Palantir Technologies Inc. (PLTR)

Deep Dive Research Report

Research Date: May 16, 2026 | NYSE: PLTR | Sector: Technology

1. What the Company Does

Palantir builds software that turns fragmented, incompatible enterprise data into operational decisions at scale - for governments fighting wars and for corporations trying to run their businesses. The pitch sounds simple. The execution is among the hardest things in enterprise software.

The company was founded in 2003 by Peter Thiel, Alex Karp, Stephen Cohen, Joe Lonsdale, and Nathan Gettings - most of them connected through Thiel's PayPal network. The founding impulse was post-9/11. The intelligence failure of September 2001 was not a lack of data; the signals were there. It was a failure to connect the dots across fragmented, siloed systems. Thiel and his co-founders believed that PayPal's fraud detection logic - which caught illicit transactions by finding anomalous patterns across millions of data points - could do for counterterrorism what it had done for financial crime. They named the company after the palantiri, the seeing stones in Tolkien's Middle-earth that allowed their bearers to see events at great distances.

The CIA's venture arm, In-Q-Tel, seeded the company with roughly $2 million in 2005, validating the intelligence-community hypothesis and opening doors to classified environments that most commercial software vendors could never enter. For its first decade, Palantir was effectively a defense and intelligence contractor. Gotham, the first platform, was deployed in Iraq and Afghanistan, fusing signals intelligence, imagery, human intelligence, and open-source data to help analysts find patterns. A famous early result: Gotham identified the connection between garage door openers and improvised explosive devices - a pattern buried in millions of unconnected intelligence reports that human analysts had missed.

The pivotal transformation came in 2017. Palantir's engineers realized they had been operating on a false assumption. They had built Gotham presupposing that customers had integrated data. In practice, the data was a disaster everywhere - ERP exports sitting in SharePoint, databases maintained by one analyst who left two years ago, twelve different definitions of "customer" spread across twelve systems, and critical business knowledge living entirely in Excel spreadsheets no one had documented. The real value Palantir was delivering was not analytics. It was data integration. And data integration, they realized, was a general-purpose problem that every large organization in the world had. This insight produced Foundry - the commercial version of the platform, designed to solve data integration for industrial, healthcare, financial, and government civilian customers.

In 2023, as large language models became commercially viable, Palantir launched AIP - the Artificial Intelligence Platform - which brought LLMs directly into its governance-first architecture. Rather than letting customers "chat with their data" - the approach of most competitors - AIP layered LLMs onto the Ontology, Palantir's semantic model of the business, so that AI agents could not merely retrieve information but propose and take real actions on real business objects within a governed, auditable framework.

The mechanism that makes all of this hard to replicate is the Palantir Ontology. Think of it as a digital twin of an organization. Every meaningful entity in the business - a patient, a contract, a weapons platform, a supplier, a field officer - is represented in the Ontology as an Object, with typed properties, defined relationships to other objects, and explicit Action Types that govern what can be done to it. Data pipelines from every underlying source - SAP, Salesforce, Oracle, legacy ERPs, IoT sensors, satellite feeds - flow into this semantic layer. When an analyst queries the system, or an AI agent reasons about a decision, they work with Ontology objects, not raw database tables. LLMs operating in AIP don't hallucinate about raw data - they reason about structured, governed, computable representations of the business.

The way Palantir installs itself in an organization is through Forward Deployed Engineers (FDEs). Unlike a typical SaaS vendor that ships a product and provides documentation, Palantir embeds a team of engineers directly inside the customer organization for months or years. These engineers observe how decisions are actually made, model the customer's operations in the Ontology, build applications on top, and solve real problems in production. The result is deep institutional integration. A competitor who arrives later with a technically equivalent product faces not just a feature comparison but years of embedded customer-specific Ontology knowledge they would need to rebuild from scratch.

In 2023, Palantir added the AIP Bootcamp to accelerate the initial FDE engagement. A Bootcamp is a 2-5 day intensive workshop where a potential customer brings their own data, and with Palantir engineers guiding them, builds a working AI application solving a real problem. Over 1,300 Bootcamps had been completed through Q4 2025. The conversion rate from Bootcamp to paid contract is reported at 30-40%. What a Bootcamp produces that no sales presentation can is visceral: the customer's own people, working on their own data, seeing a problem solved that they've been unable to solve for years.

"Almost every single highlighted example of AI actually producing results in the U.S. is actually Palantir Technologies."

• CEO Alexander Karp, Q1 2026 earnings call, May 4, 2026

2. Business Segments

Palantir reports across two primary axes: Government versus Commercial customers, and US versus International geography. The 10-K presents four reportable sub-segments. In FY 2025, the mix was approximately 54% Government and 46% Commercial; 74% US and 26% International.

US Government

The US Government segment represents 41% of FY 2025 revenue and is Palantir's oldest, most stable business. It encompasses contracts with the Department of Defense, intelligence community, federal civilian agencies, and law enforcement. Gotham and Apollo are the primary products here, with AIP accelerating as an overlay as the Pentagon modernizes its AI capabilities.

The core capability this segment represents is years of classified-environment infrastructure, security clearance depth, and a track record of mission-critical performance in active combat theaters. Getting software accredited to run in Top Secret/SCI networks takes years and costs millions. Most commercial software vendors have never attempted it. Palantir's products operate across the full classification spectrum, from unclassified through the highest levels. Apollo, the deployment platform, handles continuous software delivery to air-gapped classified networks where the public internet does not exist, pushing updates in an average of 3.5 minutes per environment.

Growth in this segment was 55% in FY 2025 and accelerated to 84% in Q1 2026. The acceleration reflects a structural shift: the US military is no longer running experiments with AI. It is standardizing on it. The $10 billion Army Enterprise Agreement signed July 31, 2025, consolidated 75 separate contracts into one framework contract and formally made Palantir the Army's primary software vendor for data integration and AI. A subsequent Army directive mandated that all Army data operations consolidate on Palantir's Vantage platform. Maven Smart System - Palantir's AI-powered intelligence targeting platform - had usage double in four months as of Q1 2026, and the DoD confirmed in March 2026 that Maven will become a formal program of record by September 2026, transforming it from a discretionary experiment into a permanently funded capability.

The segment's competitive position is strong and structurally defended. Booz Allen Hamilton, CACI, and Leidos have government relationships that Palantir lacks, but their software product capability is significantly weaker. None have built an equivalent data-integration-plus-Ontology platform. Anduril's Lattice is the credible emerging competitor - well-funded, technically capable, with strong DoD relationships - but its focus is autonomous weapons systems management, which overlaps with Gotham only in narrow command-and-control scenarios. The two can coexist in a DoD network; the question is whether Anduril's ambitions expand to broader intelligence and data fusion.

