Pershing Square Challenge 2026决赛选手推介 Amadeus $AMS｜全球旅行的收费站

Today, I have another team on from the Pershing Square Challenge. This is Team Amadeus. They were finalists, and I think they did great work. Amadeus is a really interesting company, and you're going to hear them talk about it throughout the podcast.

Amadeus has come onto my radar multiple times over the past couple of months because I've had a lot of people poke me and say, “Hey, Andrew, this company is down 30% over the past few months. This is about the most AI-proof company in SaaS land that I can think of.” The team has done really great work, and you're going to listen to this and understand why so many people think this is an AI-proof company.

The other reason it has come onto my radar is because Constellation Software, the big company up in Canada, invested 10% in a publicly traded competitor of Amadeus. So, if you're interested in Constellation Software and Sabre, which is the competitor—and Sabre is quite levered—you'll get some background on the industry. Those are the 2 reasons it has come onto my radar. The team does a great job diving into both of those and a whole bunch of other things.

So, guys, before we dive into Amadeus, which is what you pitched, I'd love it if you could just give a little bit of a 10-second background on who you are and how you came into the Pershing Square Challenge. Kabir, you're in the top left of my screen, so, if you don't mind, I'll start with you, and we'll go clockwise—or counterclockwise—just so we hit everyone.

I'm Kabir. Prior to the MBA, I worked in consulting. I'm very passionate about investment management, and I took up the Pershing Square Challenge to learn more and dive deep into a company that I wouldn't be able to do in any other way.

Awesome. Fran, you want to go next?

Yeah, sure. Thanks, Andrew. Hi, everyone. My name is Fran. Prior to CBS, similar to Kabir, I was a consultant. I was based in the Middle East, actually. Likewise, I was very excited to work on this project because we look at 1 company for around 4 to 5 months, which I think was a very unique opportunity.

Cool. And best for last.

Before the MBA, I worked on the buy side and then in corporate strategy. I took up the challenge of working on this pitch before and was super excited going into CBS that I would have the chance to do this. I'm happy we got to the finals.

Awesome. I'll just share a little inside baseball. I've said it on a few of the podcasts I've done so far: Every judge told me this was the best set of contestants they'd had in the Pershing Square Challenge by far. I know 1 judge who was over the moon about Amadeus. I think you might be long Amadeus, but I know 1 judge who was really pounding the table on your pitch and everything.

Let's dive into it. You guys chose Amadeus, and before we get into why you chose it, maybe quickly, who wants to start with what Amadeus is? They are certainly not a household name, even though I would guess every single consumer and every single podcast listener has indirectly interacted with their products.

To anyone viewing, I think Amadeus is like the tollbooth on global travel. They're a company that makes money every time a flight is booked through a travel agency or an airline boards a passenger. They get a fee on the booking.

It is the number 1 player in that space, with roughly 50% share in its core segments, and they've been growing revenue at 7% and compounding EPS at 11%. They operate across 3 segments. Air Distribution is the 2-sided platform connecting 400-plus airlines to travel sellers. Air IT Solutions provides those systems that serve passengers, which are the backbone of airline operations, and the Hospitality segment provides reservation and guest-management systems for hotels.

In general, they hold a lot of data and handle a lot of transactions per second, which is somewhat similar to Google Search.

That is a fantastic background, and I'd love to dive more into the company in a second. I've been asking all the teams from the Pershing Square Challenge this question, so I'd love to ask you: This is a semester-long project for you guys, with the final pitch being the deck. You can choose pretty much any company in the world with sufficient size.

How does the process look for you guys when you're sorting through a lot of things? How does Amadeus come onto the radar? How do you decide, “Hey, I want to spend a semester researching this company and crafting a pitch”? You've got an eye toward the pitch, too, so what makes you think Amadeus will make a good pitch?

I can talk a little bit about this one. Of course, we wanted to pitch something that would be interesting for the judges, and we knew what the portfolio of Pershing Square was, so we took that into consideration. We also wanted to pick something that was interesting for us to do research on and might be useful down the line.

The process was really fun. We thought that every 3 days we would meet, and each of us would present a primer on 2 companies. We did that for a couple of weeks, and then we ranked them and voted. We came to the conclusion that Amadeus was interesting, not only because it was related to travel, but also because it had implications for software and AI.

We also looked for feedback from our professors and TAs, and they liked it. It was a consensus in the end that this was a good idea to work on.

Does anyone else want to add anything there?

No, just what was said. That's exactly how we approached it. On top of that, we did look at other names. We were each responsible for bringing 2 to 3 names of companies that we found interesting and that fit the Pershing Square criteria we were given.

