A Philosophy PhD Who Studied Aggression Built the CIA's Favorite Weapon. Now His Company Is Worth More Than Goldman Sachs. Nobody Can Explain What It Does.

In the offices of a company named after a dangerous artifact from The Lord of the Rings, the CEO — a biracial philosophy PhD who wrote his doctoral thesis about aggression, jargon, and the weaponization of language — keeps Tai Chi swords by his desk. When visitors come, he has been known to accidentally knock them over while demonstrating a move, apologize, then run off to fetch a printout of Goethe's Faust in the original German and read it aloud to prove it's better than the English translation.

This is the man running America's most important military AI company.

His name is Alex Karp. His company is Palantir Technologies. Its software is used by the CIA, the NSA, the FBI, the Department of Defense, and the United States Army, which signed a $10 billion contract with the company in July 2025 — the largest in Palantir's history. Karp personally met with Volodymyr Zelenskyy after Russia's invasion of Ukraine and later stated, with the casualness of someone ordering coffee, that Palantir is "responsible for most of the targeting in Ukraine."

The company was co-founded by Peter Thiel — the PayPal co-founder, first Facebook investor, and Tolkien obsessive who names his offices after locations in Middle-earth. Thiel funded Palantir with a simple thesis: the fraud detection software that saved PayPal from Russian hackers could be adapted to help the CIA find terrorists. The CIA's venture capital arm invested $2 million. Twenty-three years later, the company is worth $328 billion.

For seventeen of those years, Palantir was private. Almost nobody understood what it did. The word "vaporware" was used frequently. The word "scam" was used occasionally. Analysts said the opportunities in defense contracting were limited. The stock, when it finally went public in 2020, crashed from $45 to single digits.

Then AI happened. And the company that had been building the infrastructure to connect artificial intelligence to real-world organizational data for two decades suddenly became the most important enterprise software company in the world — or at least the most expensive. The stock went up 340% in 2024. Then it went up another 135% in 2025.

Palantir now trades at 218 times earnings. That's not a typo. The S&P 500 average is 22. The company generates $4.5 billion in revenue and is valued at $328 billion. One analysis concluded that Palantir would need to grow revenue 1,500% over 25 years to justify its current stock price. Yahoo Finance published an article titled "Palantir Could Be the Most Overvalued Company That Ever Existed."

Both of the following things might be true simultaneously: Palantir is the most important AI infrastructure company in the world, and its stock is the most absurdly priced asset that has ever existed. If that tension makes you uncomfortable, good. It should. Because the question of which one wins out is worth $328 billion, and nobody — not the bulls, not the bears, not the philosophy PhD with the Tai Chi swords — actually knows the answer.

To understand Palantir, you need to understand a fraud detection system called Igor, named after a Russian who was one of PayPal's most persistent abusers.

In the early 2000s, PayPal was fighting for its life against sophisticated Eastern European fraud rings. These weren't script kiddies running brute-force attacks. They were organized operations that adapted in real time, probing PayPal's defenses, finding patterns in the detection algorithms, and exploiting every gap before it could be closed. Automated detection wasn't enough because the fraudsters were adapting faster than the algorithms could learn.

PayPal's breakthrough was a hybrid approach: software flagged suspicious transactions, but human analysts made the final calls. The machines were fast. The humans were smart. Together, they were, in the company's own words, "devastatingly effective against adaptive fraudsters." The FBI took notice and asked if PayPal's tools could be adapted for financial crime detection.

Peter Thiel looked at this and saw something bigger. If a man-machine partnership could find credit card fraudsters in a river of financial data, could the same approach find terrorists in a river of intelligence data?

The timing was September 2001. America had just discovered, in the most catastrophic way possible, that its intelligence agencies couldn't connect dots across organizational silos. The CIA had pieces. The FBI had pieces. The NSA had pieces. Nobody had the picture. The 9/11 Commission would later identify this failure of information sharing as one of the primary reasons the attacks succeeded.

In May 2003, Thiel incorporated Palantir Technologies. He recruited four co-founders: Alex Karp, a philosophy PhD he'd met at Stanford Law School; Stephen Cohen, an engineer; Joe Lonsdale, a Stanford undergraduate; and Nathan Gettings, an engineer from Thiel's venture network.