US Commercial

The US Commercial segment represents 33% of FY 2025 revenue and is the primary growth engine. It delivers Foundry and AIP to American enterprises across healthcare, manufacturing, financial services, energy, and other sectors.

US Commercial revenue grew 109% in FY 2025 and 133% in Q1 2026. No software company of comparable scale is growing at this rate. The Bootcamp model is the mechanism. By removing the friction of a 6-12 month sales cycle and replacing it with a 2-5 day demonstration on the customer's own data, Palantir has expanded its US commercial customer base from approximately 295 in Q4 2023 to over 590 by end of FY 2025. Customer count reached 1,007 in Q1 2026, the first time Palantir has had more than 1,000 customers.

Net Dollar Retention of 150% in Q1 2026 is the key metric. It means existing customers - those already on the platform - spent 50% more in the trailing year than the prior year. That's not passive renewal. That's active expansion into more use cases, more business units, more decision workflows. Companies that go through a Bootcamp for supply chain AI and sign a $15M contract often expand to $80-100M deals within 18 months as the Ontology, once built, becomes the foundation for every subsequent AI application in the enterprise.

Major Q4 2025 wins included a $96 million healthcare deal, an $80 million engineering services contract, and GE Aerospace as a flagship enterprise customer publicly deploying AIP as its enterprise AI operating layer.

The competitive position in this segment is differentiated but contested. No single competitor offers the combination of deep multi-source data integration, an Ontology governance layer, and an AI execution framework in one product. But Microsoft, Databricks, and Snowflake are all moving toward this combination from adjacent positions, and the gap is narrowing.

International Government

International Government represents 12% of FY 2025 revenue and covers defense, intelligence, and law enforcement contracts with non-US allied governments. Primary markets include the UK, Germany, France, NATO member states, and increasingly the Middle East.

This segment grew 47% in FY 2025 - healthy but well below the US pace. The structural constraints are real. Palantir's software was built for and within American legal and security frameworks. European governments face political pressure around data sovereignty, civil liberties concerns (a German court blocked certain Gotham deployments on privacy grounds, ruling that profiling individuals not accused of crimes violates their privacy), and regulatory friction around GDPR implementation in intelligence contexts.

The Middle East is a more dynamic market. Defense spending in Saudi Arabia, UAE, and Israel is rising sharply, and intelligence software procurement is accelerating. Rosenblatt Securities noted in March 2026 that Middle East tensions are a specific pipeline tailwind for Palantir's government business in the region, citing demand for targeting, border security, and threat intelligence systems.

International Commercial

International Commercial is Palantir's weakest segment at 14% of FY 2025 revenue. The 2% full-year growth in FY 2025 followed a year where some individual quarters were negative year-over-year. The reasons combine macro and structural factors: Europe's GDP growth was 0.8% in 2024, suppressing corporate technology spending; GDPR and data residency requirements complicate Palantir's typical deep-integration approach; and Palantir's associations with US defense and immigration enforcement have created reputational friction with some European buyers.

On the Q3 2025 call, CEO Karp was unusually direct: "The region is going through a very structural change and doesn't quite get AI." Palantir has responded by consciously deprioritizing European commercial expansion and concentrating FDE resources on the US opportunity.

This segment is not a strategic focus. It represents legacy deployments and opportunistic wins. Management has been transparent about this and the market does not appear to price in meaningful international commercial recovery.

Segment Summary Table

3. Products and Business Detail

Gotham

Gotham is Palantir's original product, built for the intelligence community and deployed in classified networks since the mid-2000s. It functions as the operating system for defense decision-making. The platform fuses data from multiple intelligence sources - signals intelligence (SIGINT), imagery intelligence (IMINT), human intelligence (HUMINT), open-source intelligence (OSINT), geospatial feeds, and IoT/sensor streams - into a unified operational picture.

The security architecture is what makes Gotham genuinely hard to replicate. It enforces granular access controls based on classification level, need-to-know restrictions, organizational affiliation, and time-bound access windows, while maintaining a comprehensive audit log of every interaction with every piece of data. An intelligence analyst at one agency can see exactly what they are cleared to see and nothing more. A commanding officer can see their order of battle. A targeting cell can see the target package. None of them can see each other's data unless explicitly authorized - and every access is logged.

Gotham powers several of the most strategically important US military programs active today:

Maven Smart System: The DoD's primary AI-powered intelligence analysis and targeting platform, now used across combatant commands. Usage doubled in four months and grew 4x year-over-year as of Q1 2026. Maven is being formalized as a program of record by September 2026, elevating it from an experiment to a permanently funded capability.

TITAN (Tactical Intelligence Targeting Access Node): The Army's first AI-defined intelligence ground vehicle - a mobile system that processes multi-source intelligence using Gotham-derived software to support targeting decisions at the tactical edge. Palantir delivered the first two TITAN systems in early 2025 under a $178M contract. TITAN represents the edge-deployment of Gotham: not a server rack in a secure facility but an AI-enabled vehicle deployable in contested environments.

Ship OS: Palantir's naval deployment of AIP and Foundry, which reduced the US Navy's manufacturing approval process from 200 hours to 15 seconds. The $448M Navy shipbuilding modernization contract signed in Q4 2025 is built on this capability.

Foundry

Foundry is the commercial-and-civilian-government version of Palantir's data integration and Ontology platform. It was designed to solve a universal enterprise problem: critical operational decisions are being made using spreadsheets because no software has successfully integrated data from SAP, Salesforce, Oracle, custom ERPs, and departmental databases into a coherent, actionable picture of the business.

Foundry operates in a multi-stage process. First, it connects to existing data sources through hundreds of pre-built connectors plus a flexible ETL pipeline. It materializes raw data into a governed data environment. Then it models the business in the Ontology - the differentiating step - creating typed, governed business objects (Order, Patient, Supplier, Asset, Component, Pilot, Vessel) that represent real-world entities with their properties and relationships. Finally, it provides a development environment where analysts, data scientists, and engineers can build applications on top of these Ontology objects without needing to understand the underlying data complexity.

A manufacturing company using Foundry can answer: "Show me all suppliers whose components face supply chain disruption, ranked by downstream production impact, with the specific purchase orders at risk" - in real time, through an application any plant manager can use. Previously, answering that question required a data analyst spending two weeks pulling reports from five systems.

Foundry is deeply embedded in operations. Once a company has built its Ontology - typically 3-18 months of intensive FDE work - the semantic model encodes how the organization thinks about itself, in a form that doesn't exist in any underlying system. Switching to a competitor means rebuilding that from scratch. This is the core switching cost, and it is substantial.

