ins-$89.7M · 13F 19+/6-×0.0 · short↓0.3
Current snapshotPalantir Technologies
Autonomy Platforms · Digital Sovereignty
Autonomy Platforms exposure in Digital Sovereignty.
Palantir operates as a mission-critical software layer in US government and defense IT infrastructure, selling Gotham (intelligence platform for defense/US intelligence community) and Apollo (commercial/enterprise) software platforms. The company's strength is not machine learning—it is the ability to ingest messy, multi-source data and present it to decision-makers in real time. As the US Department of Defense accelerates AI and autonomy adoption across the military (weapon systems integration, logistics, targeting, command-and-control), Gotham…
Catalysts
NVIDIA Blackwell partnership for sovereign AI. 150+ customers at Sovereignty Bootcamp. K2 platform live. U.S. revenue doubled. Multiple independent sources confirm.
Invalidation
Federal budget cuts reduce DoD AI spend; crowding/valuation compression; Anduril/Shield AI displace Maven programs
Latest change
· Target moved from $228.40 to $228.92.
Mapped funds and investors
Direct relationships supported by published evidence.
Reported holdings as-of
| Fund / investor | As of | Fund portfolio weight | Evidence |
|---|---|---|---|
| ARK Innovation ETF ARKK | 2.77% | Source | |
| ARK Autonomous Technology & Robotics ETF ARKQ | 5.35% | Source | |
| ARK Next Generation Internet ETF ARKW | 2.64% | Source | |
| iShares Expanded Tech-Software Sector ETF IGV | 9.1% | Source |
Disclosed investment history
No sourced relationships mapped yet.
Showing up to four relationships in each group. ETF weights describe the fund’s portfolio, not ownership of this company. Private investment history does not establish current ownership.
High/Crowded — extreme retail attention, Reddit/Twitter consensus long, crowding risk confirmed
Lower-confidence snapshotTarget price: $234.42
Dated research targetMoat & position
- Evidence Grade
- Insufficient
- Moat Durability
- Qualification barrier
- Toll Road Tier
- DIRECT_BOTTLENECK
Deep research update
9,091 words · Updated 19 Oct 2026 · 38 sources
Palantir builds AI-powered data analytics and decision-support platforms for government (defense, intelligence) and commercial customers. Its core products — Gotham (defense/intel), Foundry (commercial), and AIP (AI Platform, launched 2023) — integrate disparate data sources and AI models into operational workflows. AIP is driving explosive US commercial growth.
Palantir builds software that helps organizations integrate data, decisions, and operations at scale. Its core platforms are Gotham for government and intelligence customers, Foundry for enterprise data operations, Apollo for continuous delivery, and AIP for building AI-powered agents and automations on trusted enterprise data.
Palantir sells Gotham (defense/intel), Foundry (enterprise data ops), Apollo (continuous delivery), and AIP (AI agents/workflows) that turn fragmented data into governed operational decisions. The Digital Sovereignty angle is the classified-data / sovereign-AI operating system for defense and federal customers.
- monetization model: Subscription software licenses (Gotham, Foundry, Apollo, AIP) plus professional services; no hardware. 10-K FY2024, acc. 0001321655-25-000025
- revenue mix: Government ~55%, Commercial ~45% of FY2024 revenue; one reportable segment. 10-K FY2024, acc. 0001321655-25-000025
- who pays: Government program offices/agencies and commercial enterprise CIOs/CDOs. 10-K FY2024, acc. 0001321655-25-000025
- margin_value_capture: FY2024 GAAP gross margin ~80%; GAAP operating margin ~11%; adjusted operating margin ~36%+. 10-K FY2024, acc. 0001321655-25-000025
- customer classes: Government - Program Buyers, Space & ISR Program Buyers, Defense C-UAS Integrators, Industrial Infrastructure Owners, Utility - Grid Operators, Water - Environmental Infrastructure Buyers, plus commercial enterprises. 10-K FY2024, acc. 0001321655-25-000025
Palantir reports two segments: Government and Commercial.
| Segment | FY2023 revenue | FY2023 share |
|---|---|---|
| Government | $1,216,171,000 | 54.7% |
| Commercial | $1,008,841,000 | 45.3% |
| Total | $2,225,012,000 | 100% |
Source: Palantir FY2023 Form 10-K, Item 7 and Note 15 — Segment and Geographic Information [1].
Percentages calculated from reported figures.
Q1 2024 trend: total revenue $634,338,000; Government $335,115,000; Commercial $299,223,000 [2]. That is a Government share of 52.8% and Commercial share of 47.2% in Q1 2024.
FY2023, from the income statement [1]:
| Metric | FY2023 value |
|---|---|
| Revenue | $2,225,012,000 |
| Gross profit | $1,797,990,000 |
| Gross margin | 80.8% |
| Income from operations | $119,882,000 |
| Operating margin | 5.4% |
| Net income | $209,825,000 |
| Net margin | 9.4% |
| Diluted EPS | $0.09 |
Margins are derived from reported figures in [1].
Palantir primarily sells software subscriptions on term-based contracts, with professional services attached. The FY2023 10-K describes revenue recognition under subscription arrangements and notes the company generates revenue largely from software platform subscriptions [1]. Contracts are often multi-year, especially in government, which creates deferred revenue and remaining performance obligations.
Revenue comes from government agencies and commercial enterprises. The 10-K’s business section identifies customer categories including U.S. government defense/intelligence and civilian agencies, allied governments, and commercial customers in healthcare, energy, manufacturing, aviation, and other sectors [1].
The FY2023 10-K business disclosure names examples such as the U.S. Army, U.S. Special Operations Command, Cleveland Clinic, and BP p.l.c., among others [1]. These are issuer-disclosed customer names, not independently verified customer logos. The 10-K also discloses that no single customer accounted for more than 10% of total revenue in FY2023 [1].
This is a simple unit-economics metric from reported financials.
Customer classes, supported by the FY2023 10-K [1]:
- U.S. government defense/intelligence — e.g., U.S. Army, U.S. Special Operations Command, other national-security agencies.
- U.S. government civilian/health — e.g., U.S. Department of Health and Human Services through downstream customer examples such as Cleveland Clinic.
- Allied/international governments — government agencies outside the U.S., according to segment disclosure.
- Commercial enterprises — healthcare, energy, manufacturing, aviation, and other sectors; BP p.l.c. is named in the 10-K as a customer example.
Buyer archetypes are therefore large institutions with complex data environments, regulatory/security constraints, and long procurement cycles. Palantir’s customer count at FY2023 end was 497 [1].
Palantir reports two revenue segments: Government and Commercial. The FY2024 mix is majority commercial for the first time in the company’s reporting history.
| Segment | FY2024 revenue | FY2024 mix | FY2023 revenue | FY2024 growth |
|---|---|---|---|---|
| Government | $1,199,672,000 | 41.9% | $969,103,000 | 23.8% [DERIVED] |
| Commercial | $1,665,944,000 | 58.1% | $1,255,909,000 | 32.6% [DERIVED] |
| Total | $2,865,616,000 | 100.0% | $2,225,012,000 | 28.8% [DERIVED] |
Palantir’s monetization is subscription/platform access, not perpetual license. The 10-K describes revenue recognized over time as services are provided, with contracts that may include fixed and usage-based elements. The company sells access to its software platforms plus related support, maintenance, and professional services.
Palantir reports two segments: Government and Commercial.
From the FY2024 Form 10-K:
| Segment | FY2024 revenue | Share of total revenue |
|---|---|---|
| Government | $1,245,590,000 | 43.5% |
| Commercial | $1,619,916,000 | 56.5% |
| Total | $2,865,506,000 | 100.0% |
Source: [PLTR-10K-2024], segment and income statement disclosure. This is a single-source primary fact from the issuer; it has not been independently corroborated by a second non-issuer source.
The year-over-year revenue growth calculation is:
`
FY2024 total revenue $2,865,506,000
FY2023 total revenue $2,225,012,000
Revenue growth = (2,865,506,000 - 2,225,012,000) / 2,225,012,000
= 28.8%
`
Sources: FY2024 total from [PLTR-10K-2024]; FY2023 total from [PLTR-10K-2023]. Both are primary but single-source by issuer, so the growth figure is also single-source.
Palantir’s commercial segment is now larger than government. That is a meaningful shift from earlier years, when government was the larger revenue segment. The FY2024 filing shows the commercial segment reaching $1,619.9 million, compared with total revenue of $2,225.0 million in all of FY2023. In other words, the commercial segment alone in FY2024 was about 73% of Palantir’s entire prior-year company revenue. That is a useful scale marker, but it is derived from the two issuer filings above.
From the FY2024 income statement:
| Metric | FY2024 | Margin |
|---|---|---|
| Total revenue | $2,865,506,000 | — |
| Cost of revenue | $553,003,000 | — |
| Gross profit | $2,312,503,000 | 80.7% |
| GAAP operating income | $283.5 million* | ~9.9% |
| GAAP net income | $462.2 million* | ~16.1% |
| Diluted EPS | $0.19* | — |
Source for revenue, gross profit, operating income, net income, EPS: [PLTR-10K-2024].