The choice of Karp as CEO was deliberate and unusual. Thiel wanted a non-technical, non-business person at the helm — a philosopher who would serve as an "ethical compass" for a company handling the most sensitive data on Earth. Karp had a PhD in neoclassical social theory from Goethe University Frankfurt. His thesis was titled "Aggression in the Lifeworld" and dealt with the connection between jargon, aggression, and culture. He studied how language could be weaponized.

Then he built a company that weaponizes data.

Explaining what Palantir does is approximately as easy as explaining electricity to someone in 1890. "It's this invisible force that can light your house, run your factory, and also kill you." People would nod politely and go back to their candles.

Here's my best attempt.

Every large organization — a government agency, a military command, a Fortune 500 company — has data spread across dozens or hundreds of disconnected systems. The CIA has databases that don't talk to the FBI's databases that don't talk to the NSA's databases. A hospital has patient records in one system, billing in another, scheduling in a third, and research data in a fourth. An automobile manufacturer has supply chain data, factory sensors, quality metrics, and financial systems all running in parallel universes.

Palantir's software connects all of it. Not by replacing existing systems, but by sitting on top of them and creating a unified layer — what they call the "Ontology" — that links data about the same real-world entities across every database.

Example: "John Smith" exists as employee ID #4521 in HR, john.smith@company.com in email, badge #A-4521 in building access logs, and an account holder in the financial system. Palantir's Ontology creates a single "Person" object that connects all four identifiers. Multiply that by every person, vehicle, transaction, event, and physical asset in an organization, and you have a digital twin of the entire operation.

Then add AI. Palantir's newest platform, AIP — launched in 2023 — lets organizations deploy large language models on top of this Ontology. Instead of asking a generic AI to answer a question, you're asking an AI that has context — it understands your organization's actual data, relationships, and operations. And it does this with full security boundaries, access controls, and audit trails.

Palantir has four platforms. Gotham serves government and intelligence agencies — real-time military decision-making, threat assessment, targeting. Foundry serves commercial enterprises — supply chain optimization, manufacturing analytics, fraud detection. AIP deploys generative AI across both. And Apollo handles the deployment infrastructure that makes everything work in sensitive, disconnected environments.

Notable clients: JPMorgan Chase uses Palantir for fraud detection. Airbus used it to merge 25 data silos and quadruple A350 jet production. Ferrari uses it for predictive maintenance. The U.S. Army is paying up to $10 billion for it over the next decade.

And in Ukraine, Palantir's Gotham platform is being used for what Karp calls the "AI-powered kill chain" — pulling together satellite imagery, drone footage, and battlefield data into operational maps that help the Ukrainian military strike Russian targets. By September 2022, Ukraine had struck more than 400 Russian targets with HIMARS alone, with analysts crediting software-enabled targeting pipelines.

A philosophy PhD who studied aggression is now running the software that decides where missiles land. If you're looking for the through-line in his career, that's it.

Palantir was founded in 2003. It went public in 2020. Seventeen years private, in an era when most startups IPO within seven to ten.

The CIA's venture arm, In-Q-Tel, provided the first outside investment: roughly $2 million. Small money, but it opened every door in Washington. More importantly, it gave Palantir direct access to CIA analysts at Langley — the ultimate user testing lab. The feedback loops between intelligence analysts and Palantir engineers shaped the product in ways that no amount of product-market-fit theorizing could replicate.

But selling to the government was brutally hard. Traditional defense procurement was built for buying hardware — jets, tanks, ships. Not software. The entrenched contractors — Lockheed Martin, Raytheon, Boeing — had decades-long relationships with Pentagon procurement officers and knew how to navigate the bureaucracy. Palantir was an upstart from Silicon Valley trying to sell something nobody could physically see to people who measured value in tonnage.

Palantir's response was unconventional. Instead of salespeople, they deployed "Forward Deployed Engineers" — actual engineers who embedded with clients for months, solving problems directly rather than running sales demos. It built fierce loyalty but was hideously expensive. Every customer required custom engineering work. Scaling was nearly impossible.