AIP (Artificial Intelligence Platform)

AIP, launched April 2023, layers large language models directly into the Foundry and Gotham Ontology architecture. The technical problem it solves is that LLMs deployed naively in enterprise settings are unreliable, ungoverned, and unsafe. If an LLM queries raw enterprise data, it produces hallucinations, inconsistent outputs, access control violations, and outputs that cannot be audited.

AIP's architecture addresses these problems by confining LLMs to operate within the Ontology's governed framework. The LLM doesn't query raw databases - it reasons about Ontology objects whose properties are curated, typed, and access-controlled. It can only take actions through explicitly defined Action Types. Every LLM action is logged in the Ontology's audit trail. The result is AI that can operate in regulated industries (healthcare, finance, defense) and in workflows where accountability and auditability are non-negotiable.

CTO Shyam Sankar described AIP's positioning precisely in the Q2 2025 call: "AIP isn't just software our customers use, it's software our customers are building their software on." Companies are not deploying AIP as one feature in their IT stack - they are re-platforming their entire AI strategy onto it.

The 2025-2026 generation of AIP has added two significant capabilities:

AI Hivemind (launched Q3 2025): Multi-agent AI architecture that orchestrates swarms of specialized agents to divide complex tasks, work in parallel, and synthesize results. Instead of one AI agent answering a question, Hivemind deploys a coordinated team of agents - each specialized in a domain - working simultaneously on different aspects of a problem. This is the architectural foundation for AI-driven workflow automation at enterprise scale.

AIFDE (AI Forward Deployed Engineer) (launched Q3 2025): An AI-native development agent that can rapidly build new Ontology applications by understanding customer requirements and generating code. AIFDE is designed to compress what previously required 6-12 months of FDE work into weeks. If it performs as described, it addresses the FDE capacity constraint that is one of the primary limits on Palantir's growth rate.

Apollo

Apollo is Palantir's deployment and infrastructure management platform - the least visible product but strategically critical. It manages software delivery across every environment Palantir operates in, from AWS commercial cloud to a US Navy aircraft carrier to an air-gapped facility where no internet connectivity exists, from a single control plane.

Traditional CI/CD pipelines assume internet connectivity and a common infrastructure. Apollo was built for environments where those assumptions fail. It delivers software to disconnected, air-gapped, and physically isolated environments with all artifacts cryptographically signed, transfer integrity validated, and every change fully auditable. It maintains the change control records required for FedRAMP, IL5, and IL6 accreditation automatically, handling what would otherwise be months of manual compliance documentation per deployment.

Apollo's average software update time across all environments is 3.5 minutes. Without Apollo, each new government classified deployment would require months of manual accreditation. With Apollo, it takes weeks. This operational capability is invisible to customers in normal operation but becomes existential in environments where a software vulnerability needs to be patched immediately and the network is a submarine running a classified mission.

The Ontology: The Technical Moat

The Ontology underpins every product and is the deepest source of competitive advantage. It deserves a separate treatment.

The Ontology is not a database schema, not a data catalog, not a knowledge graph. It is an operational model of an enterprise - a living, computable representation of what an organization is and how it works.

The Ontology captures three categories of knowledge simultaneously:

Semantic elements: Objects (entities like Patient, Order, Weapon System, Supplier), Properties (the typed attributes of each object), and Links (the governed relationships between objects). This is the knowledge layer - how the organization understands itself.

Kinetic elements: Action Types (the operations that can be performed on objects - "approve this purchase order," "flag this target," "escalate this patient") and Functions (the business logic that evaluates decisions). This is the decision layer - how the organization acts.

Governance layer: Fine-grained access controls, audit trails, data lineage, and change management. This is the control layer - how the organization ensures accountability.

When LLMs operate within AIP, they reason about Ontology objects, not raw data. When analysts build applications in Foundry, they drag Ontology objects onto a canvas, not SQL tables. When a defense analyst in Gotham runs a pattern-of-life analysis, they reason about Ontology-modeled entities - persons, vehicles, locations, communication records - not raw signal data.

The strategic insight from Stratechery analyst Ben Thompson, who identified Palantir as the company most positioned for enterprise AI in 2024, frames this well: Palantir has solved the enterprise's foundational problem - not analytics, but the data integration that makes analytics possible. Thompson noted that competitors claiming to solve this automatically are actually requiring significant human engineering work, which subtly proves Palantir's point about the underlying complexity.

What makes the Ontology hard to replicate for any given customer is not the architecture but the specific instance. It took Palantir's FDEs months or years to build the Ontology for GE Aerospace or the US Army. That Ontology encodes institutional knowledge that doesn't exist in any single system. A competitor can replicate the software; they cannot replicate the customer-specific Ontology without starting from scratch.

The AIP Bootcamp Model

The Bootcamp is the go-to-market innovation that transformed Palantir's commercial growth rate. The structure is simple in design and powerful in effect.

Palantir invites 10-20 potential customer employees to a 2-5 day intensive session at a Palantir facility or on-site. Participants bring their own data. With Palantir engineers guiding them, they build a working AI application that addresses a real operational problem they've been unable to solve - in days. The conversion rate from Bootcamp to paid contract is reported at 30-40%.

The power of the Bootcamp is not the demo. It's the experience. Participants don't see Palantir's software working on sanitized example data. They experience their own data, in their own business context, producing an output they immediately recognize as operationally relevant. By the end of Day 3, there are typically multiple Palantir champions inside the customer organization who have personally built something real.

Over 1,300 Bootcamps were completed through Q4 2025. US commercial customer count grew approximately 100% from Q4 2023 to Q4 2025. Karp noted on the Q1 2026 call that the company operates with only "seven salespeople who actually really sell" - meaning the Bootcamp model is effectively self-propelling.

4. Customers

Government Customer Dynamics

US Government buying decisions operate at two levels. At the program level, a program manager and the relevant CTO office define requirements, issue RFPs, and evaluate vendors. The evaluation criteria are: security accreditation, classified-environment operational track record, mission-critical performance history, and integration with existing military systems. Palantir wins on all four.

At the enterprise level - the model pioneered by the Army EA - a Chief Information Officer and senior service leadership decide to standardize the entire organization. This is a fundamentally different decision. It's not "buy this software for this program." It's "we are re-platforming the Army's data infrastructure." The Army CIO explicitly positioned the EA as a template for how the DoD should procure commercial software - suggesting the Navy, Air Force, Marine Corps, and Space Force may follow the same model.