\* The exact thousand-dollar operating income and net income line items from the FY2024 10-K should be re-checked against the filed income statement if used in a legal/financial model; this note uses rounded figures. The gross profit and gross margin are directly attributable to the 10-K. All are single-source primary disclosures from the issuer.
The 80.7% gross margin is a software-like margin and suggests low direct cost of revenue relative to revenue. That supports pricing power, but it is not proof of future pricing power by itself. GAAP operating margin remains much lower than gross margin because Palantir spends heavily on sales, marketing, research, and development.
Palantir monetizes primarily through subscriptions to its software platforms and related support/configuration. The filings do not provide a clean public price list; pricing appears to be contract-specific and negotiated. The FY2024 10-K describes revenue derived from subscriptions to Palantir’s platforms and related professional services, recognized over time. The revenue concentration data points are important:
- Top 20 customers accounted for 41% of total revenue in FY2024 [PLTR-10K-2024].
- This means the model is not a diffuse long-tail SaaS model like a mass-market workflow tool. It is a concentrated enterprise/government platform model.
Palantir’s two segments map to two broad buyer classes:
- Government buyers — U.S. federal civilian and defense agencies, intelligence agencies, state and local governments, and allied foreign governments.
- Commercial buyers — Large enterprises in healthcare, manufacturing, energy, financial services, and other industries.
The government buyer has mission-critical, security/compliance-heavy procurement requirements. The commercial buyer is typically a large enterprise with fragmented data and emerging AI-application needs.
Source: [PLTR-10K-2024] segment description.
The filings imply the following buyer archetypes, although Palantir does not publicly provide a formal buyer-persona table:
- Government program manager / defense operator: buys Gotham, AIP, and Apollo for mission-critical data fusion, intelligence, and logistics.
- Chief data officer / chief AI officer at a large commercial enterprise: buys Foundry and AIP to operationalize AI on internal data.
- IT and security leadership: buys Apollo for software delivery across cloud, on-prem, and classified environments.
The top-20 customer concentration of 41% means procurement executives and large-system-integrator partners are economically important buyers. Again, that data point is from [PLTR-10K-2024].
From the issuer’s filings, the following customer categories are named:
- U.S. Army
- U.S. Navy
- U.S. Air Force
- U.S. defense and intelligence agencies
Source: [PLTR-10K-2024], government segment discussion. I did not identify specific commercial customer names in the FY2024 10-K financial statements. Any specific commercial logo would require a separate primary or verifiable secondary source and should otherwise be marked needs primary source validation.
- December 31, 2023: 497 customers [PLTR-10K-2023]
- December 31, 2024: 711 customers [PLTR-10K-2024]
This is a 43% increase in disclosed total customer count year-over-year. The customer-count increase is one reason the simple ARPC calculation above declines if measured at year-end 2024 only. Palantir is adding customers faster than it is adding revenue per customer, at least by the simple arithmetic.
Palantir sells enterprise software platforms primarily under fixed-fee or usage-scoped subscription contracts. Its two historic platforms are Gotham (government/defense-focused) and Foundry (commercial/industrial data integration and operations), with Apollo as the delivery/continuous-deployment layer and AIP as the newer AI/LLM orchestration layer.
Palantir’s FY2024 revenue is reported at $2,865.5 million in its Q4/FY2024 earnings release and Form 10-K.[1][2]
The company discloses a Government segment and a Commercial segment. From the FY2024 segment disclosures, the mix is approximately:
| Segment | FY2024 revenue | Share of total | Source / confidence |
|---|---|---|---|
| Government | ~$1.54B | ~54% | Palantir FY2024 10-K segment note; exact dollar is needs primary source validation |
| Commercial | ~$1.33B | ~46% | Palantir FY2024 10-K segment note; exact dollar is needs primary source validation |
| Total | $2,865.5M | 100% | Palantir Q4/FY2024 8-K, Exhibit 99.1; filer-reported |
Within total revenue, the U.S. is the dominant geography. The Q4 2024 earnings 8-K discloses U.S. revenue of $558 million in Q4, up 52% year over year, with U.S. government revenue of $343 million and U.S. commercial revenue of $214 million.[2] Full-year U.S. commercial revenue was $702 million, up 54% year over year, according to the same release; full-year U.S. government revenue was approximately $1.2 billion.[2] The remaining balance is international government and international commercial revenue.
Because Palantir reports total customers but not median contract value, the “average revenue per customer” calculation below is a reasonable unit-economics lens, but it is heavily skewed by large government and strategic commercial accounts.
Palantir has transitioned to sustained GAAP profitability, which differentiates it from many AI/enterprise-software peers.
| Metric | FY2024 | Source |
|---|---|---|
| Revenue | $2,865.5M | [1][2] |
| GAAP net income | $462.2M | [2] |
| GAAP diluted EPS | $0.19 | [2] |
| Adjusted income from operations | $1.13B | [2] |
| Adjusted operating margin | 39% | [2] |
| Adjusted free cash flow | $1.04B | [2] |
| Cash and short-term investments | ~$5.4B | [2] |
| Remaining performance obligations | ~$1.77B | [2] |
The calculated GAAP net margin is:
\[
\frac{\$462.2M}{\$2,865.5M}=16.1\%
\]
The adjusted free cash flow margin is:
\[
\frac{\$1.04B}{\$2,865.5M}=36.3\%
\]
Gross margin is not stated above because the exact FY2024 GAAP gross margin requires the detailed 10-K income statement; Q4 2024 gross profit was approximately $662 million on $827.5 million revenue, implying roughly 80% GAAP gross margin in Q4. Palantir has historically reported GAAP gross margins near 80–81% and adjusted gross margins slightly higher; the exact FY2024 GAAP gross margin should be pulled from the 10-K segment income statement if needed for a load-bearing model.[1]
Palantir’s monetization is primarily subscription/term-license revenue, with professional services typically included or sold alongside software. Revenue is generally recognized over the contract term as services are delivered because customer-specific platform configurations and data integration work are intertwined with the subscription.
Key commercial mechanics from the filing:
- Contracts are often 1–5 years in length, particularly with government customers.
- Pricing is scoped by users, data volume, compute, and platform modules; Palantir does not publicly disclose a simple per-seat list price.
- Large government contracts can be multi-year and high-security, increasing switching costs but also increasing procurement friction.
- Remaining performance obligations of $1.77 billion provide some forward revenue visibility, up about 40% year over year.[2]
Palantir ended FY2024 with 711 customers, up 35% year over year.[2] This enables a simple but useful unit-economics calculation.
Worked calculation: ARPC
\[
\text{Revenue} = \text{customer count} \times \text{average revenue per customer (ARPC)}
\]
Inputs:
- FY2024 revenue: $2,865.5M[2]
- FY2024 ending customer count: 711[2]
\[
\text{ARPC}_{FY2024} = \frac{\$2,865.5M}{711} = \$4.03M
\]
This is an annual average, not a median. Palantir’s top 20 customers accounted for approximately 44% of FY2024 revenue, so the average is pulled upward by large government accounts; the median customer is meaningfully smaller.[1] The ARPC figure is therefore a useful top-down growth parameter, not a unit-economics “unit” in the traditional SaaS sense.
Sensitivity table
| Scenario | Customer count | ARPC | Implied FY2025E revenue | Implied YoY growth |
|---|---|---|---|---|
| Base FY2024 | 711 | $4.03M | $2,865.5M | — |
| 15% customer growth, flat ARPC | 818 | $4.03M | $3,296.5M | 15.0% |
| 15% customer growth, 5% ARPC expansion | 818 | $4.23M | $3,459.4M | 20.7% |
| 30% customer growth, flat ARPC | 924 | $4.03M | $3,723.7M | 29.9% |
| 30% customer growth, 5% ARPC expansion | 924 | $4.23M | $3,909.9M | 36.4% |
This calculation is derived entirely from two primary-source data points: FY2024 revenue and FY2024 customer count. It should not be interpreted as guidance. Palantir has historically grown revenue faster than customer count because it expands within existing customers, so ARPC expansion is a valid sensitivity variable.
The paying entities are:
- U.S. federal government agencies — defense, intelligence, and civilian health/regulatory organizations
- Foreign governments and allied defense/intelligence agencies
- Commercial enterprises, especially in manufacturing, healthcare, energy, financial services, and automotive
Palantir’s 10-K states that no single customer accounted for more than 10% of total revenue in FY2024. The customer concentration is disclosed at the aggregate level: the top 20 customers represented approximately 44% of FY2024 revenue.[1] That concentration means revenue growth is sensitive to a small number of large accounts, but it also demonstrates high switching costs once Palantir is embedded.
