term licenses run on the customer's own infrastructure, bundled with O&M as a single performance obligation; recognized ratably.
- Professional services — on-demand user support, UI configuration, training, and "ontology and data modeling" support; recognized over the term.
The product line in market terms: Gotham (government/defense intelligence), Foundry (commercial data-ops backbone), Apollo (deployment/CD layer), and AIP — the Artificial Intelligence Platform (launched 2023), which lets enterprises run LLMs/agents on their own governed data via "ontology." AIP boot camps have become the dominant land-and-expand motion for commercial.
Customers — 954 at YE2025, up from 711 at YE2024 (+34%). Two demand engines: U.S. government (defense, intelligence, civilian agencies — each sub-agency counted separately) and commercial (U.S. commercial is the breakout). Suppliers are essentially third-party cloud hosting (AWS/Azure resold inside Cloud/pilots) plus its own forward-deployed engineering labor — there is no meaningful hardware bill of materials. Competitors: in-house data teams, the hyperscalers' own data/AI stacks (Microsoft Fabric, AWS, Google), Databricks/Snowflake on the data-platform axis, and defense-software entrants (Anduril on the autonomy side, traditional primes on integration).
Contract structure / key payment terms — multi-year, often with customer options. The dollar-weighted average remaining contract duration is ~4 years. Government deals carry termination-for-convenience and option-exercise uncertainty (a real risk the company flags). Deferred revenue $516.9M current + $41.1M noncurrent, and customer deposits $370.1M at Q1-2026 — i.e. customers prepay, a working-capital positive. Concentration is government-heavy but diversifying toward commercial (see Lens 4).
Lens 2 · Supply Chain
Palantir is an asset-light software company, so the "supply chain" is a value chain of inputs → platform → end users, not a physical BOM. Named stakeholders along the chain:
- Upstream inputs:
- Compute / hosting — third-party cloud (AWS, Microsoft Azure) resold inside Palantir Cloud and consumed for pilots; this is the single largest direct cost-of-revenue input besides O&M labor.
- Talent — forward-deployed engineers (FDEs) and product/infra staff are the core "raw material"; the labor is in cost of revenue, S&M (pilots), and R&D. SBC is the non-cash form of this input.
- Foundation models — for AIP, Palantir is model-agnostic, orchestrating third-party LLMs (OpenAI, Anthropic, Meta Llama, etc.) on top of its ontology rather than training frontier models itself. This is a strategic choice: it rides the model-price deflation curve instead of paying for it.
- The platform (Palantir): Gotham / Foundry / Apollo / AIP — the integration + governance + deployment layer.
- Downstream / end customers (named):
- U.S. government — Army (the consolidated $10B / 10-year software ceiling; Maven Smart System, $795M modification), Navy (ShipOS), intelligence community, and numerous civilian agencies (each counted separately).
- Allied governments — UK (NHS Federated Data Platform ~£330M; defense), and other NATO-aligned states.
- Commercial enterprises — across energy, healthcare, manufacturing, financial services, and increasingly mid-market via AIP boot camps.
Chokepoints / single-source dependencies: (1) U.S. federal budget & procurement — the demand chokepoint; a continuing-resolution freeze or defense-budget shift hits the government engine directly. (2) Hyperscaler hosting — Palantir depends on the same clouds that also sell competing data/AI stacks (a supplier that is also a competitor). (3) The FDE talent pool — the deployment model is people-intensive; scaling commercial faster than it can hire/retain elite engineers is a genuine constraint, partly mitigated by AIP's productization. Lens fails generic — these names are the chain.
Lens 3 · Competitive Advantages (moats)
Where the moat is real:
- Switching costs / embeddedness — once Foundry/AIP becomes the "ontology" and operating layer for an organization's data, ripping it out means re-plumbing the enterprise. The ~4-year weighted contract duration and the expansion dynamics (see NDR below) are evidence the lock-in holds.
- Government accreditation / trust as a moat — clearances, FedRAMP/IL-level authorizations, and a 20-year track record on classified work are a regulatory-and-relationship moat that new entrants cannot speed-run. Palantir famously sued the Army and won (2016), then won the work — a signal of how entrenched the procurement position is.
- The deployment/ontology IP + data-gravity — the proprietary semantic layer that maps messy enterprise/operational data into a governed model is hard to replicate; AIP rides on top of it, which is why Palantir can be model-agnostic and still differentiated.