Then we would start discussing the company and having some high-level points on what the thesis would look like. At the end of the day, similar to what Ethan said, Amadeus just turned out to have the most interesting thesis, mostly because of the AI selloff during that period.

Perfect. Yeah.

Well, look, I think you guys gave a great description. Between Kabir's overview of the business and both of you mentioning AI, I think we can start diving into the company. I'll try not to put too many words in y'all's mouths, but I think for a long time this was viewed as kind of a tax on global airfare, right? They handle a lot of the global airfare, and as Kabir said, every time someone books a ticket, they're probably getting a cut of that ticket for handling the software.

With the business, you're starting to hear some AI risks and everything. So I'd love for you guys to talk to me about how ingrained this is, how you're thinking about AI risk, and how the market is thinking about AI risk when it comes to this business.

Oh, wait. You're on mute.

Oh, sorry. I can take this one. When we started doing research into it, it was clear from earnings call transcripts and sell-side reports that AI had completely overtaken the narrative. There were 2 components to it. The first one was the potential disintermediation of the distribution business, but also the potential for airlines to develop their own IT systems more cheaply and replace the passenger service system.

There were also implications for Air IT Solutions, and I would say we did most of our primary work around this area because we wanted to really be comfortable with the downside risk coming from this. So I would pause there, and maybe, I don't know, if Kabir or Fran want to add something.

Yeah, I can just add that the market sees the AI risk similarly to how it sees other software-as-a-service companies. We've seen multiple compression since COVID in 2021, and then again over the span of the last couple of years. So, with what Ethan said, I think the market has seen that multiple compression, which has affected the stock tremendously.

I want to provide a few thoughts. I don't have any huge disagreements with this, and there's one slide I want to get to in a second that I thought was the best slide in the deck. I thought it was one of the best slides I've seen in here.

We'll cover valuation more fully in a second, but you look at the stock chart, right? The stock has gone from the low 70s a year ago to the mid-50s today, right? That's a decent-sized move. It's a little bit more than a 20% drawdown. There is some debt here, but it's not a highly levered company or anything, so that is a pretty big drawdown for a not-that-levered company.

On the other side, I look at this and say, “Hey, they grow 6% in 2025. The core business is kind of a mid- to high-single-digit grower,” which makes sense for a pretty penetrated business. This is the number-one player in the airline distribution backend software, and it makes sense that it's growing GDP plus, because that's kind of what air travel grows at.

But it's trading at around 10 times EBITDA. So I look at this and say, “It doesn't look that obviously cheap to me.” I've had several friends pitch this, or some version of this, to me, and it's, “Hey, there's SaaS fear here.” I'm going to get to your slide in a second, but when I look at the valuation, I'm like, “It doesn't look that obviously cheap to me.”

I think the company—I was just looking at its Q1 report, which I realize is kind of after your presentation—but they've started buying back shares a little bit. Especially for a European company, it might be kind of aggressive, but I can point to several other SaaS companies where the stock is off 75% and the company is out here saying, “We're going to be okay. We're going to be okay,” and they're buying back shares.

It's just not striking me as this huge dislocation, and that might be fine for the Pershing Square challenge, right? I think a lot of people believe this is a compounder, but I feel like I came into this thinking it was a SaaS victim, and it's just not obvious that it is to me. I threw a lot out there. I do want to get to the slide in a second, but I'll let anyone who wants to respond.

I'd just say it is cheap to me. This stock used to trade around 22, 23, or 24 times earnings, and it's trading at 15 times earnings. If you run a DCF, that's like a fair multiple—22 or 23 times. Consensus EPS growth is around 12%, which is in line with guidance.

For a 12% EPS grower, as you said, unlevered and trading at 15 times—15 times GAAP, or P/E, right? If you look at other SaaS names, you would have to look at the GAAP numbers to be comparable. Even though software has come down quite a bit—more than this stock specifically—if you look at GAAP earnings, it's a lot more expensive than 15 times.

No, look, you did hit the nail on the head there, because one of the things—and I've hammered people on this before—you look at these companies and say, “Oh, it looks cheap,” and everybody says, “It trades at 10 times free cash flow.” I keep hammering people on this: “Hey, you're saying they've got $500 million of free cash flow, but they've got $700 million of stock comp, right? And that's a real expense.”

Maybe when the stock was trading at a $50 billion valuation, you could kind of ignore it. But when the stock is down 90%, it's now a $5 billion company, and they're doing $700 million in stock comp, even if you want to ignore it, that diluted share count is going to go up real fast.

