There’s a company most people outside of finance or defense circles have never heard of. But if you’ve ever wondered who’s actually helping the US military decide where to send special forces, or how ICE tracks immigrants across the country, or which software Ukraine is using to fight Russia, there’s a good chance the answer involves Palantir. It’s one of those companies that operates just quietly enough that the general public doesn’t pay much attention, but people inside governments and big corporations are paying hundreds of millions of dollars for it every year.
So what exactly is Palantir? Short version: it’s a data software company. But that doesn’t really capture what it does. A more realistic description is that Palantir builds software that takes huge, messy piles of data from different places army databases, hospital records, supply chains, financial transactions and makes sense of it all in a way that helps people make decisions fast. It’s like if you had 50 Excel sheets from 50 different departments, all in different formats, and someone built you a brain that reads all of them at once and tells you what’s happening.
When I first started reading about Palantir, I found the whole thing confusing. The company is deliberately secretive, its CEO talks like a philosophy professor who’s had too much coffee, and a lot of what it does is classified. So let me just explain it the way I understand it, from the beginning.
Some of you guys might know the CEO its Peter Theil, from the so called Paypal Mafia.
From PayPal Fraud to CIA Contracts
The story starts in 2003. Peter Thiel, who had just sold PayPal to eBay for $1.5 billion, was sitting on a lot of money and a lot of ideas. At PayPal, his team had built software to detect fraud basically to catch people who were using stolen credit cards or fake accounts. The thing is, that fraud-detection logic was actually pretty interesting. You’re looking for patterns in massive amounts of transaction data, trying to spot the one weird thing that doesn’t fit. Thiel thought: what if you applied that same logic to terrorism?
This was 2003, right after 9/11. The US government was desperate to avoid another intelligence failure. So Thiel quietly incorporated a new company and named it Palantir — after the “seeing stones” in Lord of the Rings, the magical orbs that let you see across great distances. Which, I think, tells you something about how Thiel sees the company.
He recruited Nathan Gettings from PayPal, two Stanford students named Joe Lonsdale and Stephen Cohen, and then — weirdly — he brought in his old law school roommate Alex Karp as CEO. Karp had a PhD in social theory from Goethe University in Frankfurt. He’d never worked in tech. His thesis was about aggression and political identity. But Thiel thought that was actually good. The company needed someone who could navigate governments and intelligence agencies, not someone who just wanted to ship code.
Karp once described his and Thiel’s relationship: “We argued like feral animals.” That’s the guy running a $100 billion company today.
The early years were rough. Thiel personally bankrolled the initial $30 million cost when they couldn’t find any venture capital interest. Silicon Valley didn’t get it. The product was too weird, too government-focused, too secretive. VCs want fast growth and a clear market. Palantir had none of that.
The first real break came from the CIA. In 2005, Palantir secured a partnership with In-Q-Tel, the CIA’s venture capital arm, which provided critical early funding of around $2 million and validated Palantir’s potential for the intelligence community. Two million dollars sounds small now, but it was the foot in the door they needed. After that, more government contracts followed.
What the Software Actually Does
Okay so here’s where I have to try to explain something that engineers spend years learning. I’ll keep it simple.
Imagine you’re an analyst at the FBI. You have 10 different databases surveillance records, financial data, phone metadata, immigration records, court documents. But all these databases were built by different departments, at different times, using different formats. They don’t talk to each other. So if you want to check if Person X appears in three of those databases, you have to manually search each one, download the results, and then put them together yourself in a spreadsheet. This takes days. Maybe weeks.
Palantir’s main product called Gothamm, basically solves this. It connects all those databases, cleans up the data so everything is in the same format, and lets you search across all of them at once. You type in a name or a phone number, and it shows you every mention of that person across every database you’re connected to, with a visual map of who they’re connected to, when they traveled where, what transactions they made. What used to take two weeks can take about 20 minutes.
That’s the core thing Gotham does. It’s built specifically for governments, intelligence agencies, and military use. It’s currently the backbone of the US Army’s Project TITAN.
Then there’s Foundry, which is basically the commercial version of the same idea. Instead of connecting FBI databases, it connects a company’s supply chain data with their manufacturing data with their financial data. A car company can use it to see exactly what’s happening at every step of production, spot problems early, and fix them before they become expensive. Walgreens, for example, deployed AI-powered workflows across 4,000 stores in just eight months using Palantir’s platform.