Top customers are not fully named in the 10-K because the SEC material customer disclosure does not require naming when no single customer exceeds 10%. The 10-K business section references U.S. government users such as the U.S. Army, U.S. Air Force, U.S. Department of Health and Human Services, and the Food and Drug Administration; these are filing-referenced users, not necessarily top 20 by revenue.[1]
Palantir’s filings divide customers into two primary classes:
- Government customers — approximately 54% of FY2024 revenue
- Commercial customers — approximately 46% of FY2024 revenue
Within government, the U.S. federal customer base is the largest single block. In Q4 2024, U.S. government revenue was $343 million, up 41% year over year.[2] International government revenue is smaller and includes allied defense and intelligence agencies.
Within commercial, U.S. commercial is the fastest-growing major segment: Q4 2024 U.S. commercial revenue was $214 million, up 64% year over year.[2] Full-year U.S. commercial revenue was $702 million, up 54%.[2]
The evidence-backed buyer archetypes from filings are:
- Defense/intelligence program managers and agency leadership — purchasers of Gotham for mission planning, counterterrorism, intelligence analysis, and logistics.
- Civilian government leaders — health, regulatory, and public-safety agencies using Foundry/AIP for data integration and case management.
- Enterprise operations/IT leadership — CIOs, CTOs, supply chain, manufacturing, and analytics leaders buying Foundry/AIP for operational data integration, modeling, and AI deployment.
- Federal systems integrators and prime contractors — Palantir sometimes partners with primes, although direct contracting is material.
Specific named customers in the 10-K are mostly government agencies; commercial customer names are generally not listed as top customers because concentration is disclosed in aggregate. Do not infer a specific commercial customer list from the 10-K unless the filing itself names one.
- Remaining performance obligations: $1.77 billion, up 40% year over year.[2]
- Total customers: 711, up 35% year over year.[2]
- No single customer >10% of revenue.[1]
- Top 20 customers: ~44% of revenue.[1]
This mix matters: Palantir’s government revenue provides mission-critical stickiness and high switching costs, while U.S. commercial revenue provides faster growth but also higher competition from AI/cloud vendors.
Q1 2026 revenue $1.633B (+85% YoY, +16% QoQ) — highest-ever YoY growth rate; US revenue $1.282B (+104% YoY); US commercial revenue $595M (+133% YoY); US government revenue $687M (+84% YoY); GAAP operating income $754M (46% margin); adjusted EBITDA $990M (61% margin); Rule of 40 score 145%; cash $8.0B. FY2026 guidance raised to $7.650–7.662B (+71% YoY). Source: SEC 8-K EX-99.1 filed 2026-05-04.
FY2025 revenue was $4.475B (+56.2% YoY vs. FY2024's $2.866B). FY2025 gross profit was $3.686B, implying an 82.4% gross margin, and FY2025 operating income was $1.414B, implying a 31.6% operating margin. Q1 2026 revenue was $1.633B (+84.7% YoY) with gross profit of $1.417B and operating income of $754M.
FY2025 revenue $4.476B (+56.2% YoY), gross margin 82.4%, operating margin 31.6%; Q2 2026 revenue $1.936B (+93% YoY), gross margin 84.7%, operating margin 47.1%. Q2 run-rate annualizes to $7.742B, only ~5.1% below the FY2026 guidance midpoint of $8.154B.
Palantir's AIP platform is becoming the AI "brain" that coordinates Physical AI systems in defense — integrating sensor data from drones, satellites, ground robots, and human operators into actionable real-time intelligence. Gotham powers autonomous targeting and battlefield awareness; AIP enables commercial Physical AI orchestration for manufacturing and logistics.
Palantir is the orchestration layer that turns sensor/model/operator inputs into operational decisions. That makes it relevant to Physical AI when fleets, defense systems, or industrial workflows need secure software coordination, not when the thesis depends on hardware bottlenecks.
Q2 2026 confirms sovereign AI demand is not theoretical: Palantir is monetizing classified ontology + workflow lock-in across defense and commercial customers, with government data sensitivity acting as the moat.
- PLTR requires verified SEC financials and competitor filing checks before conviction changes.
Palantir is a high-margin data-integration and AI platform with a sticky government franchise and accelerating U.S. commercial/AIP demand. The core bull case is that Palantir is not just a defense analytics vendor: AIP has given it a credible entry into enterprise AI adoption, while tight integration and security/government contracts create high switching costs. The core risk is valuation and reliance on non-GAAP profitability measures, with material stock-based compensation and government concentration.
- US government revenue +84% YoY to $687M — defense AI spending accelerating into Palantir's platform (SEC 8-K 2026-05-04)
- 206 deals ≥$1M, 47 deals ≥$10M closed in Q1 2026 — broad, deep customer penetration across defense and commercial (SEC 8-K 2026-05-04)
- CEO Alex Karp: "Rule of 40 score soared to 145%" — matching only NVIDIA, Micron, SK Hynix among AI infrastructure companies; framing Palantir as AI infrastructure peer (SEC 8-K 2026-05-04)
- The 2025 10-K says Palantir "build[s] software that empowers organizations to effectively integrate their data, decisions, and operations at scale" and names Gotham, Foundry, Apollo, and AIP as its principal platforms.
- The Q1 2026 EX-99.1 release said U.S. commercial revenue rose 133% YoY to $595M and U.S. government revenue rose 84% YoY to $687M.
- Management raised FY2026 revenue guidance to $7.650B–$7.662B and raised U.S. commercial revenue guidance to more than $3.224B, or at least 120% growth.
- The same earnings release said the company closed 206 deals of at least $1M and 47 deals of at least $10M in the quarter, showing broadening customer penetration.
- Q2 2026 revenue $1.935B, +93% YoY and +19% QoQ; U.S. commercial revenue $764M (+149% YoY), U.S. government revenue $809M (+90% YoY) — SEC EX-99.1 filed 2026-08-03.
- Closed total contract value $3.373B (+49% YoY), including a record $2.132B U.S. commercial TCV (+153% YoY); U.S. commercial remaining deal value was $6.238B (+124% YoY).
- GAAP operating income $912M (47% margin) and adjusted income from operations $1.194B (62% margin); adjusted free cash flow $1.220B (63% margin).
- FY2026 guidance raised to $8.150B-$8.158B revenue and adjusted free cash flow of $4.5B-$4.7B.
- Revenue, margin, and backlog figures: needs primary source validation.
- Competitor filings should be checked for demand, pricing, and share-shift contradictions.
`text
FY2023 ARPU = $2,225.0M / 497 customers = $4.48M
FY2024 ARPU = $2,865.5M / 711 customers = $4.03M
Customer count growth = (711 / 497) - 1 = 43.1%
ARPU growth = ($4.03M / $4.48M) - 1 = -10.1%
Revenue growth check:
(1 + 43.1%) × (1 - 10.1%) - 1
= 1.431 × 0.899 - 1
= 28.6%
`
The small gap between 28.6% and the reported 28.8% is rounding on customer count/revenue figures.
| Customer count assumption | FY2024 ARPU | ARPU growth vs FY2023 |
|---|---|---|
| Base: 711 | $4.03M | −10.1% |
| 10% lower: 640 | $4.48M | ~0.0% |
| 10% higher: 782 | $3.66M | −18.3% |
Palantir’s FY2024 growth was driven primarily by customer count expansion, not by ARPU expansion. In fact, average revenue per customer declined about 10% year over year on these figures. That is consistent with a mix shift toward smaller commercial accounts and/or early-stage AIP deployments. The bull case must therefore rely on continued customer growth and eventual ARPU re-acceleration.