This company—the EBITDA and the adjustments here are very clean—and as you're saying, it's 15 times. It's a very clean—I said 10 times EBITDA, 15 times P/E, whichever you want to say. Anybody else want to say anything on the multiple, or I'd love to talk about the slide I keep alluding to.

I want to say something on the SaaS business. I wouldn't say it's directly comparable to another SaaS company just because it's not a per-seat business. They sell directly to airlines, which controls the software all the way from check-in to boarding.

Similarly, they have a distribution business, which aggregates data for all the online travel agencies. So you're not really selling per user, and it's not impacted by whether there are going to be fewer users in the future.

It's perfect. The slide I keep alluding to—and again, there'll be a link in the show notes—is slide 15 in your deck. It is titled, “Amadeus is insulated from the SaaS apocalypse and AI: Not all software is created equal.”

You lay it out in your classic 2x2 matrix. You have how likely something is to get disintermediated, and you guys have Amadeus rated as the least likely to get disintermediated by AI. You've got it better than Visa, Oracle, SAP—everything. You've got Amadeus Air IT Solutions as the best, and Amadeus Air Distribution is kind of the best there.

So I'd love to ask you why you think this is so ingrained, so sticky, so unaiable, and unreplaceable. I don't disagree that it's very hard, but I will have some gentle pushback on perhaps it being the most insulated in the entire world.

Sure. I can take a first stab at it, and then, Kabir, you can continue. As you said, we have divided it into 2: Air IT Solutions and Air Distribution. I can talk about Air IT Solutions.

If we start with the IT solutions, these are systems that were built around 40 to 50 years ago and have been updated, right? But these are systems of record that already have a lot of data from the airlines, and these are highly mission-critical. If the system isn't working, the plane will just not take off. That's it. As simple as that, right? Planes and airlines cannot afford to have any downtime whatsoever.

That's why we have Air IT Solutions as something that is highly insulated from AI, and why we have it as very different from other types of software companies. We validated this as well through calls with multiple people who are experts in the field. For example, we talked with a Microsoft AI director who told us exactly the same thing: They're not seeing, at Microsoft at least, when they're helping clients with AI, any of these types of software workloads that are deterministic in nature and highly mission-critical being replaced by AI.

It's very hard to replace them in terms of money, time, and training as well. A lot of these systems have people already trained on them. Even switching from one system to another is a nightmare for these airlines.

All right, we had some extreme technical difficulties, completely on my side. I don't know if you guys have ever looked at IWG, which is the kind of WeWork competitor that owns a bunch of different brands. I'm at an IWG brand, and it seems like they forgot to pay their cable bill because my internet completely went out yesterday.

But Fran's been kind enough to rejoin. We're going to finish the podcast, and hopefully Kabir can join. Fran, I was asking, when it cut out—and I'll try to edit everything—I was asking about slide 15, which is this great 2x2 slide showing how exposed software is to AI.

If you're all the way at the bottom left, you're the most exposed. You have Duolingo as kind of the most exposed. If you're all the way at the top right, you are the least exposed to AI: You are mission-critical and deterministic in what you do for a business. And you guys have Amadeus's 2 businesses, Air IT Solutions and Air Distribution, as basically the least exposed.

So, I was reviewing some of the stuff we’ve been talking about. I’d love it if you could clearly define, first, what Air IT is and what Air Distribution is, because I think people know, hey, they do flights at this point, but what specifically are they doing? And why do you guys think these are so unexposed to AI?

Yeah, sure. Maybe I can start with that, and then you can comment on it. Basically, as we said before, Amadeus has 3 main businesses. The 2 most famous ones, let’s say, are IT Solutions and Air Distribution.

In IT Solutions, the way that you should think about it is as the SAP of airlines. IT Solutions basically runs the whole IT infrastructure of airlines, all the way from checking in passengers and luggage, to the systems in the airport, to when the plane needs to take off, and all of that. Everything that the airline is doing under its IT infrastructure is basically done and run by Amadeus. That’s how you should think about IT Solutions: just as SAP.

Air Distribution is different. In Air Distribution, basically what Amadeus is doing is connecting and showing the content of the airlines to travel sellers. Travel sellers can be online travel agencies, like Expedia and Booking.com, which are the 2 biggest ones, let’s say, or brick-and-mortar travel agencies as well—TMCs, which are travel management companies, for example, Amex Travel and BCD Travel, which are used by a lot of big companies.

