Palantir Technologies – A Stunning Business, But a Hard Pass For Me (A Deep Dive)
This is my Palantir Technologies Deep Dive!
I have to be honest – up until recently, I never expected to be taking such a close look at Palantir Technologies. In recent years, it was simply never on my radar. The stock was a darling of the retail crowd, perpetually hyped on social media and trading at valuations that had no grounding in the underlying business — a bubble stock in the truest sense, priced for perfection years before the fundamentals remotely justified it. And what fundamentals there were didn’t exactly inspire confidence: a business generating mid-teens revenue growth, still heavily reliant on government and defense contracts, with a commercial business that seemed to promise a lot and deliver comparatively little. Layer on top of that the ethical concerns (ICE contracts, military targeting systems, civil liberties questions), and a founder-first governance structure that leaves outside shareholders with essentially no say, and Palantir was, for me, a very easy pass. I honestly never considered it.
Yet there is no denying the attractive nature of the business, its success in recent years, and, above all, that some of the hype has run out of the stock (notice it popping up less on your timeline?), with it not escaping the broader SaaS sell-off. Palantir shares are down 37% YTD and 22% over the last year, against a nearly doubling of TTM revenue. Safe to say Palantir stock has at least somewhat come back down to earth valuation-wise.
And that has drawn my attention, with Palantir’s business and performance undeniably impressive and promising, even when ethical questions remain.
You see, what changed my mind isn’t that the concerns disappeared — they haven’t, and I’ll address them head-on later in this piece. What changed is that the business underneath all the noise has become genuinely impossible to ignore. Eleven consecutive quarters of accelerating growth. Revenue nearly doubled in a year. A 150% net dollar retention rate. A 60% operating margin. A Rule of 40 score of 145%.
These aren’t the numbers of a hype-driven story held together by narrative and retail enthusiasm. They are the numbers of a business that is, right now, firing on all cylinders, with structural tailwinds that are only accelerating. Palantir spent two decades quietly building exactly the infrastructure the AI era turns out to need most, and the market is only now beginning to price that in properly.
So consider this my attempt at an honest, thorough look at a company I long dismissed — warts and all. I will cover the business model, the products, the competitive position, the financials, the concerns I still hold, and ultimately whether the valuation, at a multi-year low relative to fundamentals, makes a compelling enough case.
Without further ado, let’s delve in.
This is a paid-exclusive Deep Dive. Want full access to this 8000-word analysis? Consider upgrading to the paid tier for $9.50 a month or $95 a year! This gets you:
6–8 Research Reports per month, 2x more than free subs do, which includes 3-4 Deep Dives like this one! (that comes down to just $1 per analysis)
Full visibility into my personal portfolio, including allocations, transactions, and conviction levels
An overview of target prices and ratings, updated weekly
InvestInsights Premium is designed to help you find best-in-class compounders, identify deep-value opportunities, avoid costly mistakes, and stay focused on fundamentals when markets get noisy, with full transparency into how I make decisions over time.
This is Palantir Technologies!
Business Fundamentals
Founded in 2003 and going public in 2020 via a direct listing rather than a traditional IPO, Palantir Technologies is a software company specializing in big data analytics and artificial intelligence platforms, designed to help organizations integrate, manage, and act on massive, often siloed datasets.
The company emerged in the post-9/11 era with a founding mission rooted in counterterrorism and national security work, drawing early inspiration and funding from the CIA’s venture capital arm, but has since adopted a much broader vision, growing into a company with a $308 billion market cap, generating over $5 billion in revenue today, still under the lead of Alex Karp, one of the five original co-founders and CEO since its founding in 2003 – making Palantir founder-led. Also, Peter Thiel, another co-founder, remains chairman of the board and retains significant influence through the dual-class share structure, which concentrates voting power with the founders.
So, what is Palantir exactly? At its core, Palantir focuses on building software that helps large organizations, especially governments and big enterprises, make sense of massive, messy, fragmented data and actually act on it, rather than just visualize or report on it.
Palantir’s foundational value proposition is to take data scattered across dozens or hundreds of incompatible systems (think different databases, spreadsheets, sensors, legacy software) and fuse it into a single, coherent operational picture, giving decision-makers a live, unified view they can use to act in real time.
