February 1, 2026

What does Palantir do?

A lot of people genuinely ask what companies like Salesforce, Oracle, and Palantir do. Seeing all the AI ads at the SJC airport is cringe, lets be honest, and it does seem like griftware to see all these ads pushing 'AI'. Well, what does Palantir do?

I think the domain that Anduril and Palantir work in is very similar. Quite a few people in leadership at Anduril originate from Palantir. While they have esoteric products, the main thing they do is solve problems. Well duh. Let me get more into it.

The big data revolution happened in the mid-2000s and stayed a push for the main companies until the mid-2010s. The government eventually caught on. If you've ever worked with data, you would know that the hurdle isn't actually doing the analysis; that’s the easy part. Yes, there are lots of complex algorithms and techniques used to analyze data, but the number of people doing the highest level of innovation here is a lot smaller than you think. It's really a pyramid: you need the data first. A lot of the effort goes into collecting and storing that data.

Okay, now the government has all this data. What do they do with it? In my view, the government and companies really struggled with it. Private companies didn't struggle as much because they could work on their own timescales. Palantir wins contracts directly with the government to integrate all this data into something useful. Yes, that seems like something simple, but there are plenty of hurdles you have to jump through. Palantir worked on tightly scoped contracts to pull this data together into a known picture.

Palantir acquires all of these one-off contracts with the government (as well as private companies). Then it sees which ones have traction beyond the original contract and sells them to other customers. The book The Philosopher in the Valley covers this in some detail.

Palantir goes after very targeted data aggregation problems and then tries to build them into products. For example, their ERP migration software is something no one actually wants to do. While I'm sure they write great integrations, their prowess is in their ability to assume risk and have domain knowledge.