They’re the ones that enable these travel sellers to see and show content about flights, prices, seats, and really everything from the airlines. They handle the booking internally for them through these systems. In a nutshell, this is what these 2 lines of business are.

In terms of AI—and let me know if you have any questions—we divided them in the matrix we have because we believe they have different levels of AI-risk exposure. I can talk about IT Solutions. As I mentioned before, this is like the SAP of airlines. This is a system of record that has been around for around 50 years. Airlines rely on it fully, right? If an airline is using the Amadeus system, they rely on it fully.

To put it simply, if the system is not running, the plane will not take off. It’s as simple as that. And as we know, airlines cannot afford to have any mistakes. They need to be constantly running, and they need to be constantly flying their planes in order to make money. That’s why we have it as the most deterministic system and, at the same time, as the most critical system there is, because the airline system is very critical.

It’s very hard to replace, even based on our discussions with a lot of experts and with airlines themselves. They don’t even switch from one system to another, let alone replace the whole system just to vibe-code their own Air IT solution. This is not going to happen, especially because it only represents around 1% of their revenue. It’s just that small and that highly mission-critical. That’s why we have it as highly insulated from AI risk.

Maybe Ital or Kabir, you guys can discuss Air Distribution.

I can add something. What’s important here is that this isn’t something we just came up with out of the blue. It was informed by our research on some of the most sophisticated software investors out there. Critically, this was informed by a conversation we had with the Microsoft AI director.

He is the one who goes into a company or into a government and tries to help them develop agents to simplify their processes. You would think that he would be the most bullish on AI eating software, and we got the complete opposite. He was like, when you’re talking about mission-critical systems of record, there’s no way companies are going to take the risk to develop their own systems.

Perhaps the most interesting example he gave is that he had experience with one of the competitors in the space. So, he knew very deeply what type of systems they had, specifically talking about IT Solutions and passenger service systems. He was like, “Yeah, no way. I would bet my money that this is not going to go away in 5 or 10 years.” That’s how we formed our view: we got comfortable with the downside risk.

That’s great. And Kabir, welcome back. Thanks for joining. Do you want to add anything there? I did have some follow-up questions I wanted to ask.

No, I think they covered it pretty comprehensively. After your questions, we can probably discuss the Air Distribution segment.

No, it’s a great overview, but let me just give you one perspective. I don’t know if I firmly believe this, but the one thing I keep saying is that the AI systems we’re working with today are the worst we’re ever going to work with. I use Claude Code all the time, and I get pinged like 4 times a day: “Hey, we’ve got an update. Hey, we’ve got an update.” They’re all going so fast.

I do hear you on these things. When you’re flying planes, you can’t be down for a second, right? I think they have a slide in there—I forget if it was them or someone else—that’s looking at it like, “Hey, 99.9% of the time, it does work.” It’s got to be like the old Six Sigma thing. It’s got to be up all the time. Even if you’re just connecting to travel agencies, if you’re down for an hour, that could be so much lost revenue.

But I guess my pushback would be that I’ve interacted with the airlines. I know one of your research pieces has the photo. You guys went out to JFK and were talking to the people who are actually putting the things together. You go behind the counter and look at that computer there, and it’s like running 1990s tech sometimes, right?

That’s a great thing, in terms of, hey, it shows how hard it is to upgrade and change these. But I would just have to imagine that the technical debt has to be so high. At some point, aren’t you going to see one airline want to replace it? Airlines start up all the time, so are startup airlines going to start with tech-debt-free systems of record? I guess that would be where my pushback is. At some point, it just seems the technical debt is so high here. I’ll pause there.

Yeah, my initial thoughts on that are that the system is already working well for them. While speaking to someone who was a former senior VP of engineering at Amadeus, they said that you wouldn’t just replace a system that’s working too well, that costs about 1% of your revenue, with something that might work well.

What an implementation replacement really looks like is that you run one instance of your software that’s running today, and then you run a second one where you’re really developing another instance, because you can’t have one being created while the other one is not working. So, that adds a lot of incremental cost to a carrier to implement.

Secondly, we see that Amadeus is spending a lot on its R&D to make its software new and really adapt to airline needs. We saw that with Nevio more recently, and Finnair has been an early adopter, specifically on the Air IT Solutions. I know that doesn’t really impact the back end of what someone sees when they’re checking in at an airport, but we know that Amadeus is really looking at airlines’ needs.

We also know that they aren’t exercising much pricing power to actually gain much more out of the airline, because it’s already a very thin-margin business for airlines. So, at least in my opinion, they wouldn’t just go with any other provider instead of a rather established company like Amadeus.