So, it isn’t just about taking the data, organizing it, and visualizing it – Palantir’s pitch is one step further. It wants to be the operational layer where that intelligence is directly wired into workflows and decisions. So instead of “here’s a dashboard showing your supply chain is delayed,” it’s closer to “here’s the delay, here’s why, here’s who’s affected, and here’s the workflow to reroute the shipment, reassign the team, or alert the relevant person, and an AI agent can even initiate some of those actions for you.”
This also means Palantir doesn’t sell a single product off a shelf, but software platforms that big, complicated organizations install and build on top of to make sense of their own data and run their operations better. It doesn’t sell a narrow tool like accounting software or CRM, but a central layer that ties everything together. Think of it less like buying an app and more like hiring a contractor to build the nervous system for your organization’s data, connecting everything that wasn’t talking to each other before (different databases, spreadsheets, machines, systems) into one usable picture.
Crucial in this is Palantir’s “ontology” approach, which means building a digital model of an organization’s real-world objects, relationships, and processes so that data isn’t just stored but meaningfully connected to how the organization actually operates. In other words, it is delivered not as a standard system but highly personalized for optimal ROI. This means the software can map the data not just as numbers in a table but as the actual real-world entities, with the software understanding what an action on that data means in the real world. So, ontology means Palantir takes complex, fragmented data, organizes it into a model of how an organization actually works, and uses that to support both human decision-making and automated or semi-automated action.
With this approach and focus, Palantir targets a very specific customer base, as Palantir’s value proposition only really makes sense for organizations with genuinely messy, siloed, high-volume data environments.
A small business running on QuickBooks and a single CRM doesn’t need Palantir, as there’s nothing to fuse. But a hospital system with separate EHR platforms, lab systems, billing software, and supply chain tools, or a military with classified intel feeds, logistics systems, and sensor networks that were never designed to talk to each other, that’s exactly the mess Palantir is built to solve.
On top of that, the customer has to face high enough stakes (operational, financial, or security) from getting decisions wrong and have deep enough pockets to justify Palantir’s cost and implementation effort. Palantir implementations are expensive and labor-intensive, so the customer needs a problem where bad decisions are catastrophically costly: a misallocated ICU bed in a hospital, a missed threat in a war zone, a factory line going down, a multi-billion-dollar supply chain disruption. That’s where the ROI on Palantir’s price tag and onboarding effort clears the bar.
As a result, Palantir skews toward governments, militaries, and large enterprises rather than, say, a mid-size logistics company with admittedly messy data. Simply, Palantir targets organizations sitting at the intersection of high data complexity and high-stakes operations
These companies have data that’s both massive and trapped, spread across systems and often accumulated over decades by different vendors, acquisitions, or bureaucratic silos. The problem is that this makes it incredibly hard to get a live, cross-departmental view, and at a time when data is increasingly valuable and a key method for improving operations, that lack of visibility is critical.
Palantir’s pitch is that it removes that blindness not by replacing existing systems but by sitting atop them and creating a unified, queryable, continuously updated layer.
Makes sense, right?
With this in mind, Palantir splits revenue into two main buckets: Government (64% of revenue and Commercial (36% of revenue).
Government (and military) covers defense departments, intelligence agencies, and law enforcement. This is where Palantir started – its bedrock – with deep roots in counterterrorism, intelligence analysis, and military operations. Its platforms are used for battlefield situational awareness, predictive maintenance of military equipment, and intelligence fusion across agencies. This is still the largest by revenue.
In terms of use cases, operation awareness and decision support are the most common, taking live data from many sources and giving decision-makers a real-time, unified view so they can act faster and with better information.
Commercial covers large enterprises, such as hospitals, manufacturers, energy companies, and banks – sectors that combine messy operational/physical-world data with a high cost of failure, a profile similar to their original defense/intelligence customers. It is used in hospitals for capacity planning, supply chain coordination, and clinical operations; in energy and utilities for grid management, pipeline operations, and increasingly with the broader energy infrastructure boom tied to AI data center demand; in manufacturing for production line optimization, supply chain visibility, and predictive maintenance (take a five year deal with Stellantis); in financial services for fraud detection, risk modeling, and compliance; and in aerospace for production and program management (including a multi-year deal with Airbus).
