Transcription
[Music]
Co-founded by Peter Thiel and named after the powerful but dangerous magic seeing stones from "The Lord of the Rings," Palantir is one of the most controversial companies in the world. Some of their early clients include the CIA, the FBI, the NSA, and the U.S. Special Operations Command—four organizations most often associated with secrecy, spying, and other shadowy operations. Palantir's own marketing seems to only add to the intrigue, while interviews with their eccentric yet charismatic CEO, Alex Carp, often lead us to more questions than answers.
The actual databases were not set up to abstract out of the databases, where the abstraction is actually the reality. Questions like, "Is this really a company that we should support, let alone invest in?" To answer that, we need to answer a different question first: What does Palantir actually do?
Exactly. My name is Alex, and I'll answer both of those questions by taking a look at the science behind the stock.
So, what does Palantir do exactly? It's easy to classify Palantir as a data storage and analytics company and compare it to services like Snowflake, and that's a really good starting point. Kind of like comparing a Tesla to a Porsche; on the surface, we're looking at two reasonably similar products with similar purposes and similar features.
For example, both Snowflake and Palantir create software tools to help companies store, process, and visualize large amounts of data that come from a wide variety of sources. They also work with companies that operate in a wide variety of markets, including health care, financial services, manufacturing, and, of course, the government.
But if they're so similar, why doesn't Snowflake have the same shadowy reputation as Palantir? The comparison between Snowflake and Palantir starts to fall apart for the same reasons as comparing a Porsche to a Tesla; they're actually completely different when you look under the hood (no pun intended).
Here, let me show you. Palantir's goal is to enable data science across an entire organization, even for people who are not data scientists. The idea is to give everyone—salespeople, human resources, engineers, logistics coordinators, data analysts, executives, literally everyone—a self-consistent, relevant, real-time view of their organization. Doing that helps them make the right decisions depending on their role. Government organizations get the Gotham operating system, and commercial organizations get the Foundry operating system.
Right away, we can see how that's not the same as Snowflake. A much better comparison to make with Palantir would be something like a Microsoft Windows integrated with a Microsoft Azure, but purpose-built for large organizations with lots of new data streaming in all the time. That's a pretty tall order. So, how does Palantir do it?
Palantir automatically discovers and builds the digital twin of an organization. When you think about the phrase "digital twin," you're probably thinking about 3D graphics, virtual reality, and maybe even the whole metaverse craze that swept the internet late last year. But that's not the only way to think about digital twins. What Palantir actually does is build an organization's ontology. You may have heard them use that word once or twice. We'll present the Foundry ontology and simulation engines; our ontology continues to grow.
This ontology goes beyond integrated 360-degree views; this ontology is the representation, or the digital twin, of an organization. So, let me tell you what the heck an ontology is and why it's so important.
At a high level, an ontology is a description of things and their relationships—that's it. Let's think about a social network like Facebook. That's a network of people and their relationships. People have properties like their name, birthday, workplace, school, and location. Some of those properties are different ways that we can connect multiple people. For example, everyone with the same workplace is a co-worker, everyone with a similar location is a neighbor, and so on. Those are meaningful relationships to know about.
You can also use combinations of properties to make guesses about or derive different relationships. For example, everyone who went to the same school and has a close enough birthday was probably a classmate. It turns out that having a network view of your data, or an ontology, is a really useful tool in a lot of different ways. Sure, you can find your friend's birthday on Facebook because you're connected; that's actually saved my skin quite a few times.
But what if you want to find people that make the decisions in an organization, like a workplace, or a school, or a sports team? Well, that's probably going to be the person in the organization with the most connections in it. That's the person that you want to market your services to or make your campaign promises to, since they can help you spread your ideas the most to the rest of the people inside that organization. Finding that person would be a pretty powerful insight, wouldn't it?
Seeing your organization's data from this ontology view can help you discover valuable insights like this. There are a few special things to consider about this ontology approach that set it apart from just storing and organizing data in a new and different way.
The first is that you can have networks within networks, or ontologies within ontologies. Let's go back to the social network example. You and your family make up a small network where everyone is connected to each other. That network is part of a larger network, like your neighborhood, which is a network of nearby families and households. Your county is a larger network of neighborhoods, your state is a larger network of counties, and so on.
