Pulling Back the Curtain on Palantir (Whiteboard Confessional)

Over the course of my career, I’ve worn many different hats in the tech world: systems administrator, systems engineer, director of technical operations, and director of DevOps, to name a few. Today, I’m a cloud economist at The Duckbill Group, the author of the weekly Last Week in AWS newsletter, and the host of two podcasts: Screaming in the Cloud and, you guessed it, AWS Morning Brief, which you’re about to listen to.

Corey: This episode is brought to you by Trend Micro Cloud One™. A security services platform for organizations building in the Cloud. I know you're thinking that that's a mouthful because it is, but what's easier to say? “I'm glad we have Trend Micro Cloud One™, a security services platform for organizations building in the Cloud,” or, “Hey, bad news. It's going to be a few more weeks. I kind of forgot about that security thing.” I thought so. Trend Micro Cloud One™ is an automated, flexible all-in-one solution that protects your workflows and containers with cloud-native security. Identify and resolve security issues earlier in the pipeline, and access your cloud environments sooner, with full visibility, so you can get back to what you do best, which is generally building great applications. Discover Trend Micro Cloud One™ a security services platform for organizations building in the Cloud. Whew. At trendmicro.com/screaming.

Corey: Welcome to the AWS Morning Brief. I'm Cloud Economist Corey Quinn, And this is the Whiteboard Confessional segment that has increasingly been taken over by my colleague, Pete Cheslock, as we tear through various cloud-based companies, public filings as they race to go public and inflict their profits, or in more cases, losses on the public markets. Pete, thanks for joining me.

Pete: I am super happy to be back again, and making my mother happy that I'm actually using that MBA that I spent all that time to get.

Corey: So, we could wind up talking just about how Palantir is awful in a variety of ways. My personal favorite was the letter that their CEO attached saying that effectively engineers were stupid and didn't see the big picture, which is a weird thing to say about a whole group of people you're actively trying to hire, but all right. Let's talk about their S-1 filing. This has been anticipated for a while. What do you think?

Pete: Well, Palantir has been around for a very long time. I think it's been around a lot longer than a lot of people realize. You know, early 2000s. It was technology built to tie data together and to be honest, I only know—I’ve ever heard of one company actually using Palantir—the technology—a commercial company. They were actually using it as a SIM—SIM, whatever you want to call it—Security Information Management System—

Corey: Event management or something like that. Yeah.

Pete: Exactly. And ironically enough, that company actually—that was using Palantir—replaced it with an Elasticsearch ELK stack, which I thought was fascinating. I know nothing about their software, but I was very fascinated to read the S-1 because there's been this mythology around it and you can hear so much about insiders at Palantir, employees selling their shares in this wide secondary market. So, I was very curious to see what we were going to find, and there are definitely some interesting bits within.

Corey: There certainly are. And it's strange because for a while Palantir was doing interesting things in the market. They were offering $20,000 referral bonuses to people who referred engineers in for certain roles, and you didn't have to be a Palantir employee to do it, which was fascinating. They've recently moved headquarters from Palo Alto over to Denver, Colorado, which… okay. They are claiming it's for this whole lofty mission. Let's not kid ourselves: it's a tax play. [laughs]. 

And there's also a whole bunch of interesting stuff buried in here. But yeah, in many ways, this is a legacy company in some respects. It's been around almost 20 years. And strangely, I don't know about you, but I don't know anyone who works for Palantir. I did a little digging in preparation for this episode, and it turns out, I actually kind of do, but they're very quiet about it. It's one of those things where people don't want to be called out for working at a company that is this particular flavor of controversy, and I can't say I blame them.

Pete: Yeah, I haven't looked through my LinkedIn to see if any of my connections have ever worked there. Granted, it's such a West Coast company that me out in the East Coast, be pretty rare to run into anyone out here who's kind of taken their time and done the Palantir. I have heard, again, the rumors that they've always paid very well, and—

Corey: They would kind of have to.

Pete: You know, in the Bay Area, you kind of have to. And competing for talent against other places who pay really well, like Netflix, and Uber, and all these other big companies that are out there. So, it's a big competition for the top talent.

Corey: Oh, yeah. And most of what they do is data analytics. They take in a whole bunch of data, and they crunch a whole bunch of numbers and come out with other stuff. Historically, they have been focused on selling their services to governments, but now they're expanding in the enterprise story as well. And that is, of course, going to be a bit of a challenge for them as they expand into it, but we can talk about what they do, how they do it, and all the other challenges. Let's talk about Cloud. What do we know about their cloud environment based upon their public filing?