- Expansion economics — net dollar retention 139% (Q4 2025), and in Q1-2026 most of the revenue increase came from existing customers ($367M of the government increase and $352M of the commercial increase were from pre-existing accounts). That is a land-and-expand flywheel.
Bargaining power: Strong over commercial customers once embedded (they need Palantir more than vice-versa). More balanced with the U.S. government — Palantir needs the budget, but the government increasingly needs Palantir's integration layer for AI-enabled operations, which is shifting power toward Palantir (reflected in sole-source-style awards and the $10B ceiling). Over suppliers (hyperscalers, model labs), Palantir has moderate power because it is model/cloud-agnostic and can substitute.
Where the moat is thinner than bulls claim: AIP's orchestration layer is more contestable than the ontology layer — hyperscalers and Databricks/Snowflake are racing to own enterprise agentic-AI, and the foundation-model labs are pushing "agents-on-your-data" directly. The durable moat is the governed-ontology + government-trust core; the fast-growth commercial AIP layer sits on more contested ground.
Lens 4 · Segments
The research layer's segments.csv is header-only, so segment numbers are read from the filings' MD&A. Palantir reports along two axes: customer type (government vs. commercial) and geography (U.S. vs. international).
Q1-2026 (three months ended 2026-03-31) — by customer type:
| Segment | Q1-2026 rev | YoY | Note |
|---|
| Total revenue | $1,632.6M | +85% | fastest growth since 2020 IPO |
| Government (total) | +76% YoY (+$371M) | — | $367M of the increase from existing customers |
| — U.S. government | $687M | vs $373M PY (~+84%) | the anchor |
| Commercial (total) | +95% YoY (+$377M) | — | $352M of the increase from existing customers |
| — U.S. commercial | $595M | vs $255M PY = +133% | the breakout story |
So in Q1-2026, U.S. revenue grew ~104% YoY, with U.S. commercial (+133%) the standout — Palantir is no longer a "government stock," it is becoming a U.S.-enterprise AI stock, which is exactly what the multiple is paying for.
Full-year trajectory:
| Metric | FY2023 | FY2024 | FY2025 |
|---|
| Revenue | $2,225.0M | $2,865.5M | $4,475.4M |
| YoY growth | — | +28.8% | +56.2% |
| Gross profit | $1,793.9M | $2,299.5M | $3,686.3M |
| Income from operations | $120.0M | $310.4M | $1,414.0M |
| Operating margin | 5% | 11% | 32% |
| Net income | $217.4M | $467.9M | $1,634.6M |
| Net margin | 10% | 16% | 36% |
The trend is accelerating revenue (28.8% → 56.2%) with rapidly expanding GAAP margins (operating 5% → 11% → 32%) — the rare combination of growth acceleration and operating leverage. Cause: the AIP commercial engine plus government scale, with cost of revenue falling to ~18% of revenue and OpEx growing far slower than revenue (classic software operating leverage as bookings convert).
Phase B — Measure performance
Lens 5 · Earnings Result (latest print: Q1-2026, reported 2026-05-04)
The numbers:
- Revenue $1,632.6M, +84.7% YoY (from $883.9M). Beat the ~$1.54B consensus.
- GAAP gross profit $1,416.8M → 86.8% gross margin (88% ex-SBC). Cost of revenue only $215.8M.
- GAAP income from operations $754.0M → 46.2% operating margin (vs 19.9% PY). Adjusted operating margin 60% (vs 44% PY), adjusted income from operations $983.5M.
- GAAP net income $876.4M; net income to common $870.5M.
- GAAP diluted EPS $0.34 (basic $0.36); adjusted EPS $0.33 vs $0.28 consensus.
- Interest income $66.4M + other income $68.2M — the $8B cash pile is now a real earnings contributor.
- Contribution margin 73% (from 61%).
Which lines drove it: U.S. commercial (+133%) and U.S. government ($687M); expansion within existing accounts did most of the work.
Guidance / outlook: management raised FY2026 revenue guidance to $7.65–7.66B (+71% YoY) — a 10-point raise vs prior — well above the ~$7.27B LSEG consensus. U.S. commercial revenue guidance raised to +120% YoY. Q2-2026 guided to ~$1.80B (vs ~$1.68B consensus). FY2026 adjusted FCF guided to $4.2–4.4B. The tone got more confident, not less.