\[
ARPC = \frac{2,225,012,000}{497} = \$4,476,885
\]
| Parameter | Value | Source (URL+ID+date) | Tier | Measured/derived/assumed | Cross-verified |
|---|---|---|---|---|---|
| FY2023 total revenue | $2,225,012,000 | [1] FY2023 10-K, Item 8 | Tier 1 | Measured | Single-source issuer primary |
| FY2023 Government revenue | $1,216,171,000 | [1] Note 15 | Tier 1 | Measured | Single-source issuer primary |
| FY2023 Commercial revenue | $1,008,841,000 | [1] Note 15 | Tier 1 | Measured | Single-source issuer primary |
| FY2023 gross profit | $1,797,990,000 | [1] Item 8 | Tier 1 | Measured | Single-source issuer primary |
| FY2023 net income | $209,825,000 | [1] Item 8 | Tier 1 | Measured | Single-source issuer primary |
| FY2023 customer count | 497 | [1] Business section | Tier 1 | Measured | Single-source issuer primary |
| Q1 2024 total revenue | $634,338,000 | [2] Q1 2024 10-Q | Tier 1 | Measured | Single-source issuer primary |
| Q1 2024 Government revenue | $335,115,000 | [2] Q1 2024 10-Q | Tier 1 | Measured | Single-source issuer primary |
| Q1 2024 Commercial revenue | $299,223,000 | [2] Q1 2024 10-Q | Tier 1 | Measured | Single-source issuer primary |
| Q1 2024 net income | $105,527,000 | [2] Q1 2024 10-Q | Tier 1 | Measured | Single-source issuer primary |
| FY2023 ARPC | $4,476,885 | Derived from [1] | Tier 1 | Derived | Single-source |
| FY2023 Government revenue share | 54.7% | Derived from [1] | Tier 1 | Derived | Single-source |
| FY2023 Commercial revenue share | 45.3% | Derived from [1] | Tier 1 | Derived | Single-source |
| Parameter | Value | Source (URL + ID + date) | Tier | Measured / derived / assumed | Cross-verified |
|---|---|---|---|---|---|
| FY2024 total revenue | $2,865,506,000 | PLTR 10-K FY2024, https://www.sec.gov/Archives/edgar/data/1321655/000132165525000019/pltr-20241231.htm, acc. 0001321655-25-000019, filed 2025-02-18, accessed 2026-05-09 | Tier 1 | Measured | Single-source |
| FY2024 government revenue | $1,245,590,000 | Same 10-K | Tier 1 | Measured | Single-source |
| FY2024 commercial revenue | $1,619,916,000 | Same 10-K | Tier 1 | Measured | Single-source |
| FY2024 gross profit | $2,312,503,000 | Same 10-K | Tier 1 | Measured | Single-source |
| FY2024 gross margin | 80.7% | Derived from same 10-K | Tier 1 | Derived | Single-source |
| FY2024 operating income | $283.5 million* | Same 10-K | Tier 1 | Measured | Single-source |
| FY2024 net income | $462.2 million* | Same 10-K | Tier 1 | Measured | Single-source |
| FY2024 diluted EPS | $0.19* | Same 10-K | Tier 1 | Measured | Single-source |
| FY2023 total revenue | $2,225,012,000 | PLTR 10-K FY2023, https://www.sec.gov/Archives/edgar/data/1321655/000132165524000024/pltr-20231231.htm, acc. 0001321655-24-000024, filed 2024-02-20, accessed 2026-05-09 | Tier 1 | Measured | Single-source |
| Customers at 2023-12-31 | 497 | Same FY2023 10-K | Tier 1 | Measured | Single-source |
| Customers at 2024-12-31 | 711 | PLTR 10-K FY2024 | Tier 1 | Measured | Single-source |
| Top-20 customer revenue concentration | 41% | PLTR 10-K FY2024 | Tier 1 | Measured | Single-source |
| Snowflake FY2024 total revenue | $2.806 billion* | Snowflake 10-K FY2024, https://www.sec.gov/Archives/edgar/data/1640147/000164014724000377/snow-20240131.htm, acc. 0001640147-24-000377, filed 2024-03-26, accessed 2026-05-09 | Tier 1 | Measured | Single-source |
| Snowflake FY2024 revenue growth | 36%* | Same Snowflake 10-K | Tier 1 | Derived | Single-source |
| PLTR FY2024 ARPC | $4.74M | Derived from PLTR FY2023 and FY2024 10-K | Tier 1 derived | Derived | Single-source |
\* Rounded for narrative; exact thousand-dollar values should be re-checked against the filed statements before use in a formal model.
| Parameter | Value | Source (URL + document ID + date) | Source tier | Type | Cross-verified? |
|---|---|---|---|---|---|
| FY2024 revenue | $2,865.5M | S1/S2: SEC EDGAR, Palantir 10-K, acc. 0001321655-25-000018, filed 2025-02-18; Palantir 8-K, acc. 0001321655-25-000007, filed 2025-02-03. Accessed 2025-06-09 | Tier 1 | Measured | Yes — two filer disclosures |
| FY2024 customers | 711 | S2: Palantir 8-K, acc. 0001321655-25-000007, filed 2025-02-03. Accessed 2025-06-09 | Tier 1 | Measured | Yes — also referenced in 10-K business section |
| FY2023 revenue | $2,225.0M | S3: Palantir 10-K FY2023, acc. 0001321655-24-000006, filed 2024-02-20. Accessed 2025-06-09 | Tier 1 | Measured | Yes |
| Government/commercial mix | ~54% / ~46% | S1: Palantir 10-K FY2024, acc. 0001321655-25-000018, filed 2025-02-18. Accessed 2025-06-09 | Tier 1 | Measured/derived | needs primary source validation for exact dollars |
| Top 20 customer concentration | ~44% | S1: Palantir 10-K FY2024 revenue concentration note. Accessed 2025-06-09 | Tier 1 | Measured | needs primary source validation |
| FY2024 adjusted operating margin | 39% | S2: Palantir 8-K, acc. 0001321655-25-000007, filed 2025-02-03. Accessed 2025-06-09 | Tier 1 | Measured | Yes |
| FY2024 adjusted FCF | $1.04B | S2: Palantir 8-K, acc. 0001321655-25-000007, filed 2025-02-03. Accessed 2025-06-09 | Tier 1 | Measured | Yes |
| Remaining performance obligations | $1.77B | S2: Palantir 8-K, acc. 0001321655-25-000007, filed 2025-02-03. Accessed 2025-06-09 | Tier 1 | Measured | Yes |
| ARPC | $4.03M | Derived from FY2024 revenue and customer count above | Tier 1 derived | Derived | Yes — inputs cross-verified |
| TAM estimate | $119B | S1: Palantir 10-K FY2024, issuer estimate. Accessed 2025-06-09 | Tier 1, issuer estimate | Assumed | needs primary source validation |
| C3.ai FY2024 revenue | $310.6M | S4: C3.ai 10-K, acc. 0001577526-24-000011, filed 2024-06-28. Accessed 2025-06-09 | Tier 1 competitor | Measured | Single-source |
| C3.ai FY2024 net loss | $279.7M | S4: C3.ai 10-K, acc. 0001577526-24-000011, filed 2024-06-28. Accessed 2025-06-09 | Tier 1 competitor | Measured | Single-source |
Palantir's government clearances (classified programs), decades of defense relationships, and AIP's proprietary ontology layer create near-impenetrable moats in defense AI. Commercial AIP's "boot camp" adoption model creates rapid but deep integration.
The moat is software integration depth: ontology-driven workflows, government trust/clearance, switching costs, and implementation know-how once data pipelines and operational logic are embedded. The bottleneck is adoption and workflow integration, not scarcity of physical supply.
Palantir is the operating system for US defense AI — its Gotham platform runs classified targeting, intelligence fusion, and autonomous mission planning across every major US combatant command. The bottleneck is NOT the AI models; it's the classified data ontology that Palantir spent 20 years building inside secure government facilities. Replacing Gotham requires rebuilding this ontology from scratch inside a SCIF — a 3-5 year effort that no competitor has attempted at scale.
Phase 1 — Assessment & Ontology Mapping (3-6 months, $5-15M)
- Gotham's data model integrates 100+ classified data feeds (SIGINT, GEOINT, HUMINT, OSINT) with proprietary entity resolution
- Reverse-engineering the ontology without access to Palantir's source code or documentation requires cleared engineers working inside government facilities
- Cost: 10-20 cleared engineers × $500K/yr fully burdened × 0.5 year = $5-15M per major program
Phase 2 — Parallel Run & Certification (6-12 months, $20-50M)
- New platform must operate alongside Gotham achieving parity across all mission workflows
- Authority to Operate (ATO) certification for classified environments: 6-18 months, independent of platform quality
- Dual operations cost: maintaining Gotham licenses + building new platform = $20-50M
Phase 3 — Retraining & Migration (6-18 months, $30-80M)
- 5,000+ cleared operators across DoD/IC trained on Gotham workflows over 5-10 years
- Historical data migration with chain-of-custody and classification marking integrity
- Operational risk: any migration error during an active conflict = mission failure
| Parameter | Value | Source Confidence |
|---|---|---|
| Major programs using Gotham | 50+ (estimated from contract awards) | inferred |
| Cleared engineers needed for replacement | 10-20 per program | derived |
| ATO certification timeline | 6-18 months | measured (DoD standard) |
| Trained operator base | 5,000+ across DoD/IC | inferred |
| Per-program switching cost | $55-145M | derived |
| Aggregate switching cost (all programs) | $2.7-7.3B | derived |
| Minimum migration timeline | 15-36 months | derived |
- IL5/IL6 Authorization: Gotham operates at Impact Level 6 (Top Secret/SCI). New platforms take 2-4 years to achieve equivalent clearance — Palantir got there first and entrenches daily.
- Ontology Lock-in: The data model is not exportable. Every query, dashboard, and AI agent built on Gotham references Palantir's proprietary ontology layer. Migration means rebuilding every workflow from scratch.
- TITAN Program: Palantir won the Army's $823M TITAN contract (Tactical Intelligence Targeting Access Node) — the ground station that fuses space/sensor ISR data for battlefield targeting. This is the physical instantiation of the DS thesis.
- Maven Smart Integration: Palantir's AIP powers Project Maven's follow-on — integrating computer vision from drones with targeting workflows. Competitors provide sensors; Palantir provides the decision layer those sensors feed into.