Over those seventeen private years, Palantir raised $2.46 billion across 18 rounds from 165 investors. The valuation went from nothing to $730 million in 2010 to $20 billion in 2015. But the "vaporware" narrative persisted. Skeptics questioned whether the technology actually worked. The extreme secrecy of classified government work meant Palantir couldn't show its best results. You can't exactly publish a case study titled "How We Helped the CIA Find Someone."

Reported revenue in 2019 — the first year disclosed in the S-1 filing — was $742 million. In 2020, the year of the IPO, it was $1.09 billion. A real business, but not the kind that typically commands a $20 billion private valuation. The gap between perceived potential and demonstrable revenue fueled the skeptic narrative for years.

Palantir went public via direct listing on September 30, 2020, on the New York Stock Exchange. No underwriters. No new shares issued. Reference price: $7.25. Opening trade: $10. End of day: $9.50.

Then Reddit found it.

By the end of 2020, the stock was at $23.55 — up 136% from the opening trade. In January 2021, WallStreetBets took it to $45. The same crowd that sent GameStop to the moon decided that a CIA-backed data company run by a philosophy PhD was the next meme trade. The irony of retail investors on Reddit pumping the stock of a surveillance software company was apparently lost on everyone involved.

The inevitable crash came with the 2021-2022 growth stock sell-off. Rising interest rates destroyed the valuations of unprofitable growth companies. Palantir, which was burning cash and compensating employees with enormous stock grants, fit that category perfectly. The stock fell from $45 to single digits.

Karp's response to the years of pain was characteristically combative. When the stock eventually recovered, he took a victory lap, imagining bank analysts who had doubted Palantir driving "broken-down cars" while retail investors who had believed early drove "beautiful Teslas."

"Do you know how much money you've robbed from people with your views on Palantir?" he asked analysts at a public event. It is not a thing that normal CEOs say.

But then again, Alex Karp is not a normal CEO.

In 2023, while the rest of Silicon Valley was bolting ChatGPT onto everything from toasters to tax software, Palantir launched AIP — the Artificial Intelligence Platform. And the market suddenly realized something: the company that had spent twenty years building the infrastructure to connect AI to real-world organizational data was the only one that had actually done it.

Here's the distinction that matters. When Microsoft adds a chatbot to Bing, or when Salesforce integrates GPT into its CRM, they're layering AI onto a product. When Palantir deploys AIP, it's connecting AI to an Ontology — a living digital model of an organization's actual operations, assets, people, and processes. The AI doesn't just answer questions. It answers questions in context.

Ask a generic chatbot: "What should I do about the supply chain disruption?" It'll give you a Wikipedia-quality answer. Ask Palantir's AIP: "What should I do about the supply chain disruption?" It'll say: "Supplier X is late by 3 days, which affects production line Y, which delays order Z to Customer A. Here are three options, ranked by cost and timeline, with the relevant contracts and capacity data attached." Because it has the Ontology. It knows your supply chain. It knows your contracts. It knows your customers.

Palantir's go-to-market for AIP was a move of genius: "Boot Camps." Five-day intensive workshops where customers go from zero to production use case on their actual data. Not a demo. Not a proof of concept. A working application in five days. Nearly 300 organizations used AIP within five months of launch.

The revenue numbers tell the story:

Revenue growth is accelerating. Not decelerating, not stabilizing — accelerating. From 17% to 29% to 56% to a guided 61% in 2026. For a company that has been around for 23 years and generates $4.5 billion in revenue, this is almost unheard of. It's like a twenty-year-old marathon runner who suddenly starts getting faster in the twentieth mile.

U.S. commercial revenue growth hit 137% year-over-year in Q4 2025. The company guided for 115%+ growth in U.S. commercial for 2026 — $3.14 billion in a segment that barely existed five years ago. Net dollar retention hit 139%, meaning existing customers are spending 39% more each year.

On September 23, 2024, Palantir was added to the S&P 500, replacing American Airlines. Two months later, it transferred from the NYSE to the Nasdaq to qualify for the Nasdaq-100 index. The stock hit an all-time high of $207.52 in November 2025.

The vaporware company had become the most important AI stock on Wall Street.

There is a version of the Palantir story that is inspiring: a company built to protect democracy by giving intelligence agencies the tools to find terrorists without violating civil liberties. Karp has framed it this way explicitly: "Our project is to make America so strong we never fight. That's very different than being almost strong enough, so you always fight."