Balance-sheet flags:
- Cash & equivalents $2,291.6M + marketable securities $5,734.8M = $8.03B liquidity; zero debt.
- Operating cash flow $899M for the quarter (from $310M PY) — cash generation is real, not accrual-flattered.
- Receivables jumped to $1,405.6M from $1,042.1M (Dec 2025) — receivables growing faster than even this revenue rate; watch DSO (see Lens 10), though customer deposits ($370M) and deferred revenue ($517M) offset on the liability side.
- Accumulated deficit narrowed to $(2,691.9)M from $(3,562.4)M as profits compound.
Market reaction: despite the beat-and-raise, the stock fell ~7% on May 5, on valuation/durability concerns. That is the single most important tape signal in this dossier: good news is no longer good enough at this multiple. Post-earnings PTs ranged ~$180–210.
Unusual vs. its own history: 85% growth is the fastest since the 2020 direct listing — an acceleration at $4.5B+ scale is genuinely abnormal for enterprise software and is the crux of the bull case.
Lens 6 · Earnings Calls (sentiment trend)
No transcripts in the research layer (transcripts/ empty), so this is ``.
Management's narrative arc across the last several calls:
- Consistent escalation of conviction. Karp's framing moved from "AIP demand is unprecedented" (2024) to explicit triumphalism about a U.S.-enterprise "ferocious" adoption and the company's role in a Western AI buildout (2025–2026). The tone is the most bullish it has been.
- Recurring phrases: "ontology," "AIP," "boot camps," "ferocious"/"unprecedented" demand, "rule of 40 / rule of 80," "the West," "agentic." The rhetoric is unusually ideological for an earnings call — a Karp signature.
- What they lean into now vs. before: much heavier emphasis on U.S. commercial acceleration and on Palantir as critical AI infrastructure for the U.S. government, and on free-cash-flow + GAAP-profitability milestones (a deliberate answer to the "it's all SBC" critique).
- Sentiment shift: monotonically more confident, with management explicitly leaning into the valuation debate (arguing the market underappreciates the durability) rather than hedging. The risk in that posture: it raises the bar — the "stock fell on a beat-and-raise" reaction (Lens 5) shows the market is now pricing the triumphalism, so tone alone no longer moves it up.
Lens 7 · Comps
Peers: high-growth enterprise/data/AI-software names. Multiples are `` with source/date; never fabricated. (Index-peer pull n/a — Palantir's index topic is robotics, whose census is hardware names, not software peers; peers below are selected on business comparability.)
| Company | Ticker | Mkt cap | EV/Sales | EV/EBITDA | P/E | Note |
|---|
| Palantir | PLTR | ~$308B | ~55–61x | ~99x | fwd ~92x / ttm ~150x | the growth + margin outlier |
| Snowflake | SNOW | n/a | 16.0x | 102.0x | n/a | data cloud |
| Datadog | DDOG | n/a | 20.9x | 85.4x | n/a | observability |
| CrowdStrike | CRWD | n/a | 19.8x | 72.2x | n/a | security platform |
| Palo Alto Networks | PANW | n/a | ~11.2x | n/a | n/a | security |
| 5-yr avg ROE | — | n/a for any name | — | — | — | (ROE history not sourced; do not fabricate) |
Read: Palantir trades at roughly 2.5–3.5x the EV/Sales of the fastest-growing software peers (SNOW 16x, DDOG 21x, CRWD 20x) and at ~30x the software-industry median EV/Sales. The bull defense is that none of those peers is simultaneously growing 71% (guided FY26) and posting 60% adjusted operating margins — Palantir's "Rule of 40" score is off the chart (≈71 growth + ≈60 margin ≈ 130). But even granting a deserved premium, ~55x sales is a multiple that has historically only been sustained by companies that went on to grow into it flawlessly; most did not. Dividend yield: 0% (no dividend) for all growth names.
Lens 8 · Stock-Price Catalysts (>5% moves, last ~5 years)
Mostly ``. Pattern of what actually moves PLTR:
- 2023 → 2024 melt-up: from ~$6 (early 2023) up ~2,300% to record highs by late 2024 — driven by the AIP launch (2023) and the commercial inflection.
- S&P 500 inclusion (Sept 2024): forced index buying; a structural re-rating and liquidity event — "speculative → structural".