Sales cycles in government run 6-24 months for new contracts. Enterprise agreements like the Army deal took years to negotiate. But the resulting contracts have extraordinary durability. TITAN and Maven have been repeatedly extended and expanded since their inception. Government contracts are structured as multi-year enterprise agreements, often with annual option periods or ceiling-based consumption frameworks. The $10B Army EA has a 10-year term. The Maven contract ceiling was raised by $795M in May 2025 alone.

Contract revenues are predominantly software subscriptions combined with professional services (FDE personnel embedded in the customer). This means government revenue is not purely recurring SaaS - there is a services component tied to deployed headcount. But the software component is genuinely sticky, and the services component creates additional lock-in because Palantir's FDEs become integral to the customer's technical operations.

Commercial Customer Dynamics

Commercial buying decisions begin at the VP or C-suite level. The typical sponsor of an initial Bootcamp is a Chief Digital Officer, VP of Manufacturing, Head of Supply Chain, or Chief Operating Officer - someone with both operational authority and technology mandate. The decision to proceed to a multi-year enterprise contract requires CFO and CEO involvement as the investment grows.

The reason commercial customers choose Palantir over alternatives comes down to one thing: the others don't do what Palantir does. Databricks and Snowflake are excellent at data engineering and transformation but don't model business semantics. Microsoft Copilot overlays LLMs on existing Microsoft products but doesn't integrate non-Microsoft data or provide Ontology governance. C3.ai provides vertical AI applications but doesn't offer general-purpose integration. ServiceNow and Salesforce manage workflows and CRM but don't build the cross-system operational picture that Palantir provides.

Switching costs in commercial are among the highest in enterprise software. A company that has modeled its operations in the Palantir Ontology has invested 12-18 months and millions of dollars in FDE work to get there. That Ontology encodes how the organization thinks about its own operations. Switching means rebuilding that from scratch on a competitor platform, while simultaneously unwinding the workflows, dashboards, and AI applications built on top of it. Applications built on the Palantir Ontology become load-bearing infrastructure - they touch daily workflows of hundreds of people. No rational CFO authorizes this replatforming without an extreme reason.

The quantitative evidence for switching costs is Net Dollar Retention of 150% in Q1 2026. Existing customers spent 50% more in the trailing year than the year prior - not because of price increases but because they are expanding to more use cases. A company that signed a $15M supply chain AI contract in 2024 is signing an $80M enterprise expansion contract in 2025 because the Ontology, once built, is the cheapest starting point for every subsequent application.

As of December 31, 2025, average revenue from Palantir's top 20 customers was $93.9M trailing 12 months, up from $64.6M the prior year - a 45% increase in average spend among the largest customers.

5. Competitive Landscape

Government Competitors

Booz Allen Hamilton, CACI International, Leidos, Perspecta/Peraton: The legacy defense IT services firms. Their strategic advantage is deep government relationships, large workforces with security clearances, and the ability to prime large programs requiring extensive services delivery. Their weakness is software product depth. These companies built their businesses on time-and-materials contracts, not product development cycles. When the DoD evaluates AI software capability against Palantir's Ontology and FDE model, the contractors' offerings fall short. The Army EA explicitly evaluated and chose Palantir's commercial software over anything these firms offer. They often end up as subcontractors or system integrators around Palantir's software rather than competitors to it.

Anduril Industries: The most credible emerging competitor in the government segment. Anduril's Lattice is a genuinely powerful command-and-control platform for autonomous weapons systems, and Anduril has been winning significant DoD contracts for drone platforms (Replicator program) and autonomous systems management. Lattice's direct competition with Gotham is currently limited - Lattice excels at managing autonomous weapons; Gotham excels at intelligence fusion and analysis. But Anduril has stated ambitions to position Lattice as a universal command system, and if it extends into broader intelligence and data fusion, the overlap with Gotham grows. Anduril is well-funded (private), has excellent DoD relationships, and is led by a team (Palmer Luckey, Brian Schimpf) with genuine product credibility. This is a competitor to watch closely over the next 3-5 years.

Microsoft (Azure Government, M365 Copilot for Government): Microsoft has FedRAMP authorization, enormous Azure Government infrastructure, and the DoD's existing Microsoft Office ecosystem to leverage. The strategic limitation: Microsoft's government AI story is adding Copilot to products enterprises already use. It doesn't provide the cross-system Ontology layer, the classified-environment accreditation at the depth Palantir operates, or the FDE deployment model. The DoD uses both - Microsoft for productivity, Palantir for operational intelligence and decision-making. These can coexist, and likely will.

Google Cloud (Defense and Intelligence): Competing in the commercial space and attempting to establish presence in government AI. Google CEO Thomas Kurian has claimed Google's AI can automatically extract and map enterprise data relationships, making a separate Ontology unnecessary. Thompson's Stratechery analysis, confirmed by reviewing the claim carefully, notes that Google's "automatic" data mapping still requires significant human engineering work - which subtly vindicates Palantir's point about the underlying complexity of enterprise data.

Commercial Competitors

Databricks: The strongest platform for large-scale data and ML workloads. Databricks' Lakehouse architecture is genuinely excellent for data teams building and training models at scale. Its limitation for Palantir's use case is that Databricks requires significant data engineering expertise to operate, doesn't provide a semantic Ontology layer, and doesn't offer the end-to-end operational application development environment that Foundry provides. Databricks is a data platform for data teams; Palantir is an operational platform for business decision-makers. Both can coexist in the same enterprise - many customers use Databricks for data processing and Palantir for decision-making applications on top.

Snowflake: Excellent for data storage, sharing, and SQL-accessible analytics. Similar competitive dynamic to Databricks - strong at the data layer, not at the operational application layer. Snowflake's Data Cloud provides excellent data accessibility across organizations but doesn't model business semantics or provide AI governance.

Microsoft (Azure, Fabric, Copilot): The existential-level commercial competitor. Microsoft has Azure, Power BI, Teams, Dynamics, and now Fabric (its unified analytics platform) plus Copilot. The scale advantage is enormous. The competitive response from Palantir is that Microsoft's approach is additive - add AI capabilities to existing Microsoft products - while Palantir's is transformative - re-platform the organization's decision-making on the Ontology. Enterprises deeply embedded in the Microsoft ecosystem will face pressure to consolidate AI spend on Microsoft platforms. This is a real and growing competitive risk over the 3-5 year horizon.

ServiceNow, Salesforce, SAP: Workflow, CRM, and ERP vendors adding AI features. Palantir's positioning is that it doesn't replace these systems - it sits above them as the integration and decision layer, pulling data from all of them into the Ontology. This is a complement narrative, and it appears to hold in practice: most large Palantir commercial customers also use SAP and Salesforce.