Pete: Well, they talk about their commitments. So, this is something you often see in S-1s of their various cloud commitments, and I think this one was super interesting in that they listed commitments for about $1.5 billion in cloud commitments over six years, and this was an agreement they entered into at the end of last year. Just a massive, massive amount of cloud spend commitment, right?

Corey: Yeah, it’s a quarter billion dollars a year in spend. Which is, again, we see a number of customers in that range pretty frequently, it's not always typical to see the better part of a decade done to satisfy those commitments, though. Usually they're, “Well, this stuff is always changing. Let's talk about doing this for the next three years.” Six is a bit on the outside range of what we tend to see. 

What's fun to me was the breakdown of that commitment, which was just—I've been using this as a talking point for a week now—which is they have to undisclosed cloud companies in this part. They mention elsewhere that they use Azure and that they use AWS. Great. Fine. For one cloud provider, they have a six-year commitment of $1.49 billion, which is an enormous pile of money. The other provider they have is a little bit less than that; for $45 million committed over a five-year span. That's almost an order of magnitude difference. 

People ask me from time to time, “Well, why do you focus on AWS and not other cloud providers?” This is the purest example of what we've got to prove this. When we talk to companies about their expensive cloud bills, the AWS problems add a zero to the end of what they tend to be with other cloud providers. It's not out of preference on our part, and it's not out of, “Well, this is what we know and learning other things is hard.” It's that when we hear complaints about GCP or Azure bills, they are not in the same universe as what we're seeing on the AWS side. In other words, their annual Azure commit is about $9 million a year, and their AWS commit—at Palantir—is $250 million a year. Bit of a difference.

Pete: Yeah, exactly. I think outside of the fact that the commitment was so high, they obviously had some really interesting contractual terms, which kind of talked about how they could shift their spend if they didn't satisfy by a certain time and things of that nature. But what was so amazing was that they talked about this $1.49 billion commitment, and then they were like—this other sentence was like, “Oh, yeah, by the way, we also have this like $45 million one, and there's some stuff on that one as well.” So, of course, I'm like, “All right, it's either Azure or GCP. Which one is it?” So, just start doing the search through and they list Azure, they list Amazon, they just don't obviously list them in the same sentence. But us as rational viewers, we can make that assumption that they're probably not doing a $1.49 billion commitment on Azure.

Corey: And we can tell that because if some company did do a $1.49 billion commit on Azure, you wouldn't be able to roll out a bed in the morning before a sponsored ad on your alarm clock was telling you about this, thanks to Azure’s marketing campaign. They would be insufferable about this and never stop talking about it. Whereas with AWS, it’s one of those. “Okay, cool.” That's a big commitment, don't get me wrong. But remember, that’s—they’re one of the larger AWS accounts, depending on what inclusivity categories you look at, and that's a $250 million a year commit. Well, remember that AWS is north of $40 billion a year in terms of revenue. So, it's not a story where you just have one or two big customers that drive everything else.

Pete: Yeah, exactly. I mean, a $250 million swing for Amazon yearly, if Palantir didn't meet their commitment—and they have some words in here that basically says that they can push it out forward—I think their minimum they would have to cut a check for is like $30 or $50 million and they can push the rest of the spend later, but even if so, I mean, even though it’s such a huge number, it's such a small percentage of the behemoth that is Amazon and their cloud. It really is.

Corey: Yeah, there's a lot that winds up factoring in. And I think that it's, it's easy to sit here and cast stones, but these are big numbers, and it's a heck of a commitment. And it's just one of those things that I find fascinating. Especially since—let's not kid ourselves—Palantir is not exactly a logo everyone wants to have on their website. There are serious challenges with that, and, “Huh, we got a lot of attention for talking about doing business with Palantir, but it's not the good kind of attention.” To be very clear, because I don't want to get letters, Palantir is not a customer of The Duckbill Group.

Pete: Yeah. I mean, they're known as very secretive company, right? And I don't think you're going to see their advertisements on billboards like you might see—or at the airport, you might see a Slack [laughs] commercial on there. So, it's a lot different company. One of the places where I think it shows just how much of a different company this is, is how they are actually doing this IPO listing. So, if anyone has been following some other companies that have followed what's called a direct listing, Slack was one Spotify was another. This is where—

Corey: I believe Google was another, too.

Pete: Yeah, you just—you sell your shares to the public. So, for companies that have a certain hype factor them, or honestly for companies who think that they can do it where a normal IPO you're going to have underwriters, these big banks, of which you've all heard their name like Goldman Sachs, and whatnot, they will essentially work with their customers to set a price for your shares to make sure that the initial—let's say you sell 100 million shares—to make sure that those shares are sold on the IPO. And then there's a lockup period, and that's where everyone else who's holding shares just has to sit and wait and hope the company doesn't just crater like in the last bubble we dealt with. Now, a direct listing is much different and there's a lot of risk to it as well. 