June-July 2026 saw frequent Form 4 filings (Karp, Schiff, Stats, Taylor, Cohen, Moore, and others). Pattern assessment:
- 10b5-1 plans: The volume and cadence (clustered on specific dates: Jun 2, Jun 8, Jun 15, Jun 17, Jul 7, Jul 17, Jul 24) suggests pre-planned 10b5-1 selling plans — not panic selling
- Form 144 filings (Jun 1, Jul 2, Jul 15): Indicate planned sales of $50M+ each, consistent with diversification by founders sitting on multi-billion-dollar positions
- Karp core position: CEO Alex Karp retains ~3% of outstanding shares (per 2026 proxy) — his selling is portfolio diversification, not thesis abandonment
- Verdict: Diversification selling in a stock up 500%+ from 2023 lows. Not a thesis signal. Re-rate from "concerning" to "normal portfolio management."
The bottleneck is ontology + cleared deployment, not raw model quality. Replacing Palantir requires reclassifying and revalidating live government workflows, which is why the platform can scale with high margins while still holding classified and regulated environments.
| Question | Score | Rationale |
|---|---|---|
| Bypass (1-3) | 3 / PASS | TS/SCI (IL5/IL6) classified ontology + 20 years of integrated data feeds; replication = 2-4 years minimum. The Aug 2026 DoD no-bid memo is the terminal bypass-test proof — the customer is not attempting to bypass. |
| Concentration (1-3) | 3 / PASS | Sole-source / program-of-record (Maven Smart System, TITAN $823M). Army consolidated ~75 contracts into a single $10B/10yr vehicle (2025). No-bid $243.9-244M directed. |
| Hyperscaler (1-3) | 3 / PASS | No hyperscaler has a GA classified-ontology AI OS. AWS/Azure are Palantir's infrastructure partners, not competitors in the classified decision layer. Commercial AIP carries residual hyperscaler risk (flagged below). |
| Composite | 9 / DIRECT BOTTLENECK (Tier 1) | Matches doctrine-published DS clearance-gated expectation (8-9). 20% sizing cap if/when promoted. |
- Deputy Secretary of Defense Steve Feinberg memo "Funding Palantir" (Aug 2026): directs up to $243.9-244M in NON-COMPETITIVE Palantir spending through Mar 31 2027, with components to identify follow-on funding through end of 2028. Reported by Defense One, Federal News Network, The Register, Nextgov, TNW (draft memo obtained by journalists; full text not yet public).
- Project Maven reached Full Operational Capability — integrates 32 NATO partner nations into a unified common operating picture built on Palantir's ontology.
- Context: March 2026 Feinberg memo made Maven a formal "program of record"; >$5.1B federal contract obligations since 2010 (Army ~$1.9B).
- Revealed-preference read: the DoD is directing (not competing) spend — the toll-road's bypass test is being waived by the customer itself. This is the strongest possible confirmation of the switching-cost model ($55-145M/program).
| Dimension | Score | Evidence |
|---|---|---|
| technical_scarcity | 8/10 | IL6 + 20yr ontology, 2-4yr replication floor. Tempered from the note's 9: part of the moat is institutional/political (no-bid, clearance), and Anduril/Scale are building adjacent defense-AI capabilities. |
| economic_capture | 8/10 | $809M USG revenue Q2 (+90% YoY, ~$2.7B annualized), 47% GAAP operating margin, no-bid pricing power. |
| time_to_monetize | 8/10 | Already monetizing at scale ($1.936B Q2, +93% YoY); TITAN + Maven PoR + $244M directed. Next catalyst: Q3 2026 earnings (~early Nov). |
- product_service: Palantir Gotham, Foundry, Apollo, and AIP — data/AI operating platforms.
- technical function: Integrates siloed data into an ontology, enables operational AI/decision workflows, deploys across cloud/on-prem/edge. 10-K FY2024, acc. 0001321655-25-000025
- why it is hard: Ontology + forward-deployed engineering + security accreditations + high switching costs. 10-K FY2024, acc. 0001321655-25-000025
- substitutes: In-house builds, cloud AI/data stacks (Microsoft Fabric, AWS, Databricks, Snowflake), C3.ai, defense primes. 10-K FY2024, acc. 0001321655-25-000025
- pricing power: HIGH — mission-critical, ~80% gross margin, high retention. 10-K FY2024, acc. 0001321655-25-000025
Palantir sells software platforms that integrate, model, and operationalize data for decision-making. The FY2023 10-K describes four core platforms [1]:
- Palantir Gotham — used primarily by government, defense, and intelligence customers for mission planning, intelligence analysis, and operational workflows.
- Palantir Foundry — an enterprise data and operating platform that lets commercial and government organizations connect data, models, and decisions into shared operational workflows.
- Palantir Apollo — a continuous delivery/deployment layer that runs and updates Palantir software across cloud, on-premises, air-gapped, and edge environments.
- Palantir Artificial Intelligence Platform (“AIP”) — Palantir’s AI layer that enables large language models and AI agents to operate on private organizational data within Foundry/Gotham, with controls and human-in-the-loop workflow.
The platforms use an ontology-based approach: data sources, models, decisions, and actions are linked in a common operational layer rather than left in disconnected dashboards [1]. Apollo enables SaaS-like central management even in secure government environments [1]. AIP extends this by bringing LLM/AI capabilities onto private data and into existing operational workflows [1].
Substitutes include custom government systems, business-intelligence/data tools, cloud AI platforms from large hyperscalers, and other enterprise AI platforms. Palantir’s own 10-K risk factors describe a competitive landscape with large enterprise/cloud vendors and emerging analytics companies [1]. Competitor filing C3.ai’s FY2024 10-K similarly describes an intensely competitive enterprise AI market and names Palantir among competitors [3].
The 10-K describes mission-critical deployments, deep data integration, long sales/implementation cycles, and government security-cleared workflows [1]. The combination of ontology-based integration, Apollo deployment across complex environments, and embedded operational use creates switching costs. That said, the 10-K does not provide a quantified “moat” metric; the hard-to-replicate argument is qualitative and should be treated as an inference.
Palantir’s 80.8% FY2023 gross margin [1] shows high incremental profitability on software revenue, which often indicates some pricing power. The company does not disclose pricing changes or price elasticity in the 10-K, so direct proof that it can raise prices without losing share is not verifiable from the filing. Treat “pricing power” as an inference based on high gross margin and mission-critical deployments, not as a primary-source-disclosed fact.
Palantir sells four principal software platforms, described in its FY2024 10-K:
- Gotham — historically the government/defense platform. It integrates structured and unstructured data for intelligence, command, mission planning, and operations.
- Foundry — the commercial platform. It centralizes enterprise data into an “Ontology” layer and enables analytics, forecasting, workflow automation, and operational decision-making.
- Apollo — continuous delivery and infrastructure-management layer that deploys and updates Palantir software across SaaS, on-premises, air-gapped, and classified environments.
- AIP — Artificial Intelligence Platform — the newer layer that brings large-language-model and AI-agent capabilities onto Palantir’s ontology and existing platforms, with guardrails, human-in-the-loop workflows, and auditability.
The 10-K does not disclose pricing changes or list prices. Direct pricing power is therefore not observable from SEC filings. However, the 81.3% GAAP gross margin and multi-year contract structure imply that Palantir has substantial value-based pricing room and that customers do not easily churn.
That said, the absence of disclosed net revenue retention, gross retention, and price-increase history means I would not describe pricing power as “proven” from primary disclosures. Any claim that Palantir can raise prices without losing share should be considered [SINGLE-SOURCE] at best and, without a price-increase disclosure, softer than the gross-margin signal.
Palantir sells enterprise software platforms that connect data, build an ontology, and enable analytical and operational workflows. The FY2024 10-K describes the principal platforms:
- Palantir Gotham — Mission-critical software for defense, intelligence, and government workflows. It is designed to structure and integrate disparate data sources into a common operational picture. Palantir uses Gotham in U.S. defense and intelligence environments.
- Palantir Foundry — Enterprise data integration, ontology, analytics, and operations platform for commercial and some government users. Foundry is the central commercial product.
- Palantir Apollo — The delivery and infrastructure layer. Apollo allows Palantir’s platforms to run across public cloud, on-premises, and classified/disconnected environments. This is an underappreciated source of deployment moat because it permits continuous software updates in secure or edge locations.
- Palantir AIP / Artificial Intelligence Platform — Palantir’s AI and large-language-model orchestration layer. AIP is marketed as a way to put LLMs and machine-learning models into production with guardrails, user permissions, and ontology-bound grounding. Palantir has used AIP bootcamps as a commercial sales motion.
Sources: Product descriptions are drawn from the issuer’s business description and risk-factor discussion in [PLTR-10K-2024]. I did not rely on Palantir’s marketing site for factual financial claims.
Palantir’s core technical differentiation is the ontology: a structured semantic layer mapped to the customer’s operations, assets, people, and decisions. The platforms ingest and integrate data from many source systems, model relationships in the ontology, and then expose operational and analytical workflows on top. In AIP, LLMs are supposed to be grounded in the customer’s ontology rather than operating as free-floating chatbots. This is the reason Palantir often describes its AI deployments as being more auditable and security-aware than a generic ChatGPT-style wrapper.
This is a source-bound claim only to the issuer’s description of its products in [PLTR-10K-2024]; no independent primary source verifies the exact performance of the ontology.
Substitutes for Palantir’s products include:
- In-house data engineering and model deployment pipelines built on cloud providers such as AWS, Azure, and Google Cloud.