There is another version that is considerably less comfortable.

Palantir built "ImmigrationOS" for ICE — a $30 million contract to track immigration enforcement, select arrest targets, and create what ICE calls an "Immigration Lifecycle Process." The company also built a tool called ELITE that pulls addresses from HHS data, including Medicaid records, for immigration enforcement. The Electronic Frontier Foundation published a report in January 2026 exposing the use of health data for targeting immigrants. Civil liberties organizations argued that people who signed up for Medicaid never agreed to have their data used in deportation decisions.

Between 2012 and 2018, the New Orleans police secretly used Palantir's Gotham platform for "predictive policing" — without public disclosure or city council approval. The Los Angeles police department used it for threat assessment. An International Bar Association report raised rule-of-law concerns. Amnesty International raised human rights concerns.

Former Palantir employees circulated letters protesting the ICE contracts. Karp's response was to call critics "parasitic" and frame the debate in moral terms: "Not only was the patriotism right, the patriotism will make you rich."

Then there's the co-founder problem. Joe Lonsdale was accused of sexual assault by a former Stanford undergraduate in 2015 while he was her mentor. Stanford banned him from campus for at least ten years. Both lawsuits were eventually dropped; Stanford lifted the ban citing "new evidence." Lonsdale moved to Austin and continued venture investing. His VC firm Formation 8 shut down partially due to personal conflicts related to the allegations.

And then there's Peter Thiel himself — libertarian billionaire, major Republican donor, co-funder of JD Vance's Senate campaign, and a man whose political philosophy sometimes sits uncomfortably with the fact that his company builds surveillance software for the world's most powerful intelligence agencies. David Sacks, another PayPal Mafia member, became Trump's "AI & Crypto Czar." Stephen Miller, the architect of Trump's immigration policies, holds a substantial financial stake in Palantir.

None of this makes the software less effective. But it does mean that investing in Palantir requires a comfort level with the fact that your money is funding the development of tools that will be used in ways you may or may not agree with. That's not unique to Palantir — it's true of every defense contractor. But Palantir's tools are more intimate than missiles. They see data. They connect dots. They find people.

In Tolkien's Middle-earth, the Palantiri were seeing stones that could show you truth — but Sauron learned to use them to show selective truth, manipulating what the viewer saw to drive them to madness or submission.

The company named itself after a device that could drive you insane if you used it wrong, and then built tools for the most powerful intelligence agencies on Earth. Whether that's self-awareness or tone-deafness depends entirely on your faith in the people doing the watching.

Let's talk about what is measurably true about Palantir as a business, stripped of narrative.

Revenue: $4.475 billion in fiscal 2025, up 56% year-over-year. Q4 alone was $1.41 billion, growing 70% — the fastest quarterly growth rate in the company's history as a public company. Guidance for 2026: $7.18 to $7.20 billion, representing 61% growth. Revenue is accelerating.

Profitability: net income of $1.63 billion, up 252% year-over-year. Adjusted operating margin of 57% in Q4, 50% for the full year. Free cash flow of approximately $1.8 billion trailing twelve months. This is a legitimately profitable business with software-like margins.

Customer base: 954 total customers, up 34% year-over-year. Net dollar retention of 139% — existing customers are spending significantly more each year. The company has approximately 4,400 employees generating about $1 million in revenue per employee, which is an extraordinarily efficient ratio for any company, in any industry.

The revenue split: roughly 55% government and 45% commercial. U.S. commercial is the growth engine, accelerating from 71% to 121% to 137% YoY growth across 2025. International commercial growth is weaker at only 8% — a notable soft spot.

Now the valuation metrics that make grown adults either salivate or lose sleep:

The P/E ratio of 218x means you're paying $218 for every dollar of current earnings. The P/S ratio of 73x means you're paying $73 for every dollar of revenue. For context: at Cisco's dot-com peak in 2000 — the highest-profile valuation bubble in tech history — it traded at roughly 38x sales. Cisco took 25 years to recover its peak price.

Stock-based compensation was $684 million in 2025 — about 15% of revenue. CEO Alex Karp's "compensation actually paid" in 2024 was $6.8 billion, making him the highest-paid CEO in America. In that same year, insiders sold approximately $4 billion in total — Karp sold $1.9 billion, and Thiel sold $1.5 billion.