C3.ai: More direct competition in enterprise AI applications. C3.ai offers pre-built AI applications for predictive maintenance, supply chain optimization, and defense analytics. It's faster to deploy for specific verticals and use cases but doesn't offer Palantir's general-purpose integration capability. C3.ai competes in scenarios where a company wants a specific pre-built application rather than a ground-up data integration project.

Barriers to Entry

The barriers to replicating Palantir as a whole are high. Building the Ontology architecture, the Apollo deployment infrastructure, classified-environment accreditation across multiple government clients, and the FDE model takes many years and hundreds of millions in capital. The most important barrier is the installed base: each customer's Ontology is a unique, customer-specific asset built over years of FDE work that any new entrant would need to rebuild.

However, the barriers are not insurmountable. A well-funded hyperscaler (Microsoft, Google, AWS) could theoretically build the architecture. The true moat is not the architecture but the installed base of customer Ontologies plus the classified-environment infrastructure - and hyperscalers already have substantial government presence. The competitive risk is highest in commercial; the classified-environment accreditation and FDE track record provide more durable protection in government.

6. Industry

What Drives Demand

Three distinct demand forces converge for Palantir.

Defense and intelligence modernization: The US military has formally identified AI-enabled decision-making as a critical national security priority. The Joint Warfighting Concept requires AI-enabled command and control at every level. Russia's war in Ukraine, Middle East conflicts, and China's military buildup have all demonstrated that AI-enabled intelligence analysis and targeting outperforms legacy systems in active operational environments. The DoD's annual AI investment was approximately $1.8B in FY2024 and is growing at 15%+ annually. The formalization of Maven as a program of record in September 2026 marks a threshold moment: AI in defense is no longer a research investment, it is a fielded operational system.

Enterprise AI adoption maturity: Every major corporation is attempting to deploy AI into operations. The failure rate of enterprise AI pilots is commonly cited at 80-90%. The mechanism behind that failure rate is almost always the same: the underlying data is too fragmented, too dirty, and too siloed for an LLM to work effectively on it. As enterprises cycle through failed pilots that bolt LLMs onto existing data messes, they increasingly recognize they need the foundational data architecture - the Ontology layer - that Foundry provides. Karp described this dynamic on the Q2 2025 call: "Our customers are re-platforming from hyperscaler solutions to Palantir's opinionated AI stack."

US industrial reindustrialization: The Trump administration's industrial policy - tariff restructuring, defense spending increases, manufacturing reshoring incentives - is creating demand for supply chain analytics, manufacturing optimization, and government contracting intelligence. Palantir explicitly aligned itself with this theme in 2025, positioning AIP as the tool that helps American manufacturers compete and citing the USDA farmland security contract as an example. The $300M USDA contract announced in Q1 2026 suggests civilian federal agencies are now actively pursuing the same AI modernization the DoD has been doing since 2017.

Market Size

The enterprise AI software market is estimated at approximately $114.87 billion in 2026, growing at approximately 19% CAGR toward an estimated $273 billion by 2031. The broader AI software market (including development tools and infrastructure) was approximately $174 billion in 2025, growing at 25% CAGR. These are large market estimates from research firms, and the exact figures vary substantially by methodology - the relevant point is that all major research firms place the market in the hundreds of billions and growing at double-digit rates.

Defense AI spending is a sub-segment but is growing faster than the enterprise average, driven by explicit government policy mandates and urgency from active operational use. The Maven contract ceiling alone is now over $1.3 billion; the Army EA ceiling is $10 billion over 10 years. These are not market estimates but actual contracted ceilings.

Palantir's own management describes their addressable market as "the operating system for the modern enterprise" - a framing that places the TAM at the entire enterprise software market, estimated in the trillions. The near-term constraint on Palantir's addressable market is not market size but the company's capacity to deploy FDEs and run Bootcamps.

Cyclicality and Structural Dynamics

Government defense contracts are relatively acyclical. National security spending tends to hold through economic downturns, and in the current geopolitical environment is actively expanding. The multi-year nature of enterprise government contracts (10 years for the Army EA) further insulates this segment from economic cycles.

Commercial enterprise software shows moderate cyclicality - enterprise IT budgets compress in recessions. However, Palantir's positioning as cost-reduction infrastructure rather than new capability spend provides a natural recession hedge. AI systems that replace headcount become more attractive when labor costs are under pressure. The healthcare deal and the industrial manufacturing wins are explicitly justified by cost reduction metrics that become more compelling, not less, in a downturn.

Regulatory dynamics matter in two primary ways. GDPR and European data sovereignty laws constrain the international commercial segment. US export control laws (ITAR/EAR) restrict which governments can receive Palantir's defense software. Both are real constraints but primarily affect segments that are already the weakest parts of the business.

7. Growth Triggers

All triggers sourced directly from the four concall transcripts. Every trigger is cited with the specific call date.

• Maven formalization as a Program of Record by September 2026: DoD committed in a March 2026 letter from Defense Under Secretary Steve Feinberg that Maven will become a formal program of record by September 2026. This transforms Maven from a discretionary experimental program into permanently funded defense infrastructure, removing the annual budget-cycle uncertainty that has periodically affected the contract. (Q1 2026 concall, May 4, 2026)

"Maven Smart System usage doubled over four months and increased 4x year-over-year." - Karp, Q1 2026

• Ship OS scaling across the full Navy enterprise: Palantir's Ship OS deployment reduced the Navy's manufacturing approval process from 200 hours to 15 seconds. The $448M Navy shipbuilding modernization contract signed in Q4 2025 is the vehicle for further expansion. Management cited active growth into additional naval facilities and programs. (Q1 2026 concall, May 4, 2026; Q4 2025 concall, Feb 2, 2026) [repeated]

• USDA $300M contract execution: A $300M contract with the US Department of Agriculture for farmland security and supply chain resilience was announced in Q1 2026. This represents Palantir's first major civilian federal agency win of this scale and opens a template for other civilian agencies (HHS, DHS, Transportation) to follow. (Q1 2026 concall, May 4, 2026)

• Army Vantage directive implementation: A formal Army directive was issued in Q3 2025 requiring all Army data operations to consolidate on Palantir's Vantage platform (built on Foundry and AIP). This is a mandatory directive, not a pilot. The operational ramp of this directive across hundreds of Army data systems is expected to drive US Government revenue for multiple years. (Q3 2025 concall, Nov 3, 2025)

"The US Army issued an official directive consolidating all data operations on Palantir's Vantage platform." - Q3 2025 concall, Nov 3, 2025