Here's the thing: not every country is made up of states and counties, so even this network would look pretty different if we tried to make it in different parts of the world. And that's just one type of relationship: geography. Think about all the other things that people can belong to, subscribe to, communicate with, like and dislike, purchase, own, access, and on and on, and how those things would change across cultures, languages, and countries. All of these are different kinds of possible connections in an ontology.
This idea of representing all data as an interconnected ontology works really well, no matter the size, speed, type, or quality of the data. That means this idea of an ontology can scale all the way up to entire government branches or down to just a few spreadsheets. It's all entities and relationships.
That's why Palantir's Foundry platform has a practically infinite total addressable market. And that's not all. What we've actually been talking about this whole time is a branch of math called graph theory. Don't worry, I'm not about to hit you with any homework problems, but what you should know about graph theory is that there's an absolutely massive amount of research, development, toolboxes, code bases, and artificial intelligence and machine learning models specifically designed for these types of problems.
Why? Because many of the biggest consumer-facing companies on the planet understand that this is the way to represent their data: Facebook's network of people and things, Google's network of websites and links, Amazon's network of products and vendors, a bank's network of branches and credit lines and accounts.
Let's just use Netflix as an example. People and content on Netflix are connected by thumbs up and thumbs down ratings, watch time, and other measures of interactions. So, interactions are the relationships between people and content. When you rate content on Netflix or Spotify or Amazon or wherever else, you're telling that platform what your relationship to that content is. Content can be related to other content as well: genre, theme, shared actors and directors, length, family friendliness, and on and on.
There are going to be clusters of content that are very similar: 90-minute action adventures, horror movies from the East, Disney musicals, romantic vampire movies—don’t judge me. Now a new user named Alex tries Netflix for the first time. Let's take a look at what users like Alex have already watched and enjoyed on Netflix the most and put that content in front of him first.
[Music]
It's mostly romantic vampire movies. The important thing to understand here is that Netflix matching content to users can be represented as a graph theory problem: YouTube matching content to users—graph theory; Spotify matching songs to users—graph theory; Amazon matching products to users—graph theory; Lemonade matching insurance products to users—graph theory; Teledoc matching different steps along a specific healthcare journey to users—graph theory, graph theory, graph theory.
So, when Palantir builds you an ontology of your business, you can suddenly leverage the same research, toolboxes, code bases, and machine learning models as Facebook, Google, Amazon, and all these other great tech companies. You can answer questions like: What part of this network makes me the most money? What part of the network costs me the most money? When I get a new customer, what should I show her first based on the most profitable customers that look the most like her? What are the one or two biggest risks to my business today, and how can I fix them tomorrow?
These are the kinds of questions that apply to every kind of business. But if you want to answer them, you first need your data in the same form as the best businesses answering them today. That's this digital twin or this ontology. The rest of Palantir's offerings are tools and packages and modules and plugins that help answer these questions, depending on the specific use case.
That's why there's a Foundry for automotive, a Foundry for retail, a Foundry for energy and supply chain, and healthcare, and even crypto. They're industry toolkits that work best with different kinds of digital twins depending on the nature of the actual business—kind of like how you wouldn't build a stadium in a hospital with the exact same set of tools.
As you can see, that's why you also can't compare Palantir to Snowflake, even though Snowflake is a great platform for storing, securing, processing, analyzing, and visualizing data. And actually, Palantir knows that too, which is why you can plug data stored with Snowflake into Foundry or stored with AWS or Azure. Those aren't the features that set Palantir apart any more than Tesla's tires or doors or windows set it apart from other cars.
In the case of both Palantir and Tesla, the differentiator isn't in the what; it's in the how and why. Hopefully, this episode helped you understand a little more about Palantir and why it could be such a good long-term investment.
I actually think that Alex Carp wants Palantir to remain shrouded in mystery, wants it to stay misunderstood by the market, and wants to keep choosing their next partners and customers themselves—not just to build up their own knowledge of the different kinds of ontologies like the ones we just talked about, but to ensure that this kind of power stays in the right hands.
But Palantir doesn't just pick which companies can be on Foundry; they also invest in some of them. If you want to know what kinds of crazy futuristic publicly traded tech companies Palantir is investing in right now, check out this episode next. And if you feel I've earned it, hit those like and subscribe buttons to let me know that you enjoyed the science behind this stock.
Either way, thanks for watching, and until next time, this is Ticker Symbol U. My name is Alex, reminding you that the best investment you can make is in you.