When the people who built a company and understand it better than anyone are selling at this pace, it's worth asking whether they're diversifying a concentrated position (rational) or signaling something about the relationship between the stock price and reality (concerning).

The bull case for Palantir is deceptively simple: if AI is the most important technology of the 21st century, and Palantir is the only company trusted by both governments and enterprises to deploy AI on their most sensitive data, then $328 billion might be cheap.

The infrastructure argument. Palantir spent twenty years building the Ontology — the data integration layer that AI needs to be useful in real-world operations. Everyone else is bolting AI onto existing products. Palantir built the foundation first and then added AI on top. That sequencing might prove decisive. As one analyst put it: "If AI is the new electricity, Palantir is the utility company."

The government moat. Security clearances create barriers to entry that no amount of venture capital can overcome. The $10 billion Army contract spans up to ten years. Once you're embedded in military and intelligence operations, switching costs are measured in national security risk, not dollars. Palantir is the only company with both deep government trust and cutting-edge AI capability. Amazon has the cloud. Google has the models. Neither has the clearances and the operational track record.

The commercial acceleration. U.S. commercial revenue growing 137% is not a number that established software companies produce. The boot camp model is converting enterprises in five days. Net dollar retention of 139% means the installed base is expanding rapidly. The company is simultaneously growing new customers and deepening existing relationships.

The revenue acceleration. 17% to 29% to 56% to 61% guided. This is the opposite of the normal pattern for maturing companies. It suggests a genuine inflection point, not a slow fade into commodity software.

The bear case for Palantir is also deceptively simple: the math doesn't math.

The valuation problem. A $328 billion company generating $4.5 billion in revenue. Even at 50% margins, that's $2.25 billion in profit on a $328 billion market cap — a yield of 0.7%. You would need the company to grow revenue 1,500% over 25 years to justify the current price. That requires sustained high-30s percentage growth for a quarter century. The number of companies that have achieved this: approximately zero.

The historical precedent. Cisco in 2000 traded at 38x sales and was the most valuable company in the world. It was also genuinely growing and genuinely important — it built the internet's infrastructure. The stock still fell 80% and took 25 years to recover. Palantir trades at 73x sales. If you believe history rhymes, this rhymes loudly.

The insider selling. $4 billion in 2024. Karp personally sold $1.9 billion. Thiel sold a third of his holding. These are not options exercises for tax purposes. These are massive, sustained liquidation events by the people with the most information about the company's future.

The competition. Snowflake and Databricks are building unified data platforms with AI capabilities. AWS, Microsoft Azure, and Google Cloud are building comprehensive AI and data analytics offerings. AWS is Palantir's biggest partner but also its biggest competitive threat — "up-the-stack applications that eat into partners' margins." If the Ontology is Palantir's moat, the question is whether generic AI models can eventually achieve the same contextual understanding without a twenty-year head start.

The concentration risk. 55% of revenue comes from government contracts, which are subject to political cycles, budget sequestration, and the fact that peace is bad for defense stocks. International commercial growth is only 8% — suggesting the AIP magic hasn't translated globally.

Stock-based compensation. $684 million in SBC on $4.5 billion in revenue means 15 cents of every revenue dollar goes to employee stock grants. On a trailing basis, SBC was $1.65 billion. This dilutes shareholders and inflates "adjusted" earnings metrics that exclude it. Real profitability is lower than the headline numbers suggest.

Alex Karp is, by any reasonable standard, the most unusual CEO of a major American corporation.

He was born in New York City in 1967 to a Jewish clinical pediatrician father and an African American artist mother. He studied philosophy at Haverford College, got a law degree from Stanford (where he met Thiel), then moved to Germany for a PhD in neoclassical social theory at Goethe University Frankfurt. His doctoral thesis — "Aggression in the Lifeworld: The Extension of Parsons' Concept of Aggression Through a Description of the Connection Between Jargon, Aggression, and Culture" — analyzed how language weaponizes social interactions, drawing on Theodor Adorno and critical theory.