- Government mega-awards: the consolidated $10B / 10-year Army ceiling and the Maven Smart System $795M modification (2025); Navy ShipOS ($446M); UK defense (~£750M) — defense headlines repeatedly spike the stock.
- Peak ~$207 (Nov 3, 2025), then a ~38% drawdown to ~$127 (June 2026) — driven by valuation de-rating, "AI bubble" fears, a publicized Michael Burry bearish bet, and heavy insider selling, not by any deterioration in the business.
- Earnings reactions have flipped sign: the same magnitude of beat that sent the stock up in 2024 now sends it down (the −7% on the Q1-2026 beat-and-raise).
What the pattern reveals: the market reacts to (1) U.S. commercial acceleration (the multiple's justification), (2) defense/government award headlines, and (3) — now dominantly — valuation/positioning and macro AI-sentiment. At this multiple, PLTR trades as much on flows and narrative (index, retail, "AI trade" risk-on/off) as on fundamentals. That cuts both ways and raises realized volatility.
Phase C — Judge people & books
Lens 9 · Management
- Track record: Co-founders Alex Karp (CEO) and Peter Thiel (Chairman) built Palantir from a 2003 counter-terrorism tool into a $4.5B-revenue, GAAP-profitable, S&P 500 company — a 20-year compounding story and one of the few "DARPA-adjacent → commercial" successes. Shyam Sankar (CTO / President) is the long-tenured operating engine behind the FDE model and the AIP productization. This is a founder-led team with a demonstrated ability to win in the hardest procurement environment on earth (and to sue the U.S. Army and win).
- Tenure & skin in the game: Founders since 2003. Class F super-voting structure (1,005 Class F shares) gives the founders entrenched control disproportionate to economics — governance is founder-controlled, full stop. Insider ownership is meaningful but the direction of travel is selling (below).
- Capital allocation: Reinvest-and-hoard.
$8B net cash, zero debt, growing FCF ($4.2–4.4B guided FY26). Buybacks are token ($75M FY2025) and are dwarfed by SBC ($684M FY2025) — so net share count rises. No dividend. The capital-allocation story is "let cash compound on the balance sheet and fund growth with equity," which is defensible at this growth rate but means shareholders bear ongoing dilution. ROE/ROIC is structurally high and rising as GAAP net income compounds on an asset-light base (precise ROE not sourced — do not fabricate).
- Red flags: Relentless insider selling. Over the trailing 6 months, insiders executed ~227 open-market sales and 0 purchases. Karp sold ~$132M (incl. a $65.9M block Feb 20, 2026) under a 10b5-1 plan authorizing ~10M shares through Sept 12; Sankar ~$50M (incl. $22.5M May 20, much tax-withholding-driven); Thiel filed to offload large blocks. CEO cash comp is modest ($8.6M 2025) — the wealth is equity, and they are monetizing it. 10b5-1 plans soften the signal, but the unanimity and scale of selling, with zero insider buying through a 38% drawdown, is a real "insiders think it's fully valued" tell.
- Founder vs. professional manager: Emphatically founder-led / mission-driven — Karp's ideological framing ("the West," defending Western AI supremacy) is a feature for government trust and a recruiting magnet, but it also injects political/reputational risk (Lens 10) and key-person dependency. For this stage (hyper-growth, government-entangled), founder control is mostly an asset.
Lens 10 · Forensic Red Flags
Forensic lens — every figure labeled.
- Stock-based compensation flattering non-GAAP. Q1-2026 SBC $201.6M (~12.4% of revenue); FY2025 SBC ~$684M on $4.5B revenue (
15%). GAAP is now genuinely profitable (Q1 GAAP op margin 46%), so the historic "all the profit is non-GAAP" critique has weakened — but the gap between adjusted operating margin (60%) and GAAP (46%) is still SBC + payroll taxes ($201.6M + $28.0M add-backs). Dilution is the real cost: net share count rises because buybacks ($75M) << SBC.
- Receivables outrunning revenue. Accounts receivable rose to $1,405.6M from $1,042.1M (Dec→Mar), a ~35% jump in one quarter. Revenue grew 85% YoY but the sequential AR build is worth monitoring for DSO creep / large government billings timing. Offsetting: operating cash flow was strong ($899M) and customer deposits ($370M) + deferred revenue ($517M) show customers prepaying — so this looks like billing-cycle/government timing, not a revenue-quality red flag, but it is the line to watch next quarter.