A direct listing is anyone who has shares, any registered shareholder, which is like an employee, a former employee, the hundreds or thousands of people who hold shares in Palantir can just sell it. There'll be someone in the stock market called a market maker who will facilitate those trades, but when they list their IPO, if you have shares in Palantir, you can sell them for whatever someone's willing to pay for them. It's a much different model. But what was super fascinating is that when you do that, you could have a lot of people sell their shares, and then you could theoretically lose control. 

Palantir did something extremely fascinating, very much like how Facebook did it, where they've actually created these multiple classes of shares. Class A shares, or what you're selling, those just have a single vote, but Class B shares—of which Peter Thiel has actually the largest block—has 10 times the number of votes. But then they created this new type of share—I have never heard this before. I’m curious if another company has done this—called a Class F share. And what this does, this gives the three founders 49.999999 of the total voting power. Did I get all the nines in there? Yeah. Six nines.

Corey: In other words, the entire world has to unify against them—

Pete: Exactly.

Corey: —everyone participating in order to overrule them, which is, frankly, not happening.

Pete: And what it means is that anyone can sell as many shares as they want because a direct listing means from the IPO, anyone can sell any actual shares in the company. And they can do that without the three founders essentially having to all be overrun like you said. You'd have to get everyone else on the other side of it. So, they have to be included somehow. That was just super fascinating, but it's how they're going to maintain control while allowing people to just sell their shares.

Corey: This episode is sponsored in part by ChaosSearch. Now their name isn’t in all caps, so they’re definitely worth talking to. What is ChaosSearch? A scalable log analysis service that lets you add new workloads in minutes, not days or weeks. Click. Boom. Done. ChaosSearch is for you if you’re trying to get a handle on processing multiple terabytes, or more, of log and event data per day, at a disruptive price. One more thing, for those of you that have been down this path of disappointment before, ChaosSearch is a fully managed solution that isn’t playing marketing games when they say “fully managed.” The data lives within your S3 buckets, and that’s really all you have to care about. No managing of servers, but also no data movement. Check them out at chaossearch.io and tell them Corey sent you. Watch for the wince when you say my name. That’s chaossearch.io.

Corey: So, changing gears slightly, is this company making money? I mean, there is enough of a marketing problem that they have; there's a clear question of if people do business with them, they're not going to be clamoring to do a testimonial; and anything to do with government, that's just a given, regardless. So, are they making enormous piles of money for the trade-offs that they have agreed to make?

Pete: Well, there's a lot of people online that are really happy to go and crap all over how much money this company loses, and I'm happy to be yet another one of those people that's going to crap on how much money they lose every year. Their revenue is growing. They're doing, I think, over a billion dollars a year in revenue; that's huge. But they're also burning half a billion dollars a year. $500 million a year burned—

Corey: On top of that revenue. So, they're basically making a billion dollars a year and losing half a billion dollars a year because they're spending at one and a half.

Pete: Exactly. And in some cases, their margins are thin, and their cloud spend is high. People maybe not renew. But again, what's interesting about this one is that they're—I think they talked about like the average deal sizes, these are measured in the millions of dollars. So, they're not selling Datadog: “Oh, I got my five agents and they cost me 100 bucks.” 

They're selling seven-figure deals to large organizations: the biggest enterprises, to military, and government, etc. These are really large deals. And having been in that space, having to sell to the government, having to sell the big companies, those deals take years to complete. So, it's kind of no wonder that it takes them so long and that they burn so much cash to do it.

Corey: One question that comes up is what are the alternatives to using something like Palantir? I don't know enough about what they do and how but have got to be other options because again, this feels like it's relatively old tech. There has to be something relatively new that would come out and solve for this problem in varying ways. I have no idea what they might be, but I can't accept that this company is just sitting around with zero competition, not with only making a billion a year.

Pete: Yeah, and they even list in their S-1—they don't actually list direct competitors. They speak more abstractly into—that they are this big data, data play type of thing, and again: secretive company. It's hard to know from the outside what they're doing. It kind of feels like they're going into these places with their top tier engineering talent, with a core product that has the ability to make inferences between data, between different data sets, and give people insight into something they didn't have before. My guess is, is that what they're doing is that they're maybe not building custom solutions for everyone, but I bet there's a good amount of custom development work that they're probably doing when they go in and they sell one of these seven-figure deals because if you have to remember, you could be in a sales cycle for one to two years, you're probably building some custom stuff based on what the customers are asking for. 

Maybe those features then tie into what someone else wants to buy as well, but they did talk about the different platforms they have, but they really didn't talk at any sort of length in their S-1 about really what the thing is they're offering other than a way of tying data together, making inferences into this data from different data sources. I almost wonder if the secretive nature of them actually is their marketing. Like, that is what gives them the hype and the aura because of how secretive they are.

