
Palantir's 'Ontological Fraud'
Ontology is database modeling. The only function of the term "ontology" is to make people who don't understand databases think it's something new, and then willingly pay a thousand times the price for something old.
1. Two Expressions
By the end of 2025, Palantir's market capitalization surged to $400 billion, multiplying more than twenty times in two years. At every roadshow, in every white paper, and on every tech blog, this company keeps repeating the same word: Ontology.
This term is packaged as Palantir's core technical moat, its fortress, and its soul. Investors listen with reverence, Pentagon generals think it's the future of information warfare, and corporate executives feel they'll be left behind if they don't buy it.
At an AI summit for corporate CXOs, Palantir's VP of Sales flashed a slide with "Ontology" in large letters. The executives in the audience nodded slightly, wearing expressions that said, "I don't quite get it, but it sounds impressive." A few architects who had been dragged along to accompany them exchanged glances, and one whispered:
"Is he talking about tables and stored procedures?"
His colleague looked at the architecture diagram on the PPT, paused for three seconds, and said, "...Yes."

This is Palantir's most ingenious business tactic: using a term that carries the weight of 2,300 years of philosophical history to make non-technical decision-makers feel it's some kind of cutting-edge breakthrough, while leaving the engineers who understand the technology unable to find a suitable way to refute it in the meeting room. Because you can't exactly say to the VP, "Boss, what they're selling us is essentially just creating tables."
But lately, some people have been mystifying and hyping this thing. So today, I'm going to expose this emperor's new clothes.
2. The Rosetta Stone
Let's lay the facts on the table first. Palantir Ontology has four core concepts: Object Type, Property, Link, and Action. Scour all their documentation, white papers, and roadshow materials, and these four concepts are the starting point for everything.
Now, look at this table:

Four columns. Four terminology systems. The same thing. It's not "can be analogized," not "kind of similar," but completely overlapping, strictly isomorphic. Palantir defines sub-concepts in its documentation like Interface (polymorphism), Function (code logic), and Virtual Table. Translated, these are View, UDF, and Materialized View.

If you've studied database modeling, you already fully grasp what Palantir calls "ontology." No one ever told you that what you know can be wrapped in a philosophical term and sold for contract prices of millions to tens of millions of dollars per year.
Palantir's 2025 annual report disclosed: The top 20 customers each contributed an average of $93.9 million per year. All 954 customers contributed an average of about $4.7 million per year. That's the price tag for "tables and stored procedures."
3. The Same Idea Sold Five Times
Palantir is not the first to "invent" this. The same core idea has been repackaged repeatedly over 2,300 years, each time with a new name, and each time a group of people thinks it's a groundbreaking innovation.
First time: 350 BC, Aristotle. In the Categories, he proposed that the world consists of substances, which have properties and relationships between them. Translated into SQL, that's CREATE TABLE person (height INTEGER); teacher_id REFERENCES person(id). This isn't an analogy; it's the same mental operation in a different notation.
Second time: 1976, Peter Chen. Published the Entity-Relationship (ER) model: entities, attributes, relationships. Exactly the same as Aristotle, just translated from Greek prose into rectangles and diamonds. This paper spawned the entire relational database industry. Every programmer in the world who writes CREATE TABLE practices "ontology" daily, but no one told them it has a philosophical name.
Third time: 1990s, the Object-Oriented wave. Classes, attributes, associations, methods. The same set, with a "behavior" dimension added. Databases followed the trend with object-relational models. The object-oriented designs in PostgreSQL were added during this trend.
Fourth time: 2001, the Semantic Web. Tim Berners-Lee's vision. The core concepts of the OWL ontology language: classes, properties, relations, instances — completely isomorphic to the ER model. The term "Ontology" formally entered the computer science field during this wave.
Fifth time: 2016 to present, Palantir Foundry. Object Type, Property, Link, Action.
Notice the pattern: Each "reinvention" is accompanied by a wave of market frenzy. The ER model spawned the relational database market, OOP sparked the Java craze, the Semantic Web created a wave of academic and startup bubbles that then burst. Now, Palantir's Ontology rides the AI narrative, and the company's market cap has surged from under $20 billion to over $400 billion in just over two years.
This isn't technological evolution. It's conceptual reincarnation. In each cycle, what truly changes isn't the idea, but the wrapping paper around it and the people willing to pay for that wrapping paper.
4. The Cognitive Tax of Philosophical Terms
Now let's talk about the wrapping paper itself.
"Ontology" comes from the Greek on (being) + logos (study), literally meaning "the study of being." Aristotle studied it, Kant discussed it, Heidegger wrote an entire book, Being and Time, to redefine it. Just reading the Wikipedia entry on ontology takes half an hour and a strong cup of coffee.

This is the true power of the word: it creates knowledge asymmetry.
When a Palantir salesperson tells a manufacturing VP, "We use ontology to build your company's digital twin," the VP's internal monologue is probably: "Ontology? Sounds like some profound field of study. This must be some cutting-edge technology I don't understand."
If the same thing were translated into engineering language — "We'll help you create tables, define fields, set up foreign keys, and write stored procedures" — the VP's reaction would be: "Isn't that what the IT department has always been doing? Why should I pay you tens of millions for that?"
The same thing, with a different name, costs three orders of magnitude more. That's the cognitive tax of philosophical terms.