• Army Enterprise Agreement consumption ramp over 10 years: The $10B Army EA, signed July 2025, consolidated 75 contracts and established a 10-year consumption framework. Revenue recognized from the EA reflects actual usage against the ceiling, which management expects to grow significantly as Army units onboard. The Army CIO described the EA as a template for how other DoD services will procure commercial software. (Q2 2025 concall, Aug 4, 2025; Q3 2025 concall, Nov 3, 2025) [repeated]

• Bootcamp flywheel driving US commercial expansion into Fortune 500: Management explicitly stated the Fortune 500 addressable market is still largely unpenetrated. With over 1,300 Bootcamps completed and a 30-40% conversion rate, the funnel continues expanding. Karp stated in the Q1 2026 call that the company still operates with only "seven salespeople who actually really sell," implying growth is far from sales-capacity-constrained. (All 4 concalls) [repeated]

• AI Hivemind and AIFDE compressing deployment cycles: Launched in Q3 2025, Hivemind (multi-agent AI swarms) and AIFDE (AI-native application development) are expected to dramatically compress the FDE deployment timeline. If AIFDE can replace what previously required 6-12 months of FDE work with weeks of AI-assisted development, Palantir's throughput per customer grows substantially without proportional headcount growth. (Q3 2025 concall, Nov 3, 2025; Q1 2026 concall, May 4, 2026) [repeated]

• Net Dollar Retention expansion from existing customer base: NDR of 150% in Q1 2026 (up from 134% in Q3 2025, 139% in Q4 2025) means existing customers are already driving growth acceleration. Customers who deployed one AIP use case in 2024 are expanding to five to ten use cases in 2025-2026. Management expects this expansion dynamic to continue as the installed base matures. (Q1 2026 concall, May 4, 2026; Q4 2025 concall, Feb 2, 2026) [repeated]

Growth Trigger Summary

8. Key Risks

Stock-Based Compensation and Structural Dilution

Palantir has consistently issued substantial equity compensation: $565M in 2022, $476M in 2023, $692M in 2024, and $684M in 2025. Against a buyback program of $64M in 2024 and $75M in 2025, the net effect is persistent dilution. The company is effectively buying back approximately 10 cents of stock for every dollar issued to employees in equity. Basic shares outstanding have grown from approximately 2.06 billion in 2022 to approximately 2.25 billion by end of 2024, with diluted shares (including options and unvested RSUs) growing further. Over a four-to-five-year horizon, a 4-6% annual dilution rate materially impacts per-share value creation even as the aggregate business grows strongly.

Management has not given any indication that buyback scale will increase to offset SBC. The risk is structural and ongoing, not acute - but shareholders who hold for 5+ years will find their economic interest meaningfully diluted unless this changes.

US Government Budget Concentration

Approximately 54% of FY 2025 revenue was government-derived, predominantly US. Two specific mechanisms could disrupt this. First, DoD budget process risk: a continuing resolution that delays appropriations, a broad defense spending review, or a budget sequester would delay contract renewals and expansions. The Army EA mitigates some of this (it's a 10-year framework), but Maven's formalization as a program of record is not yet complete, and programs not yet on record remain subject to annual budget battles.

Second, DOGE-type efficiency reviews. The current administration's focus on reducing federal spending creates a non-trivial risk that certain Palantir contracts get reviewed, renegotiated, or canceled in the name of cost reduction. Paradoxically, if AI software demonstrably reduces DoD headcount (which is Palantir's pitch), an efficiency-focused administration might view AI investment favorably. But the near-term risk of contract disruption during any review period is real.

International Commercial Stagnation

International Commercial grew 2% in FY 2025 after being negative in individual quarters. GDPR enforcement, data sovereignty mandates, and reputational friction from US defense and immigration enforcement associations collectively constrain Palantir's European commercial addressable market. The risk is not that this segment shrinks dramatically, but that it fails to recover as a meaningful growth contributor. If the international commercial segment remains flat while the US drives everything, Palantir's story becomes entirely dependent on the US - which is extraordinary today but creates geographic single-point-of-failure risk for long-term investors.

Hyperscaler Capability Convergence

Microsoft, Google, and AWS each have engineering resources that dwarf Palantir's headcount. All three are actively building toward unified enterprise AI platforms. Microsoft's Fabric platform offers data integration and analytics across the Microsoft ecosystem, and Copilot adds an LLM layer. If Microsoft adds a semantic modeling layer equivalent to the Ontology within Azure, and bundles it into the M365/Azure Enterprise Agreement that Fortune 500 companies are already paying hundreds of millions annually for, Palantir's commercial market differentiation narrows substantially.

The most likely path is not direct replacement but scope compression: Microsoft handles the Ontology for Microsoft-native data (which is already a large share of most enterprises), and Palantir retains a narrower position for multi-vendor, classified, or highly specialized operational use cases. The commercial TAM looks smaller in that scenario.

Valuation Expectations Risk

Palantir's stock is priced for sustained hypergrowth. Q1 2026 came in at 85% growth with the guidance beat of approximately $98M. Consensus expectations are now recalibrated to that level. Any quarter where growth decelerates meaningfully - from macro weakness, competitive pressure, or the law of large numbers - could produce a sharp repricing. This is not a business risk but a market expectations risk. The business generating 40% growth with 50%+ FCF margins is excellent. The repricing from "85% growth" to "40% growth" would be severe at current multiples, even though the underlying company would still be exceptional.

Key Person Concentration

Alex Karp's public persona - philosophical, contrarian, genuinely eccentric - is inseparable from Palantir's brand and culture. His relationships with NATO leadership, senior US officials, and European defense establishments are personal, developed over 20 years. Peter Thiel's political network opened doors in the intelligence community that no institutional relationship could have. Shyam Sankar is the CTO who designed AIP and whose technical vision shapes the product roadmap. None of these are easily replaced. The company has a strong second layer of leadership, and the Ontology platform is not person-dependent. But a departure of Karp or a rupture in the Karp-Thiel relationship would create strategic uncertainty that markets would price adversely.

FDE Capacity Constraint

The FDE model is powerful and also a capacity bottleneck. FDEs are senior engineers embedded in customer environments - expensive to hire, hard to train, and limited in supply. If commercial demand grows faster than FDE capacity, sales cycles lengthen or the quality of deployment decreases. AIFDE is designed to address this by compressing deployment timelines through AI assistance. If AIFDE delivers on its promise, this risk diminishes. If it underperforms, the FDE constraint becomes the binding limit on how many new commercial customers Palantir can serve simultaneously.