And then in 2023, Palantir launched AIP, the Artificial Intelligence Platform. This is the newest thing and honestly the one that’s causing the most excitement right now. The basic problem it solves is this: the most powerful AI models in the world — GPT-4, Claude, Gemini, Llama — are trained on public internet data, and they have no knowledge of your organization’s private data. They cannot tell you what your inventory levels are. They cannot analyze your proprietary customer database. AIP bridges this. It lets you plug those AI models into your private, secure data, so the AI can actually reason about things that are specific to your company or your military unit, without that data ever leaving your controlled environment.
There’s also Apollo, which is like the plumbing behind everything. It makes sure Palantir’s software can run anywhere, on a regular cloud server, on a secure government server, or even on a Humvee in the middle of a desert with no internet connection.
How They Actually Sell This Stuff
Here’s something that I find genuinely smart about Palantir’s approach, and I didn’t expect it: the way they sell AIP.
Normal enterprise software sales take forever. You set up meetings, you do demos, you go through a procurement process, legal reviews it, IT reviews it, the decision goes up the chain, six months to a year before anyone signs anything. Palantir basically decided that was stupid, so they invented something called the AIP Bootcamp.
A bootcamp is a five-day workshop where Palantir’s engineers come to your company and actually build a working AI application using your data, right there, in five days. By the end of the bootcamp, the client has already seen their own data working inside the platform making the decision to sign a contract much easier because they’ve already experienced the value.
The conversion rate from bootcamp to signed contract is reportedly around 75%. That’s not normal in enterprise software. And the bootcamp approach has made Palantir’s sales cycle dramatically shorter. Over 1,300 bootcamps had been completed by Q4 2024, with a Fortune 500 industrial company expanding 5x from their initial engagement and a Fortune 100 retailer converting a pilot to a $12 million annual contract within months.
The Money
In 2024, Palantir made $2.9 billion in total revenue — 55% from government clients and 45% from commercial clients. 66% of that came from US customers.
The government side is mostly the US military, intelligence agencies, and other federal departments. The commercial side is a mix of healthcare, manufacturing, finance, and energy companies. Government was always their main business, but commercial is growing faster now. In 2025, commercial revenue reached $2.07 billion — a jump of almost 60% from 2024 while government revenue hit $2.40 billion, also up about 53%.
The gross margins are wild. Palantir’s gross margin is around 80%. For every dollar they bring in, they keep 80 cents after direct costs. That’s what software businesses can do when the product is complex enough that nobody else can easily copy it.
But there’s one thing I want to flag about the stock, because a lot of people have been excited about PLTR lately: as of late 2025, Palantir was trading at a trailing P/E ratio exceeding 400x — meaning the stock price is pricing in decades of perfect growth. That’s not me saying it will crash, I have no idea. But if you’re buying the stock because the company is good, just know the market has already priced in a lot of “good” already.
Where Palantir Shows Up in the Real World
This is the part that I think most people don’t realize. Palantir is not just in some Pentagon basement. It’s showing up in a lot of places.
In Ukraine, Palantir’s software has been used since very early in the war. Alex Karp was reportedly the first Western tech CEO to meet Zelensky after Russia invaded. Palantir is now reportedly embedded in Ukraine’s military and civic infrastructure — from targeting Russian personnel to collecting evidence of possible war crimes. NATO also signed a contract with Palantir for an AI-powered battlefield command system.
In the UK, Palantir’s work helped manage the COVID-19 vaccine rollout. The NHS used it to track and distribute vaccines across the country at a scale that would have been nearly impossible with older systems.
In the US, the Department of Defense’s Maven Smart System, which helps identify targets — runs on Palantir. The contract for Maven expanded from $480 million to $1.3 billion after an update in May 2025. And in July 2025, the US Army announced a $10 billion enterprise agreement with Palantir that consolidated 75 separate contracts into one big one. That’s the kind of deal that basically tells you the US military is treating Palantir as long-term infrastructure, not just a vendor.
The commercial side is also everywhere, just less visibly. GE Aerospace, Fujitsu, AIG, all using Palantir’s platforms for various operational things. AIG said its Palantir setup, built with Anthropic and AWS, can speed up the underwriting process by up to 5x.
