艾普拉文

AppLovin shares are off nearly 20% after reporting second-quarter revenue and adjusted EBITDA that narrowly missed the midpoints of management’s guidance. The company blamed a slower-than-normal cadence of model improvements for the shortfall.

Why it matters: While we rarely see companies with AppLovin’s growth and profitability profile, the advertising platform is a black box, with performance driven by discrete, step-function improvements to the proprietary model rather than transparent, modelable levers. This quarter’s improvements fell short, but management states it was simply delayed until July. There has been discussion that June’s self-service consumer (nongaming) launch was slow, but management claims that spending within this vertical is 28% above the cyclically strong fourth quarter and that the overall business can compound at roughly 30% annually. We are slightly more conservative with an average annual growth rate of 26%. Despite lapping tough year-over-year comparables, we don’t think growth normalization is catastrophic for the investment case à la The Trade Desk. We like the big picture: strong free cash flow trajectory, a long tail of potential advertisers, and a solid balance sheet that enables buybacks.

The bottom line: We decrease our fair value estimate to $470 from $500, while emphasizing our Very High Uncertainty Rating due to the opaque nature of model improvements. Shares are undervalued, in our view. A major bear thesis has been neutralized now that the SEC inquiry into data collection practices has closed without any recommended action. We don’t believe AppLovin is hitting a model improvement wall, though based on our understanding of black box models, we acknowledge that saturation (too many advertisers using the same model) is a risk. Still, advertisers are onboarding and spending more money, mitigating some of this concern.

Our $470 fair value estimate implies an enterprise value equal to 19 times our 2026 sales estimate.

In our view, the primary drivers of the company’s value are the growth of digital advertising budgets, advertisers’ willingness to explore AXON 2's black-box programmatic advertising solution, the penetration of ad markets outside of gaming, and international expansion. The digital advertising market is huge, with approximately $800 billion in annual spending today. We expect the market to reach $1.5 trillion by 2034, driven by GDP growth and advertisers’ demand for programmatic, omnichannel solutions.

Since going public, AppLovin has consistently outpaced industry growth, underscoring its early market penetration. We estimate AppLovin accounts for just under 2.5% of digital advertising spending, and we expect this to increase to approximately 6% penetration by 2034.

This leads us to project sales will maintain double-digit growth annually throughout our forecast, averaging roughly 20%. We project modest operating margin expansion throughout our forecast, from just below 75% in 2026 to 77% by 2035.

We see some competing pressures on the firm's variable costs. With cloud-hosting fees (captured primarily within cost of revenue), we expect increases as AppLovin adds advertisers to the platform and demand for computing resources rises.

We project sales and marketing expenses to decline, as acquisition costs should compress with the automated customer onboarding rollout. Plus, promotions are less needed to drive game downloads after the gaming business was divested.

We project R&D costs to remain flat as a percentage of sales. We project some R&D costs to increase due to ongoing enhancements to the highly complex Axon 2, but this should be offset by AppLovin's decision to stop developing original gaming titles.

We assign AppLovin a Morningstar Economic Moat Rating of narrow, based on intangible assets.

AppLovin’s most valuable asset is Axon 2, a complex “black-box” advertising optimizer. Advertisers on the AppLovin platform specify a budget constraint (what they are willing to pay to acquire users), their desired geographic regions, and their desired return on advertising spending. Axon 2 takes these inputs and deploys advertisements to users most likely to download or engage with the advertised product or application. Effectively, the black box has three inputs and then does the rest of the work. “The rest of the work” is identifying the most susceptible users so AppLovin can collect performance-based revenue commissions. This opaqueness differs sharply from The Trade Desk’s strategy of providing the advertiser with granular control over programmatic decision-making, but it also automates a time-consuming, manual process of modifying deployment strategies, which can be attractive to advertisers.