He looks nothing like a CEO of anything. Wild, unruly hair. No suits. No corporate polish. Time magazine, which named him one of the 100 Most Influential People in 2025, called him "the embodiment of a new kind of Silicon Valley billionaire: an unashamed techno-nationalist who evangelizes Western power."

In February 2025, he published "The Technological Republic: Hard Power, Soft Belief, and the Future of the West" — which immediately became a #1 New York Times bestseller. The book's thesis: Silicon Valley lost its way. In the 1950s and '60s, Pentagon funding built Silicon Valley. Today's engineers build "photo-sharing apps and marketing algorithms" while ignoring defense and national security. The West's survival depends on redirecting technical talent back to hard problems.

His public statements are not the stuff of earnings calls:

"Somehow the corporate elite of this country thinks when it's time to make money, you stand up, and when it's time to stand up, you go play golf. And we've got to change that."

"The peace activists are war activists. We are the peace activists."

"Either we win, or China wins."

"The mission of Palantir is to give an unfair advantage to our friends."

Karp's compensation in 2024 was $6.8 billion — the highest of any CEO in America. He simultaneously sold $1.9 billion in stock and continued selling into 2025 and 2026. This creates a dissonance that every investor must reconcile: the CEO who speaks about patriotism and Western survival with messianic intensity is also extracting billions of dollars from the stock at prices he presumably believes are high enough to sell into.

You can resolve this in two ways. Either Karp is a rational actor diversifying an enormous concentrated position — a billionaire who has 90%+ of his net worth in one stock and is simply following the advice any financial advisor would give. Or he's a man who knows the difference between the story he's selling and the numbers he's reading.

Both explanations are plausible. Neither is provable. And Karp, the philosopher who studied how language can be weaponized, is too smart not to know exactly what the ambiguity does to investors who are trying to read the signal.

In Tolkien's Middle-earth, the Palantiri were created by the Elves in Valinor and brought to Middle-earth by the Númenóreans. They were instruments of communication and far-seeing — powerful, useful, and fundamentally neutral. Their danger lay not in the technology itself but in who used them and why. Sauron corrupted them. Denethor, the Steward of Gondor, used his to see truths about the world that were real but incomplete — and the incomplete truths drove him to despair and self-destruction.

The lesson of the Palantiri was never "don't look." It was "what you see depends on who's showing it to you, and what they're choosing not to show."

Peter Thiel, the man who named this company, certainly knows this. Alex Karp, the philosopher who studied how language shapes reality, certainly knows this. And yet.

Palantir is building the world's most powerful seeing stones. They connect intelligence databases, battlefield data, corporate supply chains, and AI models into unified systems that let their operators see things that were previously invisible. They are being used to target missiles in Ukraine, track immigrants in the United States, detect fraud at JPMorgan Chase, and optimize jet production at Airbus.

The technology is real. The revenue is real. The growth is accelerating. The profitability is genuine. The military applications are saving lives. The AI platform is converting enterprises at a pace that no competitor can match.

The stock price is also real — and at 218 times earnings, 73 times sales, with $4 billion in insider selling last year and a history of stock-based compensation that dilutes shareholders by billions annually, it is pricing in a future that requires sustained perfection for decades.

The company that spent seventeen years being called vaporware turned out to be one of the most important technology companies in the world. The stock that crashed from $45 to single digits turned out to be one of the greatest investments of the decade.

But "turned out to be right" and "will continue to be right at these prices" are very different statements. And the gap between them is where $328 billion dollars of investor capital currently lives.

Karp, characteristically, has framed the question in the grandest possible terms: "Either we win, or China wins."

That may be true. It may even be the most important strategic observation anyone in Silicon Valley has made in the last decade. But it doesn't tell you what a dollar of Palantir revenue is worth. It doesn't tell you whether 73x sales is rational or rapturous. And it doesn't tell you whether the Tai Chi swords on the CEO's desk are the props of a visionary or the set dressing of a performance.

The seeing stone shows you truth. But it only shows you what it wants you to see.

What you decide to believe — that the unseen parts are more of the same, or that the unseen parts are the ones that matter most — that's not a financial question. That's a philosophical one.

And for once, the company has the right CEO for that conversation.

This article is for educational and entertainment purposes only and does not constitute investment advice. The author holds no position in Palantir (PLTR) or any related securities.