Corey: And that's going to be an interesting challenge. One other area that I found fascinating to look at was something you highlighted, namely that they've committed to $250 million a year to their primary cloud provider—which let's not kid ourselves, it's AWS—they've only used about $40 million of that in six months, so that's an $80 million year run rate. Someone either over-committed massively, or they plan to have a massive growth hockey stick, which is a bit ambitious if I'm negotiating the commits. Is there any other way to read that?

Pete: That was probably the most shocking—again, there was a lot of shocking things in this S-1. Not really shocking, but things that were like, “Whoa, that's weird.” This one, from our world, we do so many negotiations on EDP commitments, seeing something like this where you've committed to essentially $250 million a year, but across six months, you've only met $40 million of that, if I was the operator there, the person who negotiated this, I might be raising my hand and be like, “Well, well, hold on a second. What just happened here?” 

And it means one of two things. Either they massively overestimated how much spend they were going to need, or there's a potential that they actually have. Maybe they have some really large deals that are on the way. And I have no idea if this is the case, but around—what was it—last year or whatever, Amazon and Microsoft have been battling over this Jedi contract, which is modernizing the government to use Cloud. Well, here's Palantir, who has these agreements with two cloud providers, and they sell heavily to the government. 

So, is this essentially their, “Yeah, we're expecting more of the government moving to Cloud.” Because they clearly want that. They want to be in more Cloud. This type of deal that you see is showing that they want to be in Cloud and this is potentially their big hedge on it. There was one line that I did think was interesting is that it didn't look like their downside commitment—like they didn't have a big downside if they didn't hit it. It basically said if the difference between what they committed to per year, and what they hit was off by, like, $30 to $50 million, that was essentially what they would have to pay and the rest would fall into a future year to use. So, I mean, I'm sure they don't want to pay money they don't have to, but it seems like their risk, maybe is not use it or lose it model.

Corey: Yeah, but there are ways that winds up structuring out. In some cases, you're right. I mean, it all depends on exactly what was negotiated at termination point. I do know that when you start committing to certain yearly spends, they start forecasting and counting that revenue, so, “Hee hee, we do enough of these where things don't hit, and we just let people pay it off whenever they want, now we have to restate earnings.” And that's something Amazon is disinclined to do.

Pete: Right. And I wonder, too, if you're a few years into this agreement and you need to renegotiate things, at this kind of spend I feel like you have a lot of flexibility especially to—

Corey: We’ve seen this publicly with public filings from both Pinterest and Snap, having done that. Snap committed to, I think, $2 billion with GCP, and then within the next six months, another billion on AWS, and it's maybe committing to $3 billion of cloud services over the next five years isn’t… the most responsible way of playing this. And sure enough, they extended it out another few years and added some token amount on top of it just to give them more time to burn through it. We saw the same thing, to a lesser extent commitment-wise, with Pinterest.

Pete: Exactly. I think that—

Corey: All through their public filings, to be clear.

Pete: Right, right. And I think that is what is fascinating is that this commitment exists; we can all read it; it's all public information, but anything can change at any time. It all depends on who's got the leverage in the negotiation. And if I was running a lot of things in a cloud provider and I had a way to run them in another cloud provider, even for a little while just to show that I had that capability, that can be a negotiating point. But you could also say to that, like, you could always renegotiate and just extend your end game out later. I think everything's up for negotiation when it comes to these EDPs. When it comes to these enterprise agreements, there's a lot of flexibility.

Corey: Yeah. I think we're going to see what happens because the nice thing about companies going public is they have to update us on these things every quarter, whether they want to or not. It's weird to me to see Palantir going public because for so long, they've taken the official position that much of what we do would not be well served by requiring transparency, which is always the sign someone's up to wonderful things. But, well, you want the money, you've got to go ahead and do the filing pieces. So, good luck.

Pete: I think it will be interesting to see their filings over the next few years, and I'm sure there's going to be a lot more people out there much smarter than us with their own thoughts and opinions on what they're doing and how they're doing it because it's all going to be pretty open to the public eye.

Corey: Yeah. Pete, thank you so much for taking the time to go through yet another one of these. We'll attempt to go in a different direction next week, but given that everything is going public, who the hell knows.

Pete: Thanks so much for having me again. It's always a blast.

Corey: [laughs]. This is the AWS Morning Brief. I'm Cloud Economist Corey Quinn, and if you've enjoyed this podcast, please leave flattering comments about me with your five-star review, whereas if you hated it, please insult Pete along with your five-star review.

Announcer: This has been a HumblePod production. Stay humble.