You guys mentioned some of the research you did in terms of talking to the Microsoft people. I mentioned that you guys went to JFK and talked to people. If I’m just glancing at the slides quickly, I think you guys talked to about 12 customers of theirs.

Were there any customers you talked to who gave any pushback along the lines of—it doesn’t even have to be, “Hey, we’re ready; we would switch”—but were there any customers who alluded to what I said? Did they say, “Hey, there’s a lot of technical debt here”? Doing an implementation, trying to rip your heart out while you’re running, but doing it on an airline that needs to be up all the time, would be like ripping your heart and brain out.

Were there any customers saying, “Hey, the technical debt is getting high enough that maybe we would consider it at some point if somebody could really hold our hands through it”?

We heard that, but mostly around the IT Solutions space. Amadeus has 50% market share. The second player is 3 times smaller—Sabre, with 15% market share—and then you have some large airlines in the U.S. that have their own internal software that they’ve used for the last 20 or 30 years.

The commentary about the technical debt that we heard was about those airlines specifically: “Hey, this is getting obsolete.” At some point, some CEO is going to come in and have to bite the bullet and either completely upgrade their systems or hire someone else to do it.

The commentary we got from experts was that this was one of the biggest opportunities for Amadeus to come in and take—

I’m glad you mentioned Sabre, who’s the number 2 player here, because this happened after you guys made the presentation, but I was laughing.

I was prepping for this call. I reviewed the Sabre Q1 call, and Sabre is interesting to a lot of value investors because Constellation Software—you know, maybe the best-performing Canadian stock of all time—bought about 10% of it on the open market. They’re quite levered, and it’s a very interesting play.

But on their Q1 call, Sabre came out and said, “Amadeus is taking share because they are a monopoly, and we want the DOJ to investigate them.” We thought they wanted everybody to investigate them. Amadeus got asked about it on their call, and I’ve never heard an analyst ask someone on a call, “Hey, your competitor accused you of being a monopoly. What do you think?”

Amadeus basically said what Ethal[?] is kind of driving to. They said, “Look, customers are choosing us, and I don’t think it’s because switching is really hard. Customers are voting with their dollar because our product is just better.” It’s like, why do people watch Netflix instead of legacy cable? Netflix has the stuff they want to watch without ads, and it’s cheaper. It’s just a better product.

I don’t know if you guys want to comment on anything there, or if some of the value heads would be interested in your take on Sabre, if you’ve done any work on that company.

Yeah, I would say what pops into my mind is that it’s funny that they’re complaining, but at some point a couple of decades ago, Sabre was the number-one player. The reason Amadeus is the dominant player now is, of course, a combination of a better product, reinvesting into the business, and constantly improving their platform. But it’s also Sabre’s own doing, right? Sabre has gone through the hands of private equity, I think, a couple of times, and was always quite levered, so they didn’t have the ability to reinvest into the product and suffered a lot through crises, most recently COVID.

COVID was a big event where Sabre lost a lot of share. Specifically, for the pandemic, what Amadeus did was brilliant. All their customers—the airlines—had all their flights grounded. They were severely constrained and were trying to get cash anywhere they could find it.

So Amadeus actually gave them better payment terms in exchange for larger content deals. That’s how they took a lot of share during the pandemic. Of course, Sabre couldn’t do that because it was—I don’t know—it was probably 5 or 6 times levered going into the pandemic, and that was one event.

But also, jumping into air IT solutions, you need to constantly do implementations with clients and improvements. This is an investment that comes out of the pocket of the PSS company. Again, since Sabre was always highly levered, it couldn’t really invest in the integrations and the developments for its clients.

Let me just—you said COVID was a big event and they lost a lot of share, right? I think that comes back to my other worry, right: How is this supposed to be irreplaceable? How was Sabre losing share? Was it just the airlines they were with went bankrupt, so no new airline would dare start on Sabre because they have so much technical debt? Or was there something else that was causing them to lose share?

Well, I think, as mentioned, it was 2 things. For example, let’s talk about air distribution first. As mentioned, a couple of decades ago, Sabre was actually the number-one player, but then what happened is that the company had internal issues. The company was private-equity-owned as well at some point, and it became highly levered. Right now, I think the leverage is around 9. It was big the last time we researched it.

Usually, when a firm is private-equity-owned, the private equity firm’s interest is really just to take cash out of the company and make returns, right? They want to take cash out of the company; the focus isn’t really on reinvesting and making the product better. Across all of this time, Amadeus was making its systems and products better, so it was able to gain market share, especially when it comes to air distribution and also IT solutions.