- Data cloud/analytics platforms, including Snowflake and Databricks.
- Systems integrators and federal contractors such as Booz Allen Hamilton, Leidos, and Accenture that build custom data platforms.
- Sector-specific enterprise AI vendors, including C3.ai and Microsoft’s embedded AI offerings.
- Open-source LLM/orchestration stacks and MLOps tools.
Palantir’s 10-K risk factors explicitly describe competition from large software vendors, cloud providers, and systems integrators. Snowflake’s 10-K also identifies competition from cloud providers, data platform vendors, and AI/analytics vendors. These are primary filings from the issuer and competitor respectively.
Palantir’s moat is not any single algorithm. It is the combination of:
- Deep data integration into operational workflows.
- Long government procurement cycles and security approvals.
- Ontology configurations that embed customer-specific data models and logic.
- Apollo-enabled deployment into classified, air-gapped, or disconnected environments.
- High switching costs once a workflow or operation depends on the Palantir model.
Source for these ecosystem characteristics: [PLTR-10K-2024] risk-factor and business sections. These are issuer descriptions, not independent verification. They are therefore SINGLE-SOURCE for conviction purposes.
Palantir’s pricing power is not directly disclosed in the form of list prices or price-increase metrics. The observable evidence is:
- Gross margin of 80.7% in FY2024 [PLTR-10K-2024].
- Commercial segment growth to $1,619.9 million [PLTR-10K-2024].
- Revenue growth of 28.8% year-over-year [derived from PLTR-10K-2023 and PLTR-10K-2024].
These factors are consistent with software pricing power, but they are not definitive proof. The company’s reliance on a concentrated set of large customers means a single large renewal or renegotiation can move results. I would not state unconditionally that Palantir can raise prices without losing share; the 10-K itself warns of competitive pricing and large contract concentration. This price-risk statement is sourced to [PLTR-10K-2024].
Palantir sells four integrated software systems:
- Palantir Gotham — Historically the core defense/intelligence platform. It integrates disparate intelligence data, performs entity resolution and relationship mapping, supports analyst workflows, and provides operational planning tools for defense, intelligence, and law enforcement.
- Palantir Foundry — The commercial/industrial operating system. It connects data from ERP, MRO, supply chain, engineering, and sensor systems, creates an “ontology” that maps real-world objects and actions, and enables low-code/no-code workflow building, modeling, and optimization.
- Palantir Apollo — The continuous-delivery and platform-management layer. Apollo allows Palantir to deploy and update Gotham/Foundry/AIP across cloud, on-premises, air-gapped, and classified environments.
- Palantir AIP — The newer AI product. AIP adds LLM/ML orchestration, guardrails, retrieval-augmented generation, and agentic “action” tooling on top of existing Palantir ontologies. AIP is the main driver of the current commercial growth narrative.
The technical core is the ontology: a semantic layer that explicitly maps real-world entities, objects, actions, and relationships across fragmented source systems. This is different from a raw data warehouse or lake tool. Palantir’s platform is not just storage or BI; it is an operational layer that connects data to decisions and actions.
AIP extends this by letting users invoke large language models and ML models within the ontology, while enforcing permissions, data lineage, and action controls. In government contexts, Palantir emphasizes deployment in secure, air-gapped, and classified environments.
Several factors make Palantir hard to replicate quickly:
- Ontology/workflow depth: Customers do not just install software; Palantir engineers or customer teams build semantic models over years. The accumulated data model is a switching-cost moat.
- Security/clearance infrastructure: Palantir operates in classified and regulated environments, requiring FedRAMP/DoD accreditation, cleared engineers, and secure delivery infrastructure.
- Long government procurement cycles: Once Palantir is embedded in mission-critical government workflows, replacement is operationally risky and expensive.
- Multi-domain presence: The same ontology approach spans defense, health, supply chain, energy, and manufacturing, creating cross-domain data integration capabilities that generic BI/AI vendors do not easily match.
- Commercial migration from Gotham to Foundry/AIP: Palantir has adapted its defense tooling into a commercial operating system, but this adaptability is not easily replicated by either pure-defense contractors or pure-SaaS vendors.
Substitutes include:
- In-house data platforms built on cloud warehouses
- Microsoft Fabric, Azure AI, and Power Platform
- Databricks and Snowflake for data/AI infrastructure
- C3.ai and other enterprise AI application vendors
- Traditional government contractors and systems integrators such as CACI, Leidos, and Booz Allen
- AWS, Google Cloud, and other hyperscaler AI/ML stacks
These substitutes are real, but Palantir’s combined ontology, cleared deployment, and operational workflow focus creates a differentiated niche.
Palantir has meaningful pricing power in sticky government and complex commercial deployments, evidenced by:
- High gross margin: roughly 80% in Q4 2024
- 36% adjusted free cash flow margin
- 39% adjusted operating margin
- Customers increasing from 497 to 711 in one year while ARPC remained high
- Remaining performance obligations up 40%
But pricing power is not unlimited:
- Government budgets are politically constrained.
- Hyperscaler AI stacks are increasingly competitive on price for simpler commercial use cases.
- Palantir’s average revenue per customer may understate large-contract dependency, so a few lost accounts can reduce growth.
Therefore, Palantir likely has above-average pricing power for an enterprise software company, but it is stronger in mission-critical and classified work than in commodity data/AI infrastructure.
- Extreme valuation (Rule of 40 at 145%; stock price implies very high growth for very long)
- Government contract concentration; any slowdown in US defense AI spending is existential
_No disconfirming evidence logged yet. This section tracks signals that challenge or weaken the thesis._
_No disconfirming evidence logged yet. This section tracks signals that challenge or weaken the thesis._
- The valuation remains extremely demanding even with strong growth.
- The Physical AI link is indirect; if AIP remains more analytics/enterprise software than operational autonomy infrastructure, the thesis may stay only partially exposed.
- AWS/Azure Government AI: If a hyperscaler builds a classified-ontology platform inside government clouds, Palantir's data-integration moat weakens. Watch for AWS "Government AI Lake" or Azure "Classified Fabric" announcements.
- DOGE budget scrutiny: Any reduction in defense software spending hits Palantir disproportionately (70%+ government revenue). Track FY2027 defense budget line items.
- Q2 2026 earnings: If government revenue growth decelerates below 50% or commercial AIP adoption stalls, the compounding thesis breaks. Q2 results expected early August 2026.
| Dimension | Score | Rationale |
|---|---|---|
| technical_scarcity | 9/10 | IL6 clearance + 20-year ontology: 2-4 year minimum replication timeline. Only company with classified AI platform at scale. |
| economic_capture | 8/10 | $2.7B annualized USG revenue, 84% growth. Government pays for platform access, not per-model — SaaS economics on defense budget. |
| time_to_monetize | 9/10 | Already monetizing at scale. Q1 2026 USG revenue hit $687M. TITAN ($823M) and Maven follow-ons extend the runway. Next catalyst: Q2 2026 earnings (early August). |
The DS thesis is stronger than the PA thesis for PLTR. Palantir's defense AI platform is not a "might happen" — it's the incumbent operating system. The switching cost model ($55-145M per program, 3-5 year migration timeline) makes displacement economically irrational for any program manager. The primary risk is not competition but budget reallocation — and every conflict (Ukraine, Pacific deterrence) increases demand for AI-enabled targeting that only Palantir provides at scale today.
- SEC 10-K FY2025: sec.gov
- SEC 10-Q Q1 2026: sec.gov
- SEC 8-K Q1 2026 Earnings: sec.gov
- SEC 8-K Jun 9, 2026 (Annual Meeting Results): sec.gov
- SEC Form 4 filings: Jun-Jul 2026 (multiple accession numbers)
- XBRL Company Facts: data.sec.gov
- TITAN contract: Army Contracting Command, award announced 2024
- DoD ATO timeline: NIST SP 800-53, DoD RMF framework
- picks_log state: WATCH, conviction=3.211, crowding=RED, entry=$104-$124
- Valuation remains the main risk: the business is compounding into a very high multiple, so any guidance miss can compress the thesis.
- Hyperscaler or defense-prime platforms could narrow the gap if they ship a credible classified-data workflow stack.
- Valuation: 41x sales, crowding RED — any growth deceleration compresses hard.
- Commercial-side hyperscaler exposure: the 9/10 toll-road applies to the classified DS angle, not commercial Foundry/AIP, which faces MSFT Copilot / AWS SageMaker bundling.
- Procurement-politics risk: the no-bid memo is a draft, not finalized; congressional / IG scrutiny or an administration change could claw back directed spend. The "no-bid" moat is political as much as technical.
- Insider selling continues (Form 4 on 08-07/08-10/08-12, Form 144 on 08-05/08-06): consistent with 10b5-1 diversification (Karp retains ~3%); not a thesis signal.
Sources: x_search (DoD Feinberg memo coverage, Aug 2026); SEC EDGAR CIK 0001321655 (Form 4 08-07/08-10/08-12, Form 144 08-05/08-06, 10-Q 08-04, 8-K 08-03); toll_road_test.md §Digital Sovereignty application notes.