9. Walk the Talk

Concalls used:

1. Q2 2025 - August 4, 2025
2. Q3 2025 - November 3, 2025
3. Q4 2025 - February 2, 2026
4. Q1 2026 - May 4, 2026 (12 days ago)

Q2 2025 (August 4, 2025) - Guidance Set

On the Q2 2025 call, Palantir reported its first-ever billion-dollar quarter ($1.004B, +48% YoY) and set Q3 2025 guidance at $1.083-$1.087 billion - representing approximately 50% YoY growth at the midpoint. CEO Karp framed the company as at a "watershed moment" in enterprise AI adoption and said: "Our primary sales force now... are going to be current customers telling other customers." CTO Shyam Sankar added that customers are "re-platforming from hyperscaler solutions to Palantir's opinionated AI stack" - a claim that, at the time, could have been dismissed as promotional language. The Army's $10B Enterprise Agreement had been signed weeks earlier, and the $795M Maven ceiling increase was fresh. Karp described these as structural inflection points, not one-time events.

The full-year 2025 guidance raised to $4.142-$4.150B implied 45% YoY growth for the full year. The US Commercial guidance was set at "exceed $1.302B" - at least 85% growth.

Q3 2025 (November 3, 2025) - First Delivery

Q3 actual revenue came in at $1.181 billion - beating the guided $1.083-1.087B range by approximately $96 million, a 9% beat above the high end of guidance. The company's previous record for a guidance beat was well below this magnitude. US Commercial grew 121% YoY, substantially above the 85% implied by full-year guidance at the time of the Q2 call.

Karp's statement that Q3 results were "arguably the best results that any software company has ever delivered" was not promotional - the Rule of 40 score of 114% was 20 points above the prior historical record for Palantir. The Army Vantage directive was disclosed publicly for the first time: all Army data operations mandated to consolidate on Palantir's platform. Q4 guidance was set at $1.329-$1.331B.

Q4 2025 (February 2, 2026) - Second Delivery

Q4 actual revenue was $1.407B, beating the guided $1.329-1.331B range by approximately $77M - another 6% beat above guidance. Full-year 2025 revenue was $4.475B, comfortably above the raised FY guide. US Commercial grew 137% YoY in Q4 alone.

Initial FY 2026 guidance of $7.182-$7.198B was set. Analysts immediately noted this was widely viewed as a conservative starting point given the momentum visible in the Q4 numbers. CRO Ryan Taylor: "We closed 61 deals over $10 million. That's because of the impact we're delivering." The $448M Navy shipbuilding contract was disclosed. Net Dollar Retention hit 139% - its own record at the time.

Q1 2026 (May 4, 2026) - Third Delivery and Escalation

Q1 actual revenue was $1.633B - beating the guided $1.532-$1.536B by approximately $98M, the largest absolute guidance beat in the company's public history. Growth accelerated from 70% in Q4 2025 to 85% in Q1 2026. US revenue crossed 100% YoY growth for the first time since the DPO.

Full-year 2026 guidance was immediately raised from $7.182-7.198B to $7.650-$7.662B - a raise of approximately $470M at the midpoint. US Commercial guidance for FY 2026 was raised from $3.073B to $3.224B+. Net Dollar Retention hit 150%.

The Verdict

The pattern across all four quarters is consistent, quantified, and unambiguous. Management sets guidance conservatively, beats by 6-9%, and raises the full-year guide materially. The magnitude of beats has increased each quarter: from $96M in Q3 2025, to $77M in Q4 2025 (a sequentially larger quarter), to $98M in Q1 2026. This is not a management team sandbagging by a small amount - these are material beats that have surprised even the most optimistic analyst consensus each time.

The one area where management signaled difficulty proactively and was vindicated: Karp's Q3 2025 acknowledgment that international commercial is structurally challenged in Europe. FY 2025 international commercial growth of 2% confirmed the warning was accurate. Management chose to be transparent rather than manage expectations.

Promise vs. Outcome Summary

This is management that consistently does what they say and typically does more. The under-promise/over-deliver pattern is structurally embedded in how CFO Glazer sets guidance. Whether this conservatism is deliberate strategy or genuine uncertainty about demand patterns, the result for shareholders is four consecutive quarters of positive earnings surprises.

10. Shareholder Friendliness Index

Dividends: Palantir has never paid a dividend since its direct listing in September 2020 and has no stated intention to initiate one. All three of the last available financial years show zero dividend per share.

Buybacks and dilution: Palantir operates a share repurchase program, but its scale is cosmetically small relative to SBC dilution. Buybacks totaled $64.2M in FY 2024 and $75.0M in FY 2025, a combined $139M over two years. Over the same period, stock-based compensation was $692M in 2024 and $684M in 2025, a combined $1.376B. For every dollar issued to employees in equity, Palantir bought back approximately 10 cents. Basic shares outstanding have grown approximately 4% annually from 2022 through 2024, and diluted shares (including unvested RSUs and options) have grown on a similar trajectory. The share count is growing, not shrinking.

Verdict: Hoards Capital. Palantir has never paid a dividend, operates a buyback program that repurchases a fraction of SBC dilution, and shareholders have experienced approximately 25% dilution in diluted share count over the four years since the DPO. Management deploys capital into growth and employee retention rather than shareholder returns.

11. Insider Activities

Primary source: SEC Form 4 filings via EDGAR, with OpenInsider.com as aggregator. All transactions cited with Form 4 filing references.

Recent Transactions (Most Recent First)

Buys - Reading the Signal

There were no open-market insider purchases in the last 12 months. No director, officer, or significant shareholder bought Palantir shares in the open market at prevailing prices during the period reviewed. This is the most notable data point in this section - not because it signals concern, but because it represents the absence of the strongest positive signal in insider analysis.

Sells - Working Out the Why

The simultaneous February 20, 2026 sales by Karp, Cohen, Sankar, Glazer, and Taylor are the largest cluster of selling activity by dollar value. The common pattern - identical date, Form 4 footnotes explicitly referencing "tax withholding obligations" on RSU vesting events - confirms these are automatic, formula-driven sales that occur at vesting dates, not discretionary decisions by insiders choosing to sell. When RSUs vest, a percentage is withheld to cover tax liability. The shares sold are not the insiders exiting their positions but paying their tax bills. These sales do not carry information about the executives' view of the company's outlook.

Peter Thiel's 2,000,000-share sale on March 2, 2026 ($289.7M) through a 10b5-1 plan adopted in November 2025 is the largest single transaction by value. Thiel is Palantir's largest individual shareholder, with a position accumulated over more than 20 years at near-zero cost basis. He has been systematically reducing his Palantir position over multiple years through pre-scheduled 10b5-1 plans, a pattern entirely consistent with a founding shareholder gradually diversifying a concentrated position that represents a disproportionate share of personal net worth. The plan was adopted months before execution in a closed window, the legally required form for this type of pre-planned monetization. It does not signal concerns about near-term performance.