But when it comes to air distribution, network effects are really important. Because Sabre went through this downturn when it was private-equity-owned and wasn’t reinvesting as much in the business, Amadeus started to make its product better.

Most airlines want to be connected with the platform that is showing content to the most agencies, and vice versa: Agencies want to be connected with a platform that has the most content from airlines. So over time, because Sabre had a downturn, because Amadeus was able to make its product better, and because of the network effects, it’s just a compounding effect of how Amadeus has been gaining market share along the way, which is why for now it’s around 50%.

Let me go—and I think you were going to talk about air distribution, so maybe I’ll switch over to you. But just to mention, Amadeus invested very smartly in COVID and gave people extended terms in return for more content on air distribution.

I guess when I look at air distribution, as just an outsider who spent half a day looking at this, I look at it and say, “Oh, well, similar to Uber, right?” With Uber drivers, Uber would always say, “We’ve got this network effect,” and then a lot of people pushed back and were like, “Well, Uber drivers—it’s just that a lot of them are buying multiple phones, right? They go on Uber, they go on Lyft; wherever the ride is coming, they’re going to grab it.”

With the distribution business, because you guys have this as the most irreplaceable business in software in your 2x2 matrix, I look at that and say, “Hey, why wouldn’t I, if I was any airline, be on Amadeus? I’m on Sabre. Somebody comes with an air-distribution network and says, ‘Hey, we’ll plug you into travel agents or whoever. We’ll plug you into the OTAs, whatever, through this.’ Why is it—why am I working with 1,000 people on this?”

I can tackle that one. Fundamentally, the GDS business is not a very good use case for LLMs, and the reason is that it’s more deterministic than probabilistic. Today, as it stands, the cost of inference is 30 times more than the cost of using an API. Cost-wise, it doesn’t make the most sense.

Even in the future—I know you mentioned that AI is only going to get better and the cost of inference is only going to go down—let’s say that happens. There are still only 2 ways to extract airline flight data.

One is through web scraping. If a new competitor is coming into the business—a disruptor, a small company that wants to get airline data—instead of forming partnerships, they scrape through the airline’s website, put all their flight prices there, and that’s one way they can aggregate content. However, that today is against the airline’s terms and conditions.

So there’s still only 1 way, which is forming partnerships with the airlines and getting their content. That’s also partly because this content is very dynamic to scrape today. Prices change very quickly based on which customer you are, because the airlines are learning more about you.

The way to do it is with an API, and Amadeus has connections with over 400 airlines, which is what they’ve built over the last 30-odd years. This is not something a disruptor can come in and replace from day 1.

I definitely agree with that. But let’s say the 4 of us pool our money and we’re going to start a new distribution network. What would preclude all these airlines? Why would an airline say, “Oh, we don’t want to connect with you?” Right?

My worry is—and I understand this has evolved into the idea that Amadeus is a mountain—but I’m just trying to understand why this isn’t just a free-for-all and everyone’s connecting to everyone, in the same way I use the Uber-Lyft driver: I’ve got 5 different phones.

Yeah. So, I guess, to begin with, airlines are trying to make the push to direct connections and eliminate any third parties overall. However, the main benefit Amadeus has today is the network effects. They have these 400-plus airlines, so for a large carrier to leave this two-sided platform means that customers don’t really see their content.

Essentially, if I’m a traveler and I no longer see an American Airlines flight through Amadeus’s content, I’m very likely not to even see that flight overall. So it’s not in the airline’s best interest to leave Amadeus.

So, if I understand what you’re saying, everyone who’s done this on Expedia or whatever knows that all the airlines always try to get you to book directly with them, right? It’s higher margin, and they’ll give you incentives. They generally can’t offer cheaper flights—hotels do this too—but they can say, “Hey, if you book through us, you get extra rewards,” all this sort of stuff.

And what you’re saying is, if we started a distribution company right now, airlines would not be incentivized. They don’t want more plug-ins coming in from new people, but they’re already on the Amadeus network. If they actually pulled out—you think about Delta pulling out—cool, you’re not on Booking and Expedia anymore.

They’d be losing so much revenue that they kind of can’t make that switch, but they’re not going to just be like, “Hey, new guy who’s going to give us no revenue. Yeah, come grab all our data. Come get all our APIs. We’d love to hook up with you.” Even though there’s no cost, it’s just not what they want to do.

Am I driving that correctly, or is there a piece of the story missing?