- Revenue recognition: ratable over multi-year terms with O&M bundled as single performance obligations — conservative, low manipulation surface. RPO $4.1B at YE2025, ~38% to be recognized in the next 12 months. Total Remaining Deal Value $11.2B, +105% YoY — note TRDV assumes all options exercised and no terminations, so it is a softer, more promotional metric than RPO; the bookable backlog is the $4.1B RPO.
- Segment reporting: clean two-axis disclosure (customer type × geography); no goodwill/intangible bloat (asset-light; PP&E only $55.7M) — low impairment risk.
- Other-income reliance: $68.2M "other income" + $66.4M interest income contributed ~$135M of pre-tax income in Q1 — a chunk of the bottom line is balance-sheet/below-the-line, not operations. Tax rate is very low (provision $12.2M on $888.6M pre-tax ≈ 1.4%) due to NOLs/valuation-allowance dynamics — a normalizing tax rate is a forward EPS headwind not in trailing numbers.
Regulatory findings (required):
- SEC: No SEC Litigation Releases and no AAERs name Palantir (EDGAR EFTS, LR + AAER, search window 2021-06-20 → 2026-06-20).
- Securities litigation (10-K Item 3 / Note 8): A putative securities class action — Cupat v. Palantir Technologies Inc., et al. (D. Colo., consolidated 1:22-cv-02834), alleging false/misleading statements under the Exchange Act §§10(b)/20(a) and Securities Act §§11/15 — was dismissed with prejudice on April 4, 2025, judgment entered for defendants; plaintiffs filed a Notice of Appeal (May 2, 2025). Net: a favorable resolution, now on appeal. Otherwise "subject to legal proceedings and claims arising in the ordinary course."
- Non-SEC / political & reputational (web): This is the live regulatory/ESG risk for Palantir, and it is material — not to the income statement (yet), but to its social license and certain government relationships:
- ICE / immigration: immigrant-rights litigation over Palantir's "ELITE" (Enhanced Leads Identification & Targeting for Enforcement) ICE tool; ACLU campaign documenting Palantir's role in deportation operations; a ~$30M ICE case-management contract (April 2025) after Palantir had declined ERO work in 2020.
- UK NHS: the cross-party House of Commons Science, Innovation and Technology Committee called for termination of NHS England's Palantir Federated Data Platform contract (June 3, 2026); the BMA advised doctors to limit engagement (Feb 2026); an earlier "covert smear campaign" against a critic was investigated by NHS England. A ~£330M contract and a ~£400M six-year relationship are at reputational risk.
- Defense: Maven Smart System reportedly used to identify strike targets — an ESG/headline risk that is, paradoxically, also the growth driver.
- No FTC/DOJ/CFPB monetary enforcement action found.
- Summary: No SEC accounting enforcement; the one securities class action was dismissed with prejudice (on appeal); the genuine, ongoing exposure is political/reputational/social-license (ICE, NHS), which could cost specific contracts and weigh on the multiple but has not impaired the financials. Verified via SEC EDGAR EFTS (LR, AAER), 10-K Item 3/Note 8, and web search as of 2026-06-20.
Phase D — Project & stress-test
Lens 11 · Forward Projection (FY2026 / FY2027 / FY2028 EPS)
Built bottom-up from the latest actuals + management guidance; output ``, inputs labeled. (No forecast.ts create — watchlist/unattended rule.)
Anchors:
- FY2026 revenue guided $7.65–7.66B (+71%); adjusted FCF $4.2–4.4B guided.
- Q1-2026 adjusted operating margin 60%; adjusted EPS run-rate ~$0.33/qtr and rising.
- Diluted share count ~2.57B and creeping up ~1–2%/yr from SBC.
Base case (most-likely):
- FY2026 revenue ~$7.66B (top of guide; Palantir has serially beaten + raised) → adjusted operating income at ~58–60% margin ≈ $4.45B; after low-but-normalizing tax and below-the-line interest income, FY2026 adjusted EPS ≈ $1.45–1.55. Note: Street adjusted-EPS consensus precision n/a — not cleanly sourced; the quarterly $0.33 actual is the firm anchor.
- FY2027 revenue +~45% to ~$11.1B (decel from 71% as the law of large numbers bites, still elite) at ~60% adj op margin → FY2027 adjusted EPS ≈ $2.05–2.25.