And here's a deep irony: Palantir's use of "ontology" is fundamentally anti-philosophical. True ontology explores open questions: "Where are the boundaries of being?" "Can categories be exhaustive?" The answers to these questions are fluid and uncertain.
But Palantir's Ontology does the exact opposite. It solidifies business entities into rigid Object Types, relationships into predefined Links, and operations into approval-flow-driven Actions. This isn't exploring the nature of being; it's pouring concrete molds over reality.
Data analyst Donald Farmer shared a real case on Substack: In the 1990s, he built a complete metadata ontology for a US auto finance company. Within months, the business team switched to new analytical tools and changed the credit risk assessment process. By the time the ontology team caught up, the business had changed again. His conclusion: an incomplete ontology isn't just lagging; it's wrong. And a wrong ontology is more dangerous than no ontology at all.
This is the fate of all rigid schemas. But for Palantir, this isn't a problem; it's the business model. Schema outdated? Spend millions to update it. Business changed? Buy another round of consulting services. The rigidity of Ontology isn't a flaw; it's a customer lock-in mechanism.
5. A Few Lines of SQL vs. $30 Million
Let's descend from the conceptual level to the engineering level. All the core primitives of Palantir Ontology can be implemented in PostgreSQL with just a few lines of code.
Object Type, Property, Link, Action, access control, audit logs, cross-source federation (FDW). Every core capability touted in the Ontology documentation is natively supported by PostgreSQL, with zero license fees.

A possible rebuttal: "A few dozen lines of SQL can write a schema, but can it deliver an end-to-end platform that a supply chain manager can use directly?" Of course not. But this precisely illustrates that Palantir's value isn't in the Ontology concept itself; it's in what lies outside Ontology.
It lies in building GUI wrappers for non-technical users, spending months on-site with clients to understand business processes, and securing procurement contracts with the Pentagon. These things have nothing to do with "ontology"; they are product engineering, consulting services, and political-business relationships. These are indeed valuable — but packaging the grunt work as a philosophical concept is Palantir's core competency and its biggest deception.
6. When You've Made Money, Scholars Will Sing Your Praises
At this point, we must answer a question: If the technology is so mediocre, how did Palantir achieve a market cap of over $300 billion, $4.48 billion in annual revenue, and 56% annual growth?
The answer isn't in the technology. The answer is in Washington D.C.
Palantir was co-founded by Peter Thiel in 2003. Thiel is no ordinary Silicon Valley investor. He was one of Trump's earliest and most important supporters in the tech world and a speaker at the 2016 Republican National Convention. Palantir's first external investment came from the CIA's venture capital arm, In-Q-Tel. From day one, the company's DNA has been driven not by technology, but by political-business relationships.
Palantir's 2025 annual report states in black and white: 54% of revenue comes from government clients. The U.S. Army awarded Palantir a $458 million battlefield intelligence contract. The Department of Defense signed a Project Maven AI contract with a $1.3 billion ceiling.
ICE (Immigration and Customs Enforcement) has committed cumulative funding of over $248 million since 2011. In 2025, driven by the Trump administration, Palantir secured a $30 million contract to build ImmigrationOS for ICE — a cross-departmental database for tracking undocumented immigrants.
How were these contracts obtained? Through the technical superiority of "ontology"? Or through Peter Thiel's position next to Trump? — Palantir spends $5 million annually on political lobbying. But Palantir won't tell investors in its roadshows: "Our core competency is Peter Thiel's political network." It will say: "Our core competency is Ontology."

This is the true function of the "Ontology": it is not a technical architecture, it is a narrative architecture. It makes a company that essentially wins contracts through government-business relationships and delivers through on-site engineers stacked as headcount look like a software platform company with an insurmountable technological moat. This is the quintessential American version of a "data middle platform" and on-site outsourcing.
VII. A Consulting Firm Wrapped in SaaS Skin
Palantir has a unique role called FDE, Forward Deployed Engineer. After a client signs, Palantir sends a team of engineers to work on-site at the client, helping sort out business processes, build data models, develop applications, and train users.
This is consulting, or more bluntly — "staff augmentation." Only Palantir insists it is a software company, not a consulting firm. Because a software company can command a valuation multiple of 70 times revenue, while a consulting firm is lucky to get 2-3 times.
Michael Burry, who is shorting Palantir, nailed this point precisely. He pointed out that Palantir classifies FDE labor costs as "R&D" or "Sales and Marketing" expenses, rather than cost of revenue. If calculated under Accenture's accounting standards, Palantir's prized high gross margin would shrink significantly.
A former FDE said one thing to Burry: "Foundry is not a perpetual license. You have to be trained to use it. Even then, you still need a lot of ongoing support."
What do those FDEs actually do on-site at clients? Write ETL pipelines to move data from SAP to Foundry, debug Kafka connectors, handle schema incompatibilities between Oracle and Snowflake, and explain to business users why a certain Link definition needs to be changed. The essence of this work is data integration and glue code — the most labor-intensive, most dependent on squatting on-site to understand the business, grunt work in software engineering.
Every engineer who has worked on enterprise data warehouse projects knows this kind of work: exhausting, tedious, no silver bullet. There are tens of thousands of SIs (system integrators) around the world doing the same thing. Accenture does it, Deloitte does it, Infosys does it, and various data middle platform vendors in China do it too. The difference is that they haven't packaged this grunt work as an "Ontology," nor do they have rock-solid government-business relationships, so their valuation multiples are one-hundredth of Palantir's.
The conceptual complexity of the Ontology precisely serves this business model. If you called Object Type a "table" and Action a "stored procedure," the client's IT department would say, "We can do this ourselves." But if you call it an "Ontology," introduce the three-layer architecture terminology of semantic layer, kinetic layer, and dynamic layer, and make the entire modeling process require dozens of minutes of manual clicking in a proprietary GUI — then the client will never be able to do without your FDEs.