In recent years, the company has emphasized rapid growth in its commercial business, partly driven by enterprise demand for AIP as organizations seek practical ways to operationalize generative AI. This has fueled rapid growth here, with the segment likely to overtake Government within a year. This is where Palantir’s largest opportunity lies, especially as data becomes more important and AI adoption explodes.
So, that should give a pretty good idea of Palantir’s core business, strategic focus, and its customer base. Let’s then delve deeper into its products.
Ultimately, Palantir’s product suite centers on a handful of core platforms.
Palantir Gotham, which accounts for about 55% of revenue, is its original offering and is widely used by defense, intelligence, and law enforcement agencies to fuse disparate data sources for operational decision-making.
Gotham was built to ingest enormous volumes of structured and unstructured data, such as intercepted communications, satellite imagery, financial records, sensor feeds, and human intelligence reports, and fuse them into a coherent picture that analysts and operators can search, query, and visualize. Its signature capability is link analysis: surfacing non-obvious relationships between people, places, events, and objects across datasets that were never designed to talk to each other.
Gotham is used across a wide range of military and intelligence applications, including battlefield situational awareness, predictive maintenance for military hardware, counterterrorism investigation, and logistics planning. It is also used in law enforcement settings, where it has supported criminal investigations by linking records across databases.
Therefore, Gotham’s interface is built for analysts working in high-stakes, often classified environments, with heavy emphasis on data provenance, access controls, and audit trails, given the sensitivity of the information it handles.
Also falling under the Gotham banner is the Maven Smart System (MSS), which is Palantir’s flagship AI-enabled military platform used by over 20,000 users across 35 military services, including a $1.3 billion deal with the Pentagon through 2029.
But MSS is used well beyond the U.S., adopted by NATO’s Communications and Information Agency for use within NATO’s Allied Command Operations. The UK also announced a partnership with Palantir on September 18, 2025, to develop AI-powered military targeting and decision-making capabilities, worth up to £750 million over five years.
So, adoption has been strong, which isn’t surprising given that the system can replace what was previously run by nine separate systems, compressing certain operations from hours to minutes, which is highly valuable in these kinds of operations.
Besides Gotham, there is Palantir Foundry, which is its commercial counterpart. It brings similar data-integration capabilities to commercial enterprises, allowing companies to build a unified operational layer across their organization’s data, applications, and workflows. Crucially, this has allowed it to diversify away from core government and defense work, thereby massively expanding its TAM.
Foundry’s core function is integrating an organization’s disparate data sources, such as ERP systems, IoT sensors, spreadsheets, legacy databases, and third-party feeds, into a single, continuously updated operational layer. It’s positioned less as a single application and more as a flexible platform on which customers build their own custom operational tools.
There is also Palantir Apollo, which is less customer-facing than Gotham or Foundry but underpins both. It’s Palantir’s continuous delivery and deployment infrastructure, designed to manage software updates and operations across wildly different environments. Apollo automates the deployment, scaling, and monitoring of Palantir’s software across disparate, often disconnected environments, which is a nontrivial problem given how many of Palantir’s government customers operate in disconnected or highly secured settings.
More recently, the company introduced the Artificial Intelligence Platform (AIP), which layers large language model capabilities onto its existing infrastructure, enabling organizations to deploy generative AI within secure, governed environments. And this isn’t a standalone chatbot – AIP is built to sit atop Foundry’s or Gotham’s ontology layer, meaning that an AI model interacting through AIP has access to an organization’s real, structured understanding of its own operations, not just generic training data.
You see, most organizations’ AI bottleneck right now isn’t access to a smart model, but that their data is too messy, siloed, and ungoverned for a general-purpose LLM to act on safely. A chatbot with no grounding in your actual operational reality will hallucinate, give answers nobody trusts, or simply lack access to the data that matters.
That dynamic right now is what plays directly to Palantir’s strength.