- FY2028 revenue +~35% to ~$15.0B → FY2028 adjusted EPS ≈ $2.75–3.05.
Bull case: U.S. commercial keeps compounding 100%+, government AI budgets expand under the current administration, AIP becomes the default enterprise agentic layer → revenue ~$8.0B / $12.5B / $18B and margins push toward 62–64% → FY2028 adjusted EPS ~$3.50+. Even here, at ~$127 the stock trades ~36x FY2028 bull EPS — rich, not absurd, if growth holds.
Bear case: commercial growth halves to ~40% as competition and AI-budget digestion bite, government lumpiness/CR risk shows up, tax rate normalizes toward 20%+, SBC keeps diluting → revenue ~$7.4B / $9.8B / $12.5B, GAAP-tax-adjusted EPS materially below base → FY2028 adjusted EPS ~$2.10. At that path, ~$127 is ~60x FY2028 bear EPS — the de-rating risk is large.
The honest summary: the business base case is excellent; the stock outcome hinges almost entirely on multiple compression, because even ~30% three-year EPS CAGR leaves the shares expensive on any normalized multiple. (Brier forecast intentionally not logged per unattended rule; if logged, the scoreable line would be PLTR FY2026 adjusted EPS >= $1.50, p≈0.55, resolves 2026-12-31.)
Lens 12 · Bull vs Bear
Bull case. Palantir is the rare asset compounding revenue faster at scale (28.8% → 56.2% → 71% guided) while posting 60% adjusted / 46% GAAP operating margins and a Rule-of-40 score near 130 — a combination essentially unmatched in software. The moat (governed ontology, switching costs, 139% NDR, government accreditation) is real and widening into U.S. commercial. AIP positions Palantir as the integration and orchestration layer for enterprise agentic AI — the "operating system" for AI deployment — and it captures that value model-agnostically, riding model-price deflation instead of paying for it. $8B net cash, zero debt, $4.2–4.4B FCF, and a founder team that wins the hardest deals on earth. If the U.S.-and-allied "Western AI buildout" thesis is right, Palantir is a structural, multi-year compounder and today's 38%-off-the-high price is an entry, not a warning.
Bear case (permanent-impairment risks). (1) Valuation is the risk — at ~55x EV/Sales / ~92x forward P/E, the stock discounts a decade of flawless execution; even a great business can be a bad stock from here, and the −7% reaction to a beat-and-raise proves the market is exhausted on good news. (2) Growth durability — U.S. commercial +133% laps into brutal comps; a deceleration to "merely" 40% would crater the multiple even as the business thrives. (3) Competitive encroachment on the AIP layer — hyperscalers, Databricks/Snowflake, and the model labs themselves are racing to own enterprise agentic AI; the contested layer is exactly the high-growth one. (4) Social-license / political risk (ICE, NHS-FDP termination push) could cost contracts and compress the multiple. (5) Government concentration / budget cyclicality and a normalizing tax rate are forward EPS headwinds.
Pre-mortem (18 months out, thesis broke): It's late 2027. U.S. commercial growth decelerated from 100%+ to ~35% as enterprises paused AI spend and hyperscaler-native agentic tools won the mid-market; a government CR/budget fight pushed two big awards right; the tax rate normalized; SBC kept diluting; and the multiple compressed from ~55x to ~20x EV/Sales — a >50% drawdown despite revenue still growing 35%. The business was fine; the price was the problem. (This is the single most plausible way to lose money in PLTR.)
Are multiples too high? Yes on any historical/relative basis (~30x the software median, ~3x the fastest peers). The bull rebuttal — unique growth+margin — is legitimate and supports a premium, but not obviously this premium. The honest read: the multiple already pays for the bull case, so the risk/reward is asymmetric to the downside even if you believe the story.
Contrarian view (what the market refuses to see): Two opposing under-appreciations. Bulls underprice multiple-compression risk — that you can be completely right on the business and still lose money. Bears underprice the GAAP-profitability + FCF inflection — Palantir quietly became a real-earnings, $4B+ FCF machine, which means it can grow into the multiple faster than the "it's all SBC/hype" crowd models. The market is fighting the last war (2021 "unprofitable SaaS") on a company that has changed character.
Lens 13 · Devil's Advocate (short-seller)
Dismantling the bull case.