The Part That Makes People Uncomfortable
So here’s where it gets complicated. Palantir is not just a neutral software company. The software they build ends up being used for things that a lot of people have strong opinions about.
The biggest controversy right now is around immigration enforcement in the US. In April 2025, the US government awarded Palantir a $30 million contract to create an “ImmigrationOS” platform, intended to track undocumented people living in the US. Palantir also built a system called ELITE — Enhanced Leads Identification and Targeting for Enforcement — which reportedly mines Medicaid and other public welfare databases to identify “high potential” targets for arrest, effectively turning social safety nets into deportation dragnets.
The ACLU has been very loud about this. Amnesty International has called on Palantir to stop this work entirely. And in May 2025, thirteen former Palantir employees former software engineers, managers, and at least one person from the company’s privacy and civil liberties team — wrote an open letter condemning the company’s work with the Trump administration. They said the software they helped build was being used in ways that violated the values Palantir claimed to have.
Palantir’s defense is basically: we just make the tool, we don’t decide how it’s used. Which, honestly, is the same thing every weapons company says. Some Palantir engineers themselves have raised concerns that building surveillance systems without sufficient oversight crosses a dangerous line from protecting civil liberties to undermining them.
There’s also the Gaza situation. Palantir has confirmed it provides support to Israel’s military. Critics have accused the company of being involved in targeting systems used in Gaza. Palantir says it is not involved in the Lavender and Gospel programs used to pick targets in Gaza. Whether you believe that or not probably depends on how much you trust the company’s statements, which, to be fair, is not easy to verify from the outside because everything is classified.
The core problem, which nobody has really solved, is the “black box” nature of Palantir’s operations when AI makes a recommendation to arrest someone or target a location, the proprietary code makes it nearly impossible to audit why the system flagged that specific individual. That’s a real issue, and it goes beyond just Palantir. It’s a problem with AI-powered decisions in general. But Palantir is one of the biggest and most powerful players in this space, so the stakes are higher.
The Person Actually Running All This
Alex Karp deserves his own section because he’s one of the stranger CEOs in tech.
He doesn’t look like a tech CEO. He doesn’t dress like one either he’s known for wearing expensive sports clothes to meetings with generals. He has a PhD in social theory, he meditates, he likes to ski, and he gives long philosophical answers to simple questions. But the guy clearly knows what he’s doing. Under his leadership, Palantir went from a company that couldn’t raise money to one with a market cap exceeding $200 billion.
What’s interesting is that Karp is publicly and loudly pro-Western. He talks openly about wanting Palantir’s software to help the US and its allies maintain global power. He doesn’t try to pretend the company is neutral — he actually thinks Western democracies should win, and Palantir should help them win. That’s a different posture than most tech companies, who generally try to say they’re just providing tools.
Some people find that refreshing. Others find it alarming. I think both reactions are reasonable.
Is This the Future or a Problem?
Here’s the thing: Palantir is probably going to keep growing. The kind of data integration they do is something that governments and companies actually need. The AI layer they’re building on top of it is only going to get more useful as the underlying models get better. And once you’ve embedded Palantir into your military or your supply chain, switching costs are so high that you basically never leave.
But there’s a version of this story where Palantir becomes something like a private intelligence infrastructure that runs a significant part of how governments make decisions about who to target, who to deport, who to watch and there’s basically no public oversight over any of it. The EFF raised exactly this concern in April 2026, asking Palantir directly what human rights due diligence it conducted for its ICE work, and whether it has used contract provisions or oversight mechanisms to prevent misuse. As of this writing, Palantir hasn’t really answered that to anyone’s satisfaction.
The company frames itself as being on the right side of history. Maybe it is. But that’s also exactly what every powerful surveillance company says, right up until it isn’t.
What I do think is this: Palantir is one of the most consequential companies most people have never properly looked at. It’s influencing real wars, real immigration enforcement, and real corporate decisions at a scale that’s hard to overstate. Whether that’s good or bad probably depends on who’s in charge of the government giving it contracts — and that changes every four years.
So yeah. That’s Palantir. A seeing stone for the modern world. Whether it shows you the future or helps someone hunt you down kind of depends on which side of the data you’re on.