Palantir’s whole architecture, with the ontology layer mapping real business objects and relationships, is, in effect, exactly the kind of grounding infrastructure that makes AI agents safe and useful to deploy operationally rather than just conversationally.
This is meant to address a common enterprise concern with generative AI: hallucination and lack of grounding. Because actions taken through AIP are tied to the ontology, outputs can be made traceable and auditable, and the platform includes guardrails to ensure AI-driven actions remain within defined, approved boundaries, relevant to both regulatory compliance and customer trust in sensitive sectors like defense and healthcare.
In plainer terms: AIP is the connective tissue that takes any LLM you choose, with AIP model-agnostic, and grounds it in your organization’s real, structured operational data, so it can reason and act with context instead of guessing. Add to that tools for generating detailed audit trails, explanations, and evaluations of model decisions to help organizations maintain accountability and historical lineage in AI operations, and AIP becomes an increasingly stronger fit in the AI era.
I genuinely believe that current developments all shift in Palantir’s favor as the preferred platform for running Agentic AI, especially for governments and complex organizations.
This is a huge opportunity for the company.
Today, AIP includes the following, among others:
AIP Logic — a no-code/low-code environment for defining step-by-step LLM workflows over Ontology objects, used by non-engineers to build automations without writing code.
AIP Chatbot Studio (formerly AIP Agent Studio) — a builder tool for developing production-ready AI-powered workflows and agents on top of the Ontology.
AIP Evals — an integrated evaluation framework that lets builders create test cases, debug and iterate on agent definitions, compare performance across different LLMs, and examine variance across executions. This matters a lot for enterprise trust — it’s the mechanism for proving an agent behaves reliably before it’s let loose on real operations.
AIP Assist — a context-aware in-app assistant available throughout the platform that changes its answers depending on which platform application is active, used for things like spellchecking and document editing in Notepad, or generating schedule configurations in the Scheduler tool.
That pretty much covers Palantir’s product suite in short. So, to round up the core business breakdown, how does Palantir generate revenue?
Simply, Palantir charges customers ongoing contracts/subscriptions to use its platforms, plus services to help implement and customize them, since the software is generally too complex to just “plug and play”, meaning Palantir’s own engineers often work hands-on with the client to set it up. Palantir refers to this as the “forward deployed engineer”. Given that this is labor-intensive work, Palantir’s revenue has historically had a heavier services/implementation component than that of typical SaaS companies, generating a solid revenue stream.
Contracts tend to be large and multi-year, with pricing based on a mix of subscription-style and usage-based models, and barely seat-based. However, unlike traditional software providers of more basic plug-and-play software, many of Palantir’s contracts are custom and tailored to individual enterprise needs, with options including full platform access and professional services, adapting to deployment scale, number of users, data processing requirements, and integration complexity.
So, in practice, large customers aren’t just paying a transparent per-compute-second rate; that consumption-based pricing gets wrapped into bespoke, negotiated multi-year contracts.
That pretty much covers it.
However, there is one risk I already want to point out: customer concentration is meaningful. Palantir’s top 20 customers generate 42% of revenue as of 2025, with one customer, likely the U.S. government, representing 25% of revenue. That is significant and inherently carries a risk worth considering.
On that note, let’s delve into Palantir’s competitive position and moat.
Competitive Position & Moat
Starting with Palantir’s competitive position, given the several different, overlapping, and huge markets Palantir operates in, it is hard to find concrete market share numbers. However, in assessing its dominance, it is definitely worth splitting the government and the commercial business again.
You see, in U.S. defense and intelligence software specifically, Palantir’s position is about as close to dominant as a single vendor gets in that space. It’s described as having become the primary software layer for the Department of Defense following a $10 billion enterprise agreement with the U.S. Army in July 2025, and its Maven targeting/AI system has seen growing standardization across NATO and allied militaries. It is one of the most embedded strategic partners for the U.S. Department of Defense.