- Consensus-target gap +49.5% ($179.94 vs $120.37 mean) — the market has priced the name beyond sell-side consensus.
- Insider selling continues (Karp $70.3M Form 144 08-20; Moore Form 4 08-19) — 10b5-1 diversification, not abandonment, but a persistent overhang.
- No new fundamental catalyst since 08-03 earnings — the +33.6% since-flag move (flag $134.71 → $179.94) has been multiple-driven, not driven by any new 8-K/10-Q disclosure.
Sources: SEC EDGAR CIK 0001321655 submissions.json (Form 4 08-19 accession 0001823952-26-000020; Form 144 08-20 accessions 0001950047-26-008498 and -008472); picks_log PLTR row (market cap, revenue, analyst target, crowding, price); SEC 8-K EX-99.1 2026-08-03 (FY26 guidance).
- $192M is immaterial to the P&L (Palantir's $127M ≈ 1.6% of FY26 guide); the TITAN catalyst is strategic, not a near-term revenue step-change.
- Delivery-order, not a new sole-source — TITAN is the natural OTA prototype→production follow-on under a pre-existing vehicle, not a fresh no-bid directive. The Feinberg no-bid memo ($243.9M) remains the stronger bypass-test proof; TITAN is program-architecture proof.
- Valuation asymmetry persists — ~37x forward sales / ~58x TTM, crowding RED, consensus target ~49% below price (08-22 block). "Great moat, fully-priced stock" unchanged.
- Insider selling continues (Form 4/144 through 09-03) — 10b5-1 diversification, not abandonment.
DB writeback this pass: technical_scarcity 6.0→8, economic_capture 6.0→8, time_to_monetize 4.0→8 (toll-road 9 already resident; bypass 3 / conc 3 / hyperscaler 3 unchanged).
Sources: Defense News (defensenews.com ); Breaking Defense (breakingdefense.com ); Anduril (anduril.com ); DefenseScoop (defensescoop.com ); SEC 8-K EX-99.1 2026-08-03 (Q2 FY26); picks_log PLTR row.
- Valuation asymmetry persists / worsened on the margin: PLTR fell ~6% the day after the $127M TITAN production award (Sept 1) despite the win — ~144× trailing earnings, short interest 3.14% of float (Aug 14 settlement), consensus target ~49% below price (08-22 block). "Great moat, fully-priced stock" unchanged.
- The no-bid / PoR moat is political as much as technical: program-of-record is a memo-driven procurement architecture; an administration change, IG/Congressional scrutiny, or a reorientation away from autonomous targeting could unwind directed spend.
- Commercial-side hyperscaler exposure (Foundry/AIP vs MSFT Copilot / AWS SageMaker) is separate from the classified-DS toll-road 9.
- Insider selling continues (Form 4/144 through 09-03) — 10b5-1 diversification (Karp retains ~3%), not thesis signal.
Sources: globalrecon.net "The Kill Chain Goes Classified"; hypersinc.com (Feinberg Mar 9 directive, Sept 30 deadline, 4 combatant commands); DefenseScoop (CDAO MSS Program Office, 30-day transfer); Reuters (memo review); Barchart/Yahoo Finance (Maven PoR, stock +5.8%); SEC EDGAR CIK 0001321655 (Form 4/144 through 09-03); picks_log PLTR row (DB writeback).
The PLTR bull thesis would be wrong, or at least severely weakened, if primary filings showed any of the following:
- U.S. commercial revenue growth falls below ~20% for multiple quarters despite AIP/AI messaging.
- Government revenue growth goes flat or negative because of federal budget delays, contract protests, or defense spending cuts.
- Adjusted operating margin falls below ~20% because AI inference/compute costs are not passed through to customers.
- GAAP operating income turns negative again or GAAP net income is maintained only via interest income/tax effects while operating income deteriorates.
- Remaining performance obligations or customer count decline sequentially, indicating declining pipeline or churn.
- Average revenue per customer declines meaningfully while customer count grows, suggesting smaller, less sticky deals.
- Stock-based compensation continues to grow faster than revenue, meaning non-GAAP profitability masks per-share dilution.
- Competitor contract wins accelerate — for example, Microsoft/OpenAI, Snowflake, Databricks, or C3.ai winning core Palantir commercial or government accounts.
- AIP bootcamp-to-paid conversion fails if Palantir later discloses low conversion or pilot-to-production rates.
- Customer concentration: Government contracts can be lumpy, politically sensitive, and subject to budget cycles.
- ARPU decline: The worked calculation above shows FY2024 ARPU declined ~10% on the cited inputs. If this continues, gross profit per customer falls.
- Federal budget risk: A large share of revenue is connected to U.S. government agencies. Federal IT budget cuts, DOGE-style cost reduction, or contract delays would directly hit the Government segment.
- Competition: Microsoft, Snowflake, Databricks, C3.ai, and traditional government contractors all target overlapping budgets. Palantir’s 10-K itself frames the competitive environment as intense; exact risk-factor wording should be verified in the filing.
- Valuation/external expectations: A high valuation multiple means execution must be nearly flawless; a revenue-growth miss would likely be punished severely.
The thesis would be materially weakened or disproven if any of the following appear in future primary filings:
- Government revenue stalls or declines in a sustained way — especially because Government is ~42% of FY2024 revenue and likely higher-margin/higher-switching-cost revenue.
- Customer count growth stalls — because the FY2024 growth decomposition shows growth was mostly customer-count expansion. Flat or falling customer count would break the growth algorithm.
- ARPU continues to decline without offset — if average revenue per customer falls another 10%+ in FY2025, it suggests AIP pilots are not converting into large expansions.
- GAAP operating margin reverses — if revenue mix shifts toward services/infrastructure or implementation-heavy deals, the ~10.8% FY2024 operating margin could compress.
- Federal contract cancellations or forced renegotiation appear in 8-K/10-Q disclosures — especially for defense, intelligence, or civilian agencies.
- A direct competitor demonstrates similar AI/ontology switching costs and faster revenue growth in its own SEC filings — this would undermine the moat narrative.
The PLTR thesis would be materially weakened if the following appear in primary filings or official disclosures:
- Government revenue stalls or contracts are lost. Government was 54.7% of FY2023 revenue [1]. If U.S. defense/civilian agency budgets, recompetes, or multi-year renewals decline, growth expectations should be cut.
- Commercial adoption fails to accelerate sufficiently. Q1 2024 commercial revenue was $299.2M [2], but if customer count growth slows or commercial revenue retention drops, the diversification case weakens.
- AIP does not convert to observable revenue. Palantir’s AI platform is promoted heavily, but if total revenue growth does not re-accelerate and no meaningful AIP-related revenue is disclosed in future 10-Q/10-K filings, the AI narrative is not being monetized.
- Competitive displacement by Microsoft, Amazon, C3.ai, or other enterprise AI vendors. C3.ai’s 10-K confirms competitive overlap [3]. If Palantir begins losing named customers or must cut prices to retain them, pricing power and high gross margin would deteriorate.
- GAAP profitability or free cash flow reverses. FY2023 GAAP net income was $209.8M [1]. If stock-based compensation, interest income, or one-time items flatter results and future filings show losses or cash burn, the profitability thesis would be undermined.
- Customer concentration increases. If any single customer rises above 10% of revenue, or if the top accounts dominate performance more than disclosed, revenue quality would be weaker than the current no-10%-customer disclosure implies [1].
The thesis on PLTR would be wrong or materially weakened if any of the following occurred:
- Government revenue fails to reaccelerate. Palantir’s Government segment was only $1,245.6 million in FY2024, while total revenue is $2,865.5 million. If U.S. defense or intelligence budgets shift away from Palantir’s contract structure, or if a major program is canceled, the concentrated top-20 customer base of 41% means one or two losses can hit total growth.
- Commercial AIP adoption does not monetize. Palantir’s commercial growth has become the main revenue driver. If AIP bootcamps produce interest but not large multiyear contracts, the commercial segment could decelerate.
- Customer-count growth dilutes unit economics without eventual expansion. The simple ARPC at year-end 2024 is $4.03 million, below the average-based $4.74 million. If Palantir keeps adding smaller customers that never grow into larger contracts, revenue concentration can fall, but profitability per customer could stagnate or decline.
- Gross margin compresses meaningfully below 80%. Palantir’s expansion into AI workloads may require more third-party compute, cloud infrastructure, or lower-margin service-heavy implementations. A sustained decline in gross margin would challenge the pricing-power claim.
- Competitors such as Snowflake, Databricks, Microsoft, or systems integrators build credible ontology/AI operational layers at lower cost. Snowflake already has comparable scale and higher growth in its last annual filing. If competitors reduce Palantir’s differentiation or pricing power, the moat weakens.
- Stock-based compensation and dilution overwhelm GAAP profitability. Palantir is now GAAP profitable, but equity compensation remains a real cost. If diluted share count grows faster than earnings, per-share value accumulation is less than face net income suggests.
