(Access a of this post PDF here)
3Q26 Update.
Airbnb reported stellar 3Q26 earnings, and the stock was up +12% the following day, putting total YTD gains over +40%.
Revenue reaccelerated to +19% y/y (+16% ex-FX) to $4.9bn. GBV reported its fastest growth since 3Q22, growing +21% y/y to $27.7bn. As a result, they increased their full year guidance to high teens growth, up from their low-double digit growth expectations at the beginning of the year.
Nights and Seats also grew +16% y/y to 155mn with Reserve Now Pay Later (RNPL) contributing an uplift of +3%. On the call, Brian Chesky noted that they are seeing more guests return to Airbnb and more people trying it for the first time. We see this with first-time booker growth accelerating to +13% y/y, which is the fastest it has been since early 2022. In addition, the fastest growing segment of nights booked is Gen Z. More importantly, nights booked on their app grew +29% y/y and now represents 66% of their total nights booked. While they have always had high direct traffic (>90%), mobile app usage is a better user connection because it doesn’t put a browser between them and allows for more customization with always logged-in accounts. It also suggests more of a long-term “commitment” to the service.
In every region, they witness accelerated growth in all five of their core markets (U.S., France, U.K., Canada, and Australia) as well as all of their expansion markets. In North America, nights booked grew at its fastest rate in over 4 years and revenues accelerated to +20% y/y. One of the reasons North American revenues accelerated in the quarter was they had the continued benefit of the World Cup. Across the expansion markets, nights booked reached their fastest growth rate in nearly three years. In particular, nights booked in Brazil increased more than +30% y/y, while India grew more than +60% y/y, showcasing growth is broad based.
On the call, Brian Chesky cites their app rebuild for this growth acceleration, as well as improving the core experience, launching new services, and introducing hotels. Now with the help of AI, they are shipping products and features faster. In 3Q alone, Airbnb released more than 50 products and features, which is nearly 70% more than the prior year. Among these new products and features are Neighborhood Pages, which provide information about local restaurants, activities, and attractions, helping guests evaluate a destination before selecting a property. Following encouraging results in their beta rollout in Paris, they plan on expanding that feature out to more than 1,500 neighborhoods globally by the end of the year.
In addition, Airbnb has also introduced social travel features that allow guests to view where friends and family have stayed and access their recommendations. On the call Brian Chesky reminisced about the early days of Facebook when it was a true social network with a social graph and friends broadly sharing. Instead he notes that social networks turned into social media platforms and this usage largely disappeared. They haven’t talked about it much, but with the push to create profiles and approved identities, they seem to also be positioning themselves to create a sort of social network. For now, the idea is namely to focus on where friends have traveled and recommendations from their trip, but hearing him speak about it, it is clear they have broader ambitions here. Adding a social element to travel recommendations can increase trust and thus reduce booking friction.
For Airbnb, it is clear why they would want this activity on their platform. It creates more differentiation versus other OTAs, makes them harder to disrupt, can increase engagement, and gives users more reasons to open their app—which in turn could provide Airbnb more opportunities to cross-sell (your friend stayed at this Airbnb, and then also went to this Experience in Paris, want to try it too? They said they loved it!).
We can see how staying at an Airbnb with your friends (and they typically get added to the stay) could allow them to back into a real social friend network. This mirrors their previously mentioned ambitions of creating an “offline” social network that helps people meet in the real world. It is still early innings, but Chesky calls this “one of the single most underestimated parts of our business.” Much like their services business, we think it makes sense in theory as it provides users something we believe they could want, but the app’s low usage frequency is likely to also be an issue to building this social networking activity. It is worth mentioning that the reason social networks turned into social media platforms is because it turned out professional creators created better content than your friends. Instagram’s “close friends” feature, Snapchat, and private groups on messaging apps have replaced a lot of this “private social networking” use case. Success here though could be a game changer, but we view it as a hard to execute on call option. Their efforts to branch beyond core accommodations have largely fallen flat so far.
Optimizing listing pricing remains one of Airbnb’s larger opportunities. Airbnb recently began introducing dynamic pricing capabilities that use AI to identify price elasticity across individual listings. The idea being that a slight price increase may be more than offset by increased bookings, driving overall revenue for the host and platform higher.
Their new Reserve Now, Pay Later (RNPL) offering has already demonstrated a measurable impact. On the call, they noted that more than 40% of GBV became eligible for RNPL and this drove 3% incremental nights booked growth. As RNPL reduces the commitment to booking, it increases lead times and often leads to bookings of higher-priced properties. However, with the incremental RNPL growth has come higher cancellation rates. Cancellation rates increased 200-300bps y/y from 16% to 19%. CFO Ellie Mertz on the call said this 19% cancellation rate is still healthy relative to the broader travel industry.