I was going to go there: if you're an airline leaving Amadeus's platform, you're leaving incremental revenue on the table, right? And if you're a travel agency and you leave the platform, you're leaving revenue from incentives that you're getting from Amadeus. So that's kind of the two reinforcing things that are going on that keep the platforms going. I think, also to that point, you should keep in mind the industry structure. There are only 3 companies in the world. That's for a reason, right? If it were so easy, then this would be a very fragmented market, and that's not the case, and it hasn't been the case for the last 20 years.

No. Look, that was—I think I'm really starting to drive home why I've had several friends who are more into really good businesses saying, “Hey, you've got to take a look at that.” And I can see why. Constellation is famous for buying businesses and holding them long term. It seems like they're just going to sit on them. I can see why they said, “Hey, Sabre, maybe a little undermanaged, but it's a damn sticky business, and it's tough to get them.”

Let me bring up one more thing. I believe this happens, humorously, after you guys submitted the presentation, but I think the day before you actually present at Pershing Square, Amadeus buys a biometrics company. We don't have to talk about the acquisition per se, but it does jump out to me: whenever I've looked at a business, one of the telltale signs that they're about to go ex-growth or they're starting to plateau is they go buy something. And the biometrics business—I saw that acquisition, and I was kind of like, “Hey, I get they're saying there's a little synergy there, but it's not core, and it seems like maybe they're trying to find a new growth avenue.” So we don't have to talk about biometrics specifically, because I know that might have occurred after, but what would you say if I was pushing back and saying, “Hey, it looks like they're really trying to find some growth vectors here”?

Yeah, we got completely caught by surprise. I remember vividly waking up the morning before the finals and getting an update on the Quartr app: a $1 billion acquisition by Amadeus. Part of our narrative, especially in the appendix slides, was that they had gone from large acquisitions before 2016–2017 to more smaller ones. So, yeah, we were caught by surprise.

I'd say when we looked at the merits of the transaction a little more deeply, we were more comfortable with it for a couple of reasons. First, our take was that this is a very high-quality asset in the biometrics space, ideally PE-owned, and they had to exit, so it was more opportunistic. They have very sticky clients in airports around the world and governments, so it's a very high-quality customer base. Second, I'd say biometrics does make sense, and the company has said it aligns with their long-term strategic vision of being an orchestrator of all travel. So, of course, they see some kind of long-term optionality value from having an asset like this. It reinforces their position, and there are some benefits of being the technology provider for the airport and the airline as well.

Finally, we like the price a lot. They paid 10 times EBITDA. When you factor in the growth outlook for that business and the cost synergies that they put out, it actually turns out to be like 6 times EBITDA by 2028–2029. And when we did a quick IRR calculation, it was actually accretive to our thesis. So that's kind of what we thought.

I'm just looking through my notes. I've got 1 or 2 more things I want to touch on, but one thing in my notes—and Kabir, I think you said it, but I'd love to touch on it one more time—is when I was asking about the LLMs on the distribution business. I think you said, “Hey, inference is 30x more expensive than deterministic,” which they do for the inference. We don't have to talk about that specifically, but I think you talked to someone at Gemini who said, “Hey, this is doable through LLMs and inference.” Ben, I just love when you talk to the AIs. Just click on that one more time on why they were telling you, “Hey, this isn't in the roadmap. There's probably easier rows to hoe than this.”

Yeah. So I think while speaking to the Gemini person, we learned that Google as a firm has already looked at this problem and, for a reason, decided not to pursue it further. Now, in the mid-2010s, they acquired ITA Software to get into the space. However, that technology became obsolete, and then they realized that instead of integrating further into this industry, which didn't make financial sense, they came back and partnered with Amadeus last year.

And what this gave us confidence in was that it's very difficult for a large AI firm to justify the spend on building a distribution system that has a combined revenue pool of $6 billion when they operate at the top of the funnel, aggregating demand, with a $65 billion revenue pool from the largest OTAs, like Expedia and Booking.com. So we learned that since they're able to perform better in that space, they might as well not really dig into development and focus on a $6 billion pool, because they're also much larger companies and are focusing on those bigger projects.

Secondly, LLMs today have an uptime requirement. I know we spoke about 99.99%, which in a year means about 50 minutes of downtime. LLMs have 99.3% uptime, which is about 87 hours. So that big gap, today as it stands, will not work for an airline, because it means grounded planes, shut airports, a loss of revenue for these large airlines, and also just disruption across the global economy. And for Gemini, it would be easier to build a platform that has a much higher uptime. It's just not worth the entire investment that they would have to make.