In the narrow but high-value defense niche overall, Palantir genuinely doesn’t have a peer operating at the same scale and depth of integration. Anduril is the closest thing to a peer in terms of financial quality and growth trajectory, described as the only Rule-of-40 competitor with a similar trajectory, though it’s private and roughly 30x less scaled than Palantir as of 2026. Beyond Anduril, traditional government services incumbents like Booz Allen, Leidos, SAIC, CACI, and ManTech compete for similar contracts, generally with more legacy government-relationship depth but less of Palantir’s product-led, software-first approach.
So, when it comes to government and defense, Palantir is definitely a global leader and is actively taking share from legacy incumbents.
In Commercial, the picture is more complicated. In commercial enterprise data/AI platforms, Palantir is a significant and fast-growing player, but it’s competing in a much more crowded field against companies like Snowflake, Databricks, Microsoft (Azure/Fabric), Salesforce, ServiceNow, and the major cloud providers’ own AI tooling and data solutions, along with countless narrower point solutions.
Databricks is frequently named as the strongest overall rival, with its lakehouse architecture built by the original creators of Apache Spark. Snowflake is the other major rival, and notably isn’t just a passive bystander, as it’s aggressively pushing its “Unistore” workload to handle transactional data alongside analytical data, directly attacking Palantir’s strength in operational workflows, and has separately built out a Department of Defense Impact Level 5 (IL5) provisional authorization on AWS GovCloud, equipping it to deliver mission-critical data and AI solutions to the national security community, meaning Snowflake is now encroaching even into Palantir’s defense stronghold, not just commercial.
So, in Commercial, Palantir definitely isn’t a dominant force, and there’s significant competition to consider. Positively, the shift to Agentic AI, with Palantir seen as a leader, is helping it gain traction and market share, according to analysts.
Ultimately, Palantir is near-dominant in a narrow, high-stakes niche (Western government/defense operational data and AI), and is a fast-growing, differentiated, but not market-leading player in the much larger and more contested commercial enterprise AI/data space.
This positioning is important to keep in mind.
In terms of a moat, Palantir does look good, driven by three main characteristics. First of all, there is Palantir’s bundling advantage, with no single competitor able to match Palantir’s feature depth. This gives Palantir an edge, especially as more and more governments and enterprises are looking for a unified solution. Second, there are switching costs – once a customer’s actual workflows are built on top of Palantir’s ontology, ripping it out isn’t a vendor swap, it’s rebuilding the organization’s operational nervous system. And third, there are two decades of accumulated trust and accreditation. Particularly in defense and intelligence, the security clearances, accreditation processes, and relationship trust Palantir has accumulated since the early 2000s aren’t things a well-funded competitor can simply buy or build quickly; they require time, track record, and government sign-off that compounds slowly. And this becomes self-reinforcing as more agencies standardize on it.
Overall, this translates into a pretty solid moat and very sticky customer relationships.
So, Palantir does have a durable competitive position, but competition is significant. It absolutely isn’t uncontested. Palantir’s position is much less strong than, say, Veeva Systems or ServiceNow, in my view, which does impact the premium one should be willing to pay here.
Growth drivers and outlook
There is no denying that Palantir operates in one of the most interesting and compelling sectors, as data increasingly becomes the new gold, especially amid rapid AI adoption.
You see, before AI, fragmented data was mainly a human bottleneck: an analyst had to manually pull data from five systems, which was slow and annoying and limited by how much one person could do in a day. AI removes that human bottleneck on the output side, but that just exposes the input problem more starkly: an AI agent acting fast on bad, ungrounded, fragmented data doesn’t just produce a slow answer; it can confidently take a wrong action quickly. So, the same fragmentation that used to just slow humans down now has the potential to scale errors fast. That’s a real structural shift in how much fragmented data “costs” an organization, not just a continuation of an existing trend. And with 40% of enterprise applications embedding task-specific AI agents by the end of 2026, up from under 5% in 2025, according to Gartner, the need for a unified, governed data layer is becoming increasingly important structurally and fast, which plays in Palantir’s favor and is driving strong growth in the industry.
Take the global data analytics market, which is forecasted to grow at a 29% CAGR through 2030. This is driven by:
Rapid adoption of intelligence tools for everything from supply chain optimization to personalized customer experience – data is becoming hugely important across industries to optimize results.