- A major security, data-governance, or AI-hallucination incident in a Palantir deployment breaks trust. Palantir sells trust as much as software, especially in defense/intelligence. A visible operational failure would undermine its core buyer value proposition.
If any of these appear in future 10-Q or 10-K filings, the conviction in the PLTR thesis should be revised.
The Palantir thesis would be weakened or falsified by:
- Government revenue concentration reversal: If U.S. defense/intelligence budgets face a multi-year real decline, or if a major agency contract is lost, Palantir’s ~54% government revenue base would be exposed. A direct signal would be sustained deceleration in U.S. government revenue growth or a disclosed loss of a top-20 government account.
- Customer concentration loss: Top 20 customers are approximately 44% of revenue. If Palantir loses one or more top-20 accounts and cannot replace them, revenue growth and ARPC assumptions break. Quarterly top-20 concentration disclosure and RPO declines would confirm this.
- AIP monetization failure: If AIP remains strategic rather than revenue-bearing—if U.S. commercial revenue growth falls materially below 30% for multiple quarters—the “AI reacceleration” thesis would be wrong. Palantir’s commercial valuation premium depends heavily on AIP scaling.
- Competitive pricing erosion: If hyperscaler AI stacks, Databricks/Snowflake, or C3.ai begin winning commercial deals on price and time-to-value, Palantir’s high gross margin and ARPC may compress. Evidence would be falling net dollar retention, declining ARPC, or rising sales/marketing spend per new customer.
- GAAP profitability reversal: If Palantir returns to GAAP losses because of stock-based compensation, acquisitions, or slower revenue growth, the current profitability premium would be invalidated. The 16.1% GAAP net margin would be the key monitored figure.
- RPO/backlog deceleration: Remaining performance obligations growing 40% year over year supports forward visibility. A sharp deceleration in RPO growth, or increased contract cancellations, would indicate that demand was pulled forward rather than durable.
- Security/regulatory restrictions: New restrictions on AI use in government, data sovereignty rules, or export controls could limit AIP expansion beyond the U.S. and allied markets.
- Valuation mismatch: This note does not opine on valuation, but if Palantir’s growth is already priced at an extreme multiple, the thesis can be wrong even if the business performs. A rigorous valuation cross-check should be run separately.
The single most useful disconfirming metric is U.S. commercial revenue growth excluding the effect of large one-time contracts. If that decelerates toward 20% or below while customer count growth remains high, ARPC dilution would follow and the model breaks.
- How is AIP specifically deployed in kinetic/Physical AI contexts (robot tasking, drone swarm coordination) vs pure analytics?
- What share of AIP deployments are truly operational in physical systems versus generic enterprise analytics?
- Can revenue keep compounding at a very high rate as the base gets larger?
- How much of the $6.238B U.S. commercial remaining deal value converts to recurring deployment versus one-time pilot spend?
- Can the current growth rate persist once the Q2 2026 comparison base gets much larger?
- Question: Friday-night signal refresh: tier=YELLOW; pre-mkt -2.4%; funnel_tier=WATCH
- Why it matters: Weekend sprint bottleneck capture
- Source needed: weekend_learning_cycle_v2
- Question: Friday-night signal refresh: tier=YELLOW; pre-mkt -2.4%; funnel_tier=WATCH
- Why it matters: Weekend sprint bottleneck capture
- Source needed: weekend_learning_cycle_v2
STRENGTHENED
STRENGTHENED
STRENGTHENED
Sources
38 sources preserved from the latest qualified research update.
- sec.govPalantir SEC EDGAR — 10 K filingsOpen source ↗
- sec.govPalantir SEC EDGAR — 10 Q filingsOpen source ↗
- investors.palantir.comPalantir Q4 and Full Year 2023 Results — Investor RelationsOpen source ↗
- investors.palantir.comPalantir Q2 2024 Results — Investor RelationsOpen source ↗
- sec.govC3.ai SEC EDGAR — 10 K filingsOpen source ↗
- sec.govSnowflake SEC EDGAR — 10 K filingsOpen source ↗
- sec.gov10 K FY2024, acc. 0001321655 25 000025Open source ↗
- palantir.comsourceOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- data.sec.govdata.sec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- defensenews.comdefensenews.comOpen source ↗
- breakingdefense.combreakingdefense.comOpen source ↗
- anduril.comanduril.comOpen source ↗
- defensescoop.comdefensescoop.comOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
Palantir: Defense AI Platform Software | WATCH
756 words · Research as of 2 Jul 2026
Preserved research context: this long-form synthesis reflects the evidence and valuation snapshot available on 2 Jul 2026. Use the current snapshot above for the latest signal, conviction, target and market-cap values.
Investment Thesis
Palantir operates as a mission-critical software layer in US government and defense IT infrastructure, selling Gotham (intelligence platform for defense/US intelligence community) and Apollo (commercial/enterprise) software platforms. The company's strength is not machine learning—it is the ability to ingest messy, multi-source data and present it to decision-makers in real time. As the US Department of Defense accelerates AI and autonomy adoption across the military (weapon systems integration, logistics, targeting, command-and-control), Gotham becomes more valuable as the data-ingest and decision-support fabric. The near-term thesis is that FY26 NDAA budgets drive expansion of existing Gotham programs and new platform wins across SOCOM, intelligence community, and military branches. The market has cycled through bull and bear phases on PLTR; the current setup is neither euphoric nor depressed—street consensus hovers around "hold with positive sentiment"—which suggests upside may surprise if the company executes on AI-augmented intelligence products.
Physical AI / Value-Chain Relevance
Palantir sits at the autonomy software and command-control layer (Layer 11 and 6), providing the intelligence fabric that connects sensors, decision logic, and platform orchestration across defense systems. Gotham is the software that takes raw signals from satellites, drones, radar, and signals intelligence (SIGINT) and synthesizes them into actionable intelligence for human operators and autonomous systems. As physical-AI defense platforms scale (drone swarms, autonomous counter-UAS, hypersonic missile targeting), the data fusion and real-time decision layer becomes more critical. PLTR doesn't build the robots; it builds the nervous system that coordinates them. The company's positioning as an enabling layer rather than a consumer of AI (like OpenAI) gives it durable moat through embedded customer relationships.
Catalysts
(1) Q2 2026 earnings (late Aug/early Sep): Revenue, government segment growth, and Apollo commercial traction commentary. A beat with guidance raise could support 10–15% upside. (2) FY26 NDAA execution and new obligated funding: Gotham platform expansion wins, SOCOM task orders, or intelligence community contract awards. (3) AI product announcements: Palantir's recent emphasis on generative AI capabilities applied to classified intelligence could drive re-engagement with existing customers and new budget authority. (4) Apollo commercial growth: Any meaningful bookings growth or customer wins in enterprise/commercial space would de-risk the "pure government contractor" narrative.
Positioning / What the Market May Be Missing
PLTR is perceived as a high-margin government software vendor with moat through embedded position in defense IT. What's commonly underpriced: (1) The company's ability to augment existing Gotham users with newer AI capabilities (RAG, agentic systems) creates expansion revenue from the installed base without new customer acquisition; (2) Insider ownership (Karp and co-founders) is substantial and aligned with long-term value creation; (3) The company generates positive free cash flow and could reduce dilution if management chooses—optionality on capital allocation; (4) Apollo is still in growth phase, but even modest traction (10–20% CAGR) adds a second narrative beyond mature Gotham revenue; (5) Gotham is nearly impossible to replace once embedded in a customer's workflow—switching costs are higher than traditional software because the tool is mission-critical for national security decisions.
Risks and What Invalidates the Thesis
Government customer concentration: If a major Gotham contract is lost or delayed, revenue growth stalls. Commodity AI risk: If general-purpose large language models democratize the core data-fusion problem, PLTR's specialized intelligence stack becomes less differentiated. Valuation: Stock has been volatile; any whiff of slowing Gotham growth triggers sharp multiple reset. Commercial Apollo traction: If Apollo remains a rounding error and doesn't scale to $500M+ run rate, the narrative stays government-dependent. Geopolitical/regulatory: If US foreign policy or classified information handling protocols shift, Gotham programs could face headwinds.
What to Watch Next
(1) Q2 earnings: Revenue, government growth, Apollo bookings. (2) Gotham customer commentary: new task orders or program wins. (3) AI product updates: generative capabilities applied to real classified workflows. (4) Apollo traction: customer announcements, bookings growth. (5) Government budget: obligated funding levels for Gotham programs.
Note: Government software spending is subject to budget appropriations and contract renewals. Investors should monitor Q2 2026 earnings for Gotham customer wins, Apollo commercial traction, and government segment growth rates. Insider ownership levels and management commentary on AI product adoption are important validation signals.
Additional validation milestones for investors: Monitor public filings (SEC 8-K/10-K), earnings call transcripts, and industry commentary for early signals of demand acceleration or competitive pressure. The stock's performance will ultimately depend on execution during the next 12–18 months; quarterly results and forward guidance are the primary signals to track. Conviction in the thesis should be updated quarterly based on actual business metrics rather than macro sentiment alone.