Perfect. I want to go back, and I believe—it's hard to remember—I can't remember if we talked about this before, but I believe in the first half, before my internet went out, we talked about valuation. I kind of said, “Hey, look, it's down 20% or 30%, but it's not screaming cheap.” That was just the pure quantitative cheapness side. But I do want to come back to the valuation and ask: forget the multiple, toss the multiple out. This is trading at €55 per share right now. How do you guys look at it, and what do you think the fair value for this company is?

Yeah, I can talk a little bit about that. I think we talked about some of it before the internet issue we had. This company has, over the last 15 years, grown 7% top line and 11% EPS, right? Through our research and the thesis that we put out, we think for the next 3 or 4 years, top line is going to grow 8% and EPS is going to grow 15%.

Historically, this is a company that has traded at 22, 23, 24 times P/E. It's now trading at 15 times. Our thesis was not—we don't expect any multiple rerating, because we don't know if this overhang from the AI narrative is going to persist. But we do think the fair P/E of this business is around 22 times, and that's actually what comes out if you run a DCF. That's a fair P/E.

At 15 times earnings, you're getting 15% growth in EPS; you're getting—I think it's probably close to a 3% dividend yield at this price. And I think just holding this, without any multiple expansion, you're going to get close to a 20% return over the next 3 to 4 years.

You know, it's an interesting one because obviously I think some of the sell-off is driven by the AI fears that you guys have hopefully dispelled throughout this podcast. But one interesting thing here is—and you guys start to allude to this in the presentation—if you really believe this is extraordinarily difficult to rip out, kind of non-AI-able, you would view them as a massive AI beneficiary, right? Because AI is going to, at minimum, really improve their efficiency. You would hope, right? You had 1,000 engineers before. Well, you can probably do it with 500 engineers now, or you can keep 1,000 and just be much more productive.

So you would think it's going to improve margins. And I would argue AI might make—I mean, travel's always a GDP-plus business because just as people get richer, they travel more. I'd argue if you see a productivity boom, you might even see more travel. It might be GDP-plus. So I'd argue both in terms of a tailwind to the growth rate and margins expanding from AI benefit, I wouldn't be surprised if these guys are ultimately an AI beneficiary. I got a little bit galaxy-brain. Anyone can talk to that if you want, and then we can probably wrap this up.

Yeah, I can take that one. I think we spoke to the SVP of engineering at Amadeus, like I mentioned, and in terms of AI, we are seeing an increase in the productivity of engineers. What we learned was that today they have more engineers at Amadeus than Microsoft Office and Outlook combined, which requires many more updates.

Then, to dive deeper into this, we dug deeper into how other similar companies compare in this space in terms of revenue per full-time employee. Sabre, which is the closest competitor, has higher productivity by 30%, which shows room for growth.

So, in general, in our conversations around the productivity boosts with experts in Gemini and Anthropic, we learned that they expect about a 1.5x to 10x productivity boost. Even if we take the lower end of that scale, which is, I think, 1.5x, we see that earnings will grow by 10%. So, in addition to what you mentioned about the GDP growth, there's also the scope for increasing earnings just by the reduction in employees needed at Amadeus.

You know, my only pushback to what you said was that they have more employees than Outlook and Office combined. I am on Microsoft Outlook, and I don't know how much engineering work is going into Microsoft Outlook these days.

And then on the Sabre point, it's interesting. You say, “Hey, Sabre gets 30% more revenue per employee.” But then you say, “Sabre is overlevered.” Maybe it's not, “Hey, Amadeus can catch up in terms of margin expansion.” Maybe it's a bull case for Amadeus on the other side. It's like, look, Sabre—they have underinvested and they have lost market share because they're underinvesting. And that gap might be more of a signal of, “Hey, they continue to really underinvest in this thing and kind of run a bare-bones product that will cede share over the long run to Amadeus,” as the Sabre CEO basically admits when he says, “Hey, I need somebody to go investigate them for an antitrust violation.”

Guys, this has been great. I think we've covered most of it. I think this is a fascinating one, and I know I've got a lot of people on the quality-compounder side who have been saying what you guys are saying. This has been unfairly sold off, but were there any last thoughts you wanted to share before we wrap this up?

No, I'll just say what a great experience it was researching it. We had help from many people who kindly answered the phone when we called them, including the head of IR, Christina, who was very nice to talk to. So, overall, a very nice project, an enjoyable experience, and I'm happy to have done it with these two.

Well, look, you guys did a great job. Congrats again. I know a couple of people who are really pushing for this. I think it's great work and a fascinating company, so I appreciate it. But guys, thanks so much, and we'll talk soon.

Thank you.

Thanks, Andrew.