The expansion of cloud-native analytics removes the capital intensity that once limited big-data adoption to only the largest, most well-resourced organizations, widening the addressable buyer base considerably.
An explosion in data volume, driven by the proliferation of IoT devices, social media, and digital transactions — simply put, there’s more raw material to process every year, mechanically growing the addressable problem regardless of any single vendor’s execution.
Increasing regulatory reporting needs are cited as a forward-looking driver — as data governance rules multiply globally, organizations need better tooling just to stay compliant, a demand driver largely independent of whether the broader economy is doing well.
And then there is AI, which doesn’t just add a new market; it makes the existing data analytics market more valuable per dollar spent, because better-grounded data makes AI outputs meaningfully better.
Meanwhile, the global enterprise agentic AI market is projected to grow at an even more impressive 44-46% CAGR through 2032, driven by rapid adoption, and this is a market Palantir wants a share of with AIP.
Across the two markets, growth forecasts are pretty spectacular, setting a strong backdrop for Palantir, which is well-positioned to outpace these growth rates given its low penetration in commercial, where it is rapidly gaining share.
On top of this, Palantir also benefits from growing defense investments. Ongoing conflicts around the world have put military spending back on the agenda in many European countries, while the U.S. continues to increase its budget as a percentage of GDP to maintain its lead. Global defense spending as a percentage of GDP is at 2.5%, the highest since 2009, but is expected to rise further through 2030, based on current commitments and targets, with NATO targeting 5% of GDP by 2035.
Furthermore, European military spending rose 14% YoY in 2025, with Germany boosting its budget by 24%. Meanwhile, U.S. spending is now over $1 trillion and could rise to $1.5 trillion by 2027, going by recent proposals. The outlook for defense spending is extremely strong, creating a durable, long-term tailwind for Palantir as one of the top Pentagon and NATO partners.
Even better for Palantir, within that growing pie, digital/software/AI spending is growing disproportionately faster. Take U.S. R&D spending, which will be up 27% YoY in 2026, one of the largest allocations in history. That 27% growth rate is roughly 5-10x the global defense spending growth rate, suggesting software/tech-adjacent categories are capturing a rapidly rising share of the marginal dollar.
We also see this outside of the U.S., with 1.5% of the 5% of GDP NATO target aimed at
AI, autonomous systems, cybersecurity, drone technology, and advanced communications networks are simply the future of defense, and Palantir will fully benefit. This includes addressing vulnerabilities in cybersecurity, communications networks, and other systems essential for modern military effectiveness, meaning nearly a third of the entire NATO spending increase is earmarked for the very digital/infrastructure category Palantir operates in and dominates, with few alternatives.
And this is an incredibly durable driver as well, geopolitically motivated, not just economically. Commercial AI adoption can slow if ROI disappoints or budgets tighten in a downturn, while defense spending growth is tied to perceived existential security threats, a fundamentally different, generally more durable kind of demand driver, less sensitive to ordinary business-cycle fluctuations.
While the defense exposure can’t always count on investor support, it is extremely favorable for its long-term outlook.
This combination of huge potential in the data analytics and agentic AI markets, along with growing defense budgets, makes for a very attractive market for Palantir, supporting considerable long-term optimism regarding its outlook. I mean, with an underlying market growing at this rate, the underlying drivers (growing datasets, AI adoption, growing defense budgets, etc.) sustainable, and Palantir in a good position to outgrow the market given low penetration, especially in Commercial, a 40-50% growth rate through 2030 really isn’t a stretch, especially if Palantir can indeed position itself as a top agentic AI layer – its core positioning and early results suggest it is.
In other words, Palantir has a mega outlook, supported by several structural drivers, no denying it. I can realistically see Palantir sustain a 40%+ growth rate through 2030 and continue to deliver exceptional growth into the next decade, given its runway.
Let’s then finally delve into its recent performance and financials.
Financial & Performance Review
Palantir released its latest financial results – Q1 2026 – back on May 4 and delivered hugely impressive results, accelerating growth for an 11th straight quarter and doing so profitably, expanding margin, and achieving a Rule of 40 score of 145%, up 18 percentage points sequentially. This really is what a business firing on all cylinders looks like.
