This is the exact type of opportunity us value investors salivate over. A great business with a broken stock. When looking for new opportunities, this is primarily what I’m looking for: a growing business that continues to operate profitably while the market loses its mind over something I deem temporary or even irrelevant.

Enter AppLovin

Imagine, for a second, you’re playing a Solitaire mobile app on your phone as you kill some time on a weekend. You finish a hand, but before the next one is dealt, a 30-second ad pops up. The ad could be anything from a clip from another game to a mattress company to a meal kit delivery service. That ad didn’t come free. An advertiser paid the mobile game’s owner for that ad slot.

AppLovin is the business that matches advertisers who want to pay to have their ads seen, with publishers (mobile game developers) who want to sell ad slots to monetize their games, while AppLovin takes a fee for brokering the transaction.

Once Adam figured he was onto something with this recommendation engine, he needed to raise money. He ended up turning to angel investors, who invested about $25 million.

For six years, AppLovin operated without a board. I’ve spoken with enough investors and board members to know that a board mostly exists to talk someone out of their best and worst ideas, in roughly equal measure. We saw this on full display with Foroughi and AppLovin:

I want to revisit the dynamic that AppLovin plays between advertisers and publishers to make it as simple as possible to understand.

On the advertising side, imagine you run marketing at Wayfair, the business that sells a medley of furniture. You are tasked with one job: turn ad spend into revenue dollars for Wayfair. The number you are laser-focused on is return on ad spend, or ROAS. You are tasked with achieving a ROAS of 500%, meaning that for every $100 of ad spend, you need $500 back in cold, hard sales.

But there are a few problems with this structure. First is wasted time. As the request trickles down to the next advertiser, the slot sits empty. And empty slots pay nothing. Second, money could be left on the table. If the company that traditionally pays the most wants the ad slot, they can get it, but perhaps a lower-ranked advertiser was willing to pay more. In that case, the publisher isn’t optimizing for the highest price.

This is where AppLovin’s MAX product comes in. Instead of using this waterfall, the process becomes an auction. If an ad slot opens, MAX will ask to allow all advertisers to bid on it. MAX will then match the publisher with the highest bidder, resolving the problems with the waterfall.

And the process works! Tripledot, a gaming studio, ended up switching to MAX and A/B testing it across their entire portfolio of games. MAX raised average revenue per daily user by 20%. A 20% revenue gain from simply picking MAX is a pretty nice growth lever.

Thanks to the strong results it delivers on the supply & demand side of programmatic digital advertising, the stock has risen meteorically post IPO, with shares compounding at 34% annually going back to 2021. Even better results were obtainable for those who snapped up shares at the 2023 low point, delivering a 30-bagger in just a few years’ time.

In theory, this means AppLovin can take a larger share of the profits without squeezing anyone. My best guess is that, on average, they’re taking about 40% of the difference between what an advertiser pays and what a publisher receives. But if the algorithm keeps improving, the gap between those numbers could widen even further.

For quite a few years, AppLovin owned a bunch of game studios. But given AppLovin’s business model, that seems kind of strange, doesn’t it? After all, game studios require a lot of money to run. It would be like Uber buying a bunch of cars to facilitate ride-hailing, rather than relying on contractors to bring their own vehicles.

The real reason for owning gaming studios was to embrace their customers’ perspective on the supply side, receiving feedback and data that could be used to improve the core business.

By 2025, the algorithm was good enough on its own — AppLovin no longer needed the game studios, so they parted ways with them. They received $800 million in cash and stock, keeping a 20% stake in Tripledot to ensure the MAX relationship stayed intact. So the strategy here was to buy the data source, extract value, then sell the shell and keep the data. Kind of similar to what Foroughi did with the matching algorithm.

I haven’t spent much time looking at AppLovin’s numbers yet because I wanted to make sure you undestood the business model. The numbers, though, speak for themselves:

Their balance sheet looks quite strong with only $400 million of net debt, and maturities stretching to 2054. With a few months of cash flow, they could extinguish the net debt entirely.

There are three things that did it:

I’d say the last two are the primary culprits. Wall Street tends to like businesses with accelerating growth rates. And even though AppLovin’s growth rates remain exceptional, I think there are real questions about how quickly these numbers may decline in the coming years. Put yourself in investors’ shoes for a second. You paid a growth multiple. Now growth is decelerating, margins are compressing for the first time ever, and volume has turned negative. I know I’d get a little nervous too.

The above may concern you, and I wouldn’t blame you, but we need to evaluate whether this company can continue growing. The days of growing revenue by 60% or more seem to be over, but that doesn’t mean this company can’t still grow and make cash. I see three primary growth drivers.

The first is e-commerce. E-commerce has grown for them, but there’s a small issue as it’s not the best natural fit. AppLovin generally shows 30-second ads inside mobile games. But most smaller e-commerce businesses tend to build static images and product catalogues for Google or Meta. If AppLovin can bridge the gap, it will do very well, but it’s not yet clear whether they can.

The second is Gist, a social media app they built from scratch. With Gist, they are copying the gaming studio playbook and trying to gather data on where users spend time to then use for the flywheel. But I’m not expecting this to be the next Instagram or TikTok, so who knows how much of a value-add this will be. Lastly, it is expanding beyond games into non-gaming apps, the open web, or connected TV. This doesn’t appear to be a priority right now, but there is upside optionality here.