Founder Dossier No. 006 · 5 min read

Alex Karp

Co-founded Palantir in 2003 to build AI for the US military when Silicon Valley treated defense work as a moral failing, then held that conviction for twenty years until geopolitics, not marketing, proved the thesis.

Company
Palantir Technologies
Sector
Defense and enterprise software
Era
2003-present

For twenty years Palantir was, in its chief executive's own description, "the freak show": a Silicon Valley company building software for the US military at a moment when the industry treated defense work as a moral failing. Alex Karp, who co-founded it in 2003 and has run it ever since, came to the job with a doctorate in neoclassical social theory and no formal training in computer science or business.

Palantir builds software and AI systems used by militaries, intelligence agencies, and large enterprises to pull scattered data into a single operational picture and act on it. Its core products help organizations such as the US Department of Defense, allied governments, and large companies integrate data and deploy AI tools inside secure environments.

Background

Karp was born in New York City and raised in Philadelphia. His father, Robert Karp, was Jewish, and his mother, Leah Jaynes Karp, was an African American artist. After his parents divorced, the family faced financial hardship, and a local Jewish jewelry store owner helped shelter them, a kindness Karp has said he spent years repaying. He attended Central High School in Philadelphia and is dyslexic, an experience he credits with teaching him early which of his own abilities were real and which were assumed.1

He studied philosophy at Haverford College, graduating in 1989, then went to Stanford Law School, a period he later called "the worst three years of my adult life." There he met Peter Thiel; the two argued nightly over politics, Karp a socialist and Thiel a libertarian, and the friendship outlasted law school. Karp then moved to Frankfurt, Germany, and completed a doctorate in neoclassical social theory at Goethe University Frankfurt in 2002. He has since spent significant stretches of his life in Germany, shaped by time around older Germans who lived through the Third Reich.1

Getting started

Before Palantir, Karp had no plan to build a technology company. He inherited about twelve thousand dollars from his grandfather and used it to start Caedmon Group, a small money-management firm, aiming to return to Berlin as an intellectual dilettante once it succeeded. The September 11, 2001 attacks changed that trajectory. Thiel, who had co-founded PayPal and watched its fraud-detection systems catch suspicious patterns across transaction data, concluded a similar approach applied to intelligence data might have helped stop such an attack, and recruited Karp, PayPal engineer Nathan Gettings, and Stanford students Joe Lonsdale and Stephen Cohen to build the idea into a company.

Palantir was founded in 2003 to build software for the US military and intelligence community, at a moment when defense-related work was widely regarded in Silicon Valley as reputationally toxic. Karp took the CEO role; his contribution was philosophical and organizational, built around a theory of American institutions, individual rights, and how to organize unconventional engineers.

What he built

Palantir's products include Foundry, an enterprise data-integration platform; Gotham (also called Palantir Government, or PG), its government suite; Apollo, a system for deploying software into secure, disconnected environments; and the Ontology, a knowledge-graph layer modeling an organization's data, logic, and permitted actions so humans and AI systems can act on a shared picture of it. Its best known deployment is Project Maven, an Air Force surveillance-analysis program, a flashpoint after Google employees petitioned to withdraw their company from it while Palantir kept building.

The company runs on a hybrid model built around forward-deployed engineers embedded with customers to configure the software against real operational problems, rather than selling a purely self-service product. Karp has argued this services-heavy structure, once seen by investors as a liability, is central to its moat, since the value lies in accumulated deployments and relationships rather than a static codebase.2

Palantir went public in September 2020 via a direct listing on the New York Stock Exchange, opening at a valuation of roughly sixteen to twenty billion dollars. Revenue grew from about 1.5 billion dollars in 2021 to roughly 4.5 billion dollars in 2025, and the company turned profitable in 2023. Its market capitalization rose into the hundreds of billions of dollars by 2026, driven largely by demand for its Artificial Intelligence Platform, along with reported use of Palantir-built software in military operations.

How he operates

Karp's description of Palantir's method centers on what he calls "anti-dislexifying" a person's playbook: rather than impose a standard process, he identifies each person's individual aptitudes, including neurodivergent ones, and inserts compensations for their weaknesses while leaving their approach intact. He has said each major Palantir product was built by essentially the one person in the world who could have built it, and that the company's advantage comes from cultivating that talent rather than standardizing around a generic process.1 His profile is filed under the Contrarian Mission-Builder archetype: he committed to an unfashionable, philosophically grounded mission and held it for two decades before outside events, not marketing, shifted opinion in Palantir's favor.

Karp believes taste, the judgment of which business problem is worth solving, is the one input large language models cannot commodify; he has summarized this as "taste plus money."2 He also believes the technology industry risks nationalization if AI companies are seen simultaneously as destroying white-collar jobs and as indifferent to the military, and argues self-regulation is the only way to avoid a heavier government hand.2 He remains an outspoken defender of defense-sector work at a time when many Silicon Valley peers avoided it.

Where things stand

As of the mid-2020s, Karp remains Palantir's CEO and its most visible public spokesman, appearing at events such as a16z's American Dynamism Summit and Palantir's own AIPCon. Palantir has grown into one of the largest defense-technology companies by market value, with an expanding government business alongside a growing commercial AI platform. Karp describes himself as an introvert who avoids most social obligations, and divides his time between the United States and Germany.1 He has used his platform increasingly to warn the technology industry about the political consequences of AI-driven job displacement, urging companies to frame AI adoption around worker upskilling rather than headcount reduction.

Key facts

  • Co-founded Palantir in 2003 with Peter Thiel, Joe Lonsdale, Stephen Cohen, and Nathan Gettings, inspired partly by PayPal's fraud-detection systems.
  • Holds a JD from Stanford Law School and a PhD in neoclassical social theory from Goethe University Frankfurt.
  • Led Palantir through roughly two decades in which defense-related tech work was viewed in Silicon Valley as reputationally toxic, before AI demand shifted decisively in the company's favor.
  • Went public via direct listing in September 2020 at a valuation of roughly sixteen to twenty billion dollars; market capitalization reached the hundreds of billions of dollars by 2026.
  • Is dyslexic, and attributes his talent strategy for neurodivergent engineers to his own experience with it.

From the Curator

The reader is directed to the file on Evan Spiegel, another builder of deliberate inversions. Karp inverted the Valley's moral consensus and waited twenty years for the world to come around; Spiegel inverts product defaults and ships the inversion the same year. Conviction on those two clocks is nearly a different instrument.

Founder Dossier No. 038Evan SpiegelBuilds each Snapchat product as a deliberate inversion of the industry's defaults, ephemeral instead of permanent, no public likes, opening into the camera instead of a feed, and funds a twelve-year bet to reinvent the computer with augmented reality out of the profitable core.

Also on the desk: AI Red Lines (Concept practiced)

Related