Director Alexander Moore's recurring monthly sales (Feb, Mar, Apr 2026) through a December 2025 10b5-1 plan are mechanical diversification by a board member whose primary compensation in the role is equity. Monthly regularity is a classic feature of 10b5-1 plans established for systematic diversification.

The CFO's September 2025 transaction combined an option exercise with a larger sale. Selling shares beyond the option exercise to cover tax is common when significant unrealized gains have accumulated; it does not indicate a negative view.

Net Assessment

Insiders are net sellers over the last 12 months, but the selling is overwhelmingly explained by routine mechanisms: RSU tax withholding events (automatic at vesting) and pre-scheduled 10b5-1 plans adopted months in advance. There are no open-market discretionary sales that would suggest management concern about the business. The absence of any open-market buying is the most notable observation - no insider has added to their position at current prices. Given that the stock has appreciated dramatically over the prior two years, the lack of open-market buying is understandable without being alarming. Signal: Neutral. Selling is fully explained by compensation mechanics; no discretionary selling and no buying produces a net neutral read.

12. Scenarios

Bull Case

In the bull case, Palantir achieves what Alexander Karp has been describing since 2023: it becomes the operating system layer for the modern enterprise in the same way AWS became the infrastructure layer for the cloud. The Bootcamp flywheel, currently driving 133% US commercial growth, sustains above 80% for two more years as Palantir penetrates deeper into the Fortune 500. Companies that deployed one AIP use case in a Bootcamp in 2025 now have ten applications running on the Ontology, and their enterprise contracts have grown from $20M to $100M. Net Dollar Retention stays at or above 150%, meaning existing customers alone generate growth rates that most software companies would celebrate.

On the government side, the Army EA template spreads across the services. The Navy follows with its own enterprise agreement, consolidating its dozens of Palantir contracts. The Air Force and Space Force do the same. Maven's formalization as a program of record in September 2026 elevates AI-enabled intelligence from a research investment to a permanently funded operational system, insulating the revenue from annual budget uncertainty. TITAN vehicles receive a production contract for hundreds of units. Allied governments - UK, Australia, NATO members - accelerate defense AI procurement using US programs as the template.

The technology advantage compounds. AIFDE fulfills its promise and compresses FDE deployment from 6-12 months to 4-6 weeks. This removes the FDE capacity bottleneck entirely, allowing Palantir to serve 2x-3x as many commercial customers simultaneously with the same headcount. The combination of faster deployments and higher NDR creates a flywheel: more customers deploy faster, expand more quickly, and refer other enterprises more aggressively. The seven salespeople Karp mentioned become irrelevant because the product sells itself through results.

International commercial recovers modestly as European enterprises, forced to cut costs by continued macro weakness, look past the political friction and adopt AI solutions that demonstrably reduce headcount. The government segment in Europe stabilizes as NATO members accelerate defense spending under sustained US pressure. China's aggressive military posture keeps allied governments investing in intelligence AI.

Base Case

The most likely path is continued strong execution with growth moderating from hypergrowth toward sustained rapid growth as the base expands. US commercial growth gradually decelerates from 133% toward 60-70% by 2027 as the early adopter market fills and the next tier of enterprises requires more sales effort. But the installed base generates substantial expansion revenue through NDR, so total commercial growth remains robust even as new logo growth slows.

Government revenue grows steadily at 40-55% annually. The Army EA ramps consistently. Maven's program of record status removes budget uncertainty. Additional USDA-type civilian agency wins add to the federal civilian revenue base. Allied defense spending supports international government.

The free cash flow machine continues generating 50%+ FCF margins on a rapidly growing revenue base. AIFDE delivers partial benefit - shortening deployment timelines by 30-40% rather than the 80-90% that would eliminate the FDE constraint. Management deploys FCF into talent, international government expansion in the Middle East and Asia-Pacific, and selective AI capability investments.

International commercial remains flat to slightly positive, making no meaningful contribution to growth. It is not a problem - the US opportunity is large enough - but it is a permanent ceiling on international commercial scale.

Bear Case

The bear case is not a collapse but a meaningful growth deceleration that disappoints the expectations embedded in the current valuation. The most credible mechanism is Microsoft closing the capability gap.

Microsoft Fabric, which already integrates data across the Microsoft ecosystem, adds a semantic modeling layer equivalent to Palantir's Ontology in 2027. Microsoft bundles this into the Azure Enterprise Agreement that Fortune 500 companies are already paying hundreds of millions for. Enterprise CIOs, under pressure to consolidate technology spend, begin routing new AI projects through Microsoft's platform rather than going through a Palantir Bootcamp. Bootcamp conversion rates fall from 35% to 20%. New commercial logos per quarter drops from 100+ to 40-50.

Simultaneously, DoD budget reviews under continued DOGE pressure delay several contract renewals for two to three quarters, slowing US Government growth from 84% to 25%. The Army EA ramps slower than modeled because onboarding 75 program offices onto a new platform is harder than the agreement anticipated.

US commercial growth slows from 133% to 40%. US Government slows to 25%. International commercial remains flat. Total company growth decelerates from 85% to 35-40%. This is still excellent in absolute terms, but it is a violent repricing from the market's current expectations. Free cash flow remains strong - 40%+ margins on a $5-6B revenue base generates substantial cash. But the stock, priced for 70%+ growth, discovers what a 35-40% growth software company is worth, and the repricing is severe. Management, who have been selling shares through pre-planned 10b5-1 programs, proves to have chosen their exit timing wisely.

13. Further Reading

• Enterprise Philosophy and the First Wave of AI - Stratechery, Ben Thompson, 2024 [free] - Thompson argues Palantir is the company most positioned for the first wave of AI because enterprise AI requires deep data integration, not just LLMs, and Palantir's Ontology is the only production architecture designed for exactly this.

• The Deployment Company, Back to the 70s, Apple and Intel - Stratechery, Ben Thompson, 2026 [paid]

Sources:

• Palantir Q1 2026 Earnings Call Transcript - The Motley Fool
• Palantir Q4 2025 Earnings Call Transcript - The Motley Fool
• Palantir Q3 2025 Earnings Call Transcript - The Motley Fool
• Palantir Q2 2025 Earnings Call Transcript - The Motley Fool
• Palantir 2025 FY 10-K Annual Report
• Palantir PLTR Insider Trading Form 4 Filings
• Enterprise Philosophy and the First Wave of AI - Stratechery
• Palantir Q1 2026 Press Release - IR
• US Army $10B Enterprise Agreement - CNBC
• Maven Smart System Contract Expansion - DefenseScoop
• Palantir Ontology System Documentation
• PLTR Insider Trading - OpenInsider