- What structurally breaks the model: Palantir's growth is increasingly a U.S.-commercial AIP story layered on a contested technology (agentic-AI orchestration). If the hyperscalers and model labs make "deploy agents on your governed data" a native, cheap feature, Palantir's fastest-growing layer commoditizes — leaving the slower, lumpier government core. The bull multiple assumes the commercial S-curve is durable and defensible; it may be neither.
- Revenue concentration & what shifts it: Government is still the anchor (U.S. gov $687M in Q1). It is exposed to U.S. budget politics, continuing resolutions, and — newly — administration/political risk cutting both ways (currently a tailwind; a future administration hostile to Palantir's defense/ICE entanglement is a tail risk). Commercial concentration in a handful of large expanding accounts (most growth from existing customers) means a few logo losses or expansion stalls hit hard.
- Why the moat is weaker than bulls think: the durable moat is the ontology + government trust; the orchestration layer riding LLMs is the contested, fast-growth part. Bulls conflate the two.
- Most dangerous competitor bulls underestimate: the hyperscalers' own agentic-AI + data stacks (Microsoft especially) bundling "good-enough" governance into platforms enterprises already own — plus Databricks (private, aggressive) on the data-ops axis and Anduril on defense autonomy.
- Worst capital-allocation / governance issues: token buybacks dwarfed by SBC = perpetual dilution; Class F founder super-voting control removes shareholder leverage; and the unanimous, large-scale insider selling with zero buying through a 38% drawdown is the insiders' own valuation verdict.
- What must hold for today's price: ~50%+ growth for years, margins holding ~60% adjusted, no tax-rate normalization shock, the commercial moat proving durable against hyperscalers, and the multiple not compressing. That is a lot of simultaneous "yes."
- If growth disappoints 20–30%: a path to ~40% growth (from 71% guided) likely triggers multiple compression from ~55x toward 20–25x EV/Sales — a 40–60% drawdown even with revenue still rising. The valuation is the short, not the business.
- Single permanently-impairing scenario & plausibility: hyperscaler/model-lab commoditization of the agentic layer coinciding with a U.S. government budget retrenchment — plausible-but-not-base-case (~15–20%); more likely is a painful de-rating (not impairment) as growth normalizes.
Lens 14 · Management Questions (ordered by information value)
- U.S. commercial grew 133% in Q1 — what is the sustainable growth rate once the boot-camp cohort laps, and how much of current commercial revenue is land (pilots) vs. durably expanded production deployments?
- As hyperscalers and foundation-model labs ship native "agents-on-your-data," what specifically keeps AIP's orchestration layer from commoditizing — and can you quantify win/loss vs. Microsoft/Databricks in competitive commercial deals?
- Total Remaining Deal Value is $11.2B but RPO is $4.1B — walk us through the gap (options vs. obligations) and what share of TRDV you'd underwrite as high-probability.
- With buybacks (
$75M) a fraction of SBC ($684M), what is the multi-year net dilution trajectory, and at what point does capital return shift from token to material?
- The effective tax rate is ~1–2% on NOLs — when does it normalize, to what, and what is the GAAP-EPS impact you're guiding the Street to model?
- How exposed is the FY2026 government guide to continuing-resolution / budget timing, and how lumpy could the quarterly cadence get on the $10B Army ceiling?
- The NHS Federated Data Platform faces a parliamentary termination push and the ICE work draws active litigation — how do you quantify the revenue at risk from social-license/political pushback, and is it changing your commercial pipeline anywhere?
- Net dollar retention is 139% — what is the ceiling, and which cohorts are decelerating first?
- Accounts receivable rose ~35% sequentially while you guide accelerating growth — what is DSO doing, and how much is government billing-cycle timing vs. structural?
- How dependent is the AIP roadmap on third-party frontier models, and what is your contingency if model-lab pricing or access terms change adversely?
- The FDE/deployment model is talent-intensive — what is the productivity curve (revenue per engineer) as AIP productizes, and is hiring a binding constraint on the commercial ramp?
- What does "winning the Western AI buildout" mean in dollars — TAM, your realistic share, and the 3-year revenue bridge to it?
- International is underpenetrated vs. the U.S. surge — what is the strategy and the realistic timeline, given the trust frictions in markets like the UK?
- Given Class F super-voting control, how should public shareholders think about governance accountability and succession/key-person risk around the founder team?
- At ~55x sales, what specific milestones over the next 8 quarters would you point to as evidence the business is growing into the valuation rather than the valuation needing to come to it?