The harder the system is to use, the more dependent the client becomes. The more obscure the concept, the more irreplaceable the FDE. This is not a bug, this is a feature.
Palantir's own data confirms this: in 2025, customer count grew 34%, but the average annual contribution from the top 20 customers grew 45%. CEO Alex Karp said something意味深长 on the earnings call: "There will be revenue growth that cannot be explained, but there will not be customer count growth that cannot be explained." Translated into plain language: we don't plan to win more customers, we plan to squeeze more money out of the customers we already have.
This is a consulting firm's growth model, not a software platform's growth model.
VIII. The Ontology Buyer Profile
Who buys Palantir?
Just look at the customer list and you'll understand. The U.S. Army, ICE, CDC, NHS (UK National Health Service), Airbus, BP. These organizations share several common characteristics:
First, the decision-makers don't understand technology. Pentagon procurement officers don't know what a foreign key is. NHS management doesn't care how ETL pipelines are written. What they need is a concept that sounds sophisticated, to justify that their tens-of-millions procurement decision was "strategic." "We introduced Palantir's Ontology-driven digital twin platform" — written on any reporting PPT, this looks ten thousand times better than "we hired an outsourcing team to build us a few tables."
Second, they're spending public money. Government contracts are characterized by ample budgets but a lack of technical audit capability. No one loses their job for spending thirty million on Palantir. But if you propose building it in-house with open-source solutions, and something goes wrong, it's your responsibility. This is the modern reincarnation of the classic "no one ever got fired for buying IBM." Only IBM has been replaced by Palantir, and the mainframe by the "Ontology."
Third, once path dependency forms, it is extremely difficult to reverse. Once your business model is encoded into Palantir's Ontology, your team is trained to think only in Palantir terms, your Object Types are not standard SQL tables, your Actions are not standard REST APIs, and your entire semantic layer is locked inside a proprietary platform. The cost of migration becomes unbearably high. This is the secret behind Palantir's net revenue retention rate of 134%. Not because the product is so good that customers voluntarily buy more, but because the lock-in effect leaves customers with no choice but to buy more.
IX. Summary
What is the Ontology? A modeling method with 2,300 years of history. A formal description of things, attributes, relationships, and operations. Every one of its core concepts can be mapped one-to-one to database primitives. The first three chapters of any database textbook cover its entire content. Palantir invented nothing new.
What is Palantir's real competitive advantage? Not the Ontology, but Peter Thiel's political network, the labor-intensive model of FDE on-site services, and the ability to create path dependency in governments and large enterprises. These three things — government-business relationships, consulting outsourcing, and lock-in effects — none of them has the slightest connection to the concept of an "Ontology."
What is the true function of the "Ontology"? A narrative tool. It allows a company that is essentially government-business relationships plus consulting outsourcing to enjoy top-tier SaaS valuation multiples in the capital markets. It makes non-technical decision-makers believe they are purchasing some unfathomable "core technology," rather than signing an overpriced system integration outsourcing contract.
At the end of the day, what Palantir does is this: take the grunt work of writing ETL, building tables, and configuring permissions that tens of thousands of engineers do every day, slap an Aristotelian noun on it, and sell it at an AI-era valuation.
It's like a Michelin chef writing "molecularly reconstructed carbohydrate crystals with organic protein gel" on the menu. What gets served is actually egg fried rice. Egg fried rice can be delicious, but you can't call "molecular reconstruction" a core technological moat. Especially when that plate of egg fried rice costs ninety million a year, and is personally served by Peter Thiel.
Next time you see a vendor going on and on in a PPT about the "Ontology," keep your wits about you — nine times out of ten, it's a load of hot air.
Note: The financial data cited in this article comes from Palantir's 2025 10-K annual report (SEC Filing), MacroTrends, and OpenSecrets public data. Michael Burry's short position information comes from his November 2025 SEC disclosure and subsequent Substack articles.