Airbnb’s implied take-rate declined slightly to 17.6% in 3Q. This was primarily attributable to RNPL, which increases GBV before the associated revenue is recognized, alongside higher incentives and credits for new initiatives. The impact was partially offset by travel insurance and the company’s migration to a simplified fee structure, in which all hosts now charge a 15.5% service fee. As a reminder, Airbnb used to charge hosts 3% and guests the rest of the fee (which varied). This arrangement made it hard for a host to price properties at parity across platforms. This pricing change has greatly simplified pricing issues across platforms. Over time, the combination of simplified fees and insurance products should provide a modest contribution to take-rate expansion.
Turning to their expansion beyond alternative accommodations, hotels are emerging as an attractive opportunity for them. Airbnb now offers thousands of boutique and independent hotels across more than 60 major destinations, with bookings accelerating again in 3Q. Brian Chesky indicated that hotel bookings are growing substantially faster than the company’s core homes business, although the segment remains relatively small. The opportunity is significant given the estimated $900bn global hotel market. Airbnb already attracts traffic from travelers searching for accommodations and by introducing hotel inventory, they can convert visitors who might otherwise leave the platform to book elsewhere. Interestingly, they revealed that ~35% of guests whose first booking on the platform is a hotel return to book a “traditional” Airbnb within a year. This suggests hotels can serve as both an incremental revenue stream and a customer acquisition channel for the core business.
Airbnb has also been expanding its services offering. They launched car rentals in the quarter, and it has become one of the company’s fastest-growing services. Additional offerings they have launched include meal delivery, laundry, baby equipment, ski equipment, and boat rentals. Essentially, they are trying to leverage their user base to expand out into a variety of different service verticals. While it is clear why Airbnb wants this business to work—and we believe the need for a unified provider of local services exists—it is less clear if Airbnb is in a good position to become that provider.
They have also started to introduce partnerships, which is something they have stayed away from for most of their history. Since their founding, Airbnb has largely grown organically. Now, they are partnering with airlines like Delta to reach travelers through new channels. Most notably, they launched an exclusive partnership with American Express, where Amex travel now has 1 million Airbnb listings through their travel portal, which in turn gives Airbnb more direct access to affluent travelers. As a result, the introduction of partnerships gives Airbnb another distribution channel and early results are positive.
Moving on to capital allocation, in the quarter, they repurchased $700mn of stock, with $2.7bn remaining in their share repurchase program. They also announced a new $12bn share repurchase program.
But despite all of this progress, there still remains one key risk. And that is AI.
AI: Opportunity or Key Risk?
AI represents both an opportunity and a potential competitive threat to Airbnb. With the help of AI they are able to roll out products and features at a faster rate than they could have done in the past. As we mentioned earlier, they released more than 50 new products and features in 3Q alone. One of the products they launched is AI search for guests, where guests can describe what they’re looking for and find the right home, service, or experience for the trip. For hosts, AI can help hosts create 3D floor plans and optimize their home’s price, while also telling them how to improve their listing’s performance.
One of the clear and immediate benefits they cite on the call is their AI assistant. Their AI assistant is now available in more than 50 languages and is able to resolve 40% of customer support issues without the need of human intervention. As a result, customer support cost per booking declined more than 17% y/y.
Now Airbnb does still have potential disintermediation risk here. After Meta rolled out their AI agent, Muse, we now had a real world example of how consumers might broadly use an AI Agent (even if it doesn’t end up being Muse). It turned out, one of the big consumer use cases for AI agents was travel. So, if in the future most travel journeys start with an AI agent, then the concern becomes that platforms like Airbnb become disintermediated. Instead of customers going directly to Airbnb, they ask their AI agent for help finding a place to stay during travel. This could result in leakage of the business that Airbnb used to always win as the AI agent considers listings from Booking, Vrbo, and other platforms that the user may never have considered before. Airbnb’s main defense against this is that most of their users who book on Airbnb think of an “Airbnb” as a distinct sort of trip and go directly to the site to book those stays. The second risk is that even if the agent finds a home on Airbnb for the guest, it might price check it across platforms and perhaps a direct listing too, and then book on the cheapest option.
On the host side, there is also a risk that AI makes it much easier to cross-post listings on multiple platforms (something that most Airbnb hosts avoid doing because it becomes more complicated running a calendar and other platforms aren’t as intuitive as Airbnb). However, if AI does all of this for