According to Northbeam, AppLovin is driving a 45% higher return on ad spending than Meta and 115% higher compared with secondary advertising platforms like TikTok, Pinterest, Snapchat, and YouTube. While we expect some return compression as more advertisers join the platform and compete for similar inventory, the recent outperformance is remarkable in magnitude. Additionally, Northbeam notes that AppLovin is extremely good at customer acquisition. For example, 85% of the purchases through AppLovin ads are from first-time buyers, which is a great justification for advertisers to join AppLovin’s DSP. The complexity of training Axon 2 increases the durability of the firm’s competitive advantage. Specifically, Axon 2 is trained on 13 years of first-party user behavior data from the gaming division. The first-party training data has been further reinforced with strategic acquisitions of exchanges and data attribution firms, which provide advertising flow data and post-user acquisition behavior information, respectively. With outperformance in return on ad spending and training complexity, AppLovin’s proprietary Axon 2 has created a durable intangible asset that drives meaningful results.

AppLovin sharply increased research and development spending to train Axon 2 prior to the 2023 launch, and AppLovin has continued this investment to maintain its competitive advantage and ensure that Axon 2 remains a moving target for competitors. AppLovin employs a large number of PhD researchers in its machine-learning division who have been working on first-party data and distillation modeling techniques for several years. While tech behemoths like Google and Meta are also leveraging AI distillation techniques and have more researchers, their focus is primarily on optimizing ads within their ecosystems, while AppLovin is aiming for the broader internet, including connected television streaming applications. Prioritizing one’s own advertising inventory, as Google and Meta do, places a "silo" constraint on the efficiency of a system that works best without constraints. We believe that AppLovin’s focus on the broader internet and CTV will enhance the durability of Axon 2’s competitive advantage against firms with the resources to displace it.

Currently, the only bottleneck for more growth is the manual onboarding process of large advertisers outside of the traditional gaming contingent. AppLovin claims it is adding nearly 100 e-commerce advertisers a month, and that demand has outpaced supply (employees dedicated to onboarding). As AppLovin onboards additional users, more bids will be placed across the digital ecosystem, giving Axon 2 more win/loss data to refine its optimization. These incremental improvements can create a positive feedback loop: as performance improves, more advertisers are drawn to the platform, increasing training data volume, improving conversion rates, and attracting even more advertisers, potentially compounding Axon 2’s effectiveness. We don't see a network effect as an underlying moat source, however, because performance could degrade as more clients join the platform. As more advertisers join the platform and compete for the same inventory, the mispriced supply that Axon identifies could be competed away.

We believe switching costs could emerge as advertisers dedicate more of their budgets to AppLovin’s black-box approach. We also like the firm’s differentiation from its main competitor, TTD, because it offers a vastly different value proposition: AppLovin is hands-off automation, while TTD provides high-touch advertiser control. That said, we believe switching costs will be more subdued for low-touch AppLovin than for a high-touch platform like TTD, since TTD encourages marketers to spend weeks refining their targeting algorithms on the TTD platform.

Overall, we acknowledge that much of AppLovin’s future hinges on securing buy-in beyond gaming inventory and gaming advertisers, but we believe that Axon 2 is delivering tangible results for diverse advertisers and warrants a narrow moat rating for the company.

Axon 2 is an innovative and complex ad optimizer that has driven robust growth within AppLovin's profitable ad tech business segment.

Acquisitions have incorporated connected television ad inventory and mediation services, which greatly increase the total addressable market and remove ad exchange middlemen.

AppLovin can benefit from a positive feedback loop in which more advertisers joining the platform results in more data, which improves the optimizer, delivering better results, and attracting more advertisers.

Axon 2 may face a saturation threshold, where performance deteriorates as more advertisers compete for the same inventory.

The success of AppLovin’s expansion into additional advertising inventory types, like connected television and the open internet, remains highly uncertain with potential shortfalls relative to market expectations.

AppLovin has been subject to multiple short-seller reports that allege inflated attribution and conversion metrics. More short reports, if released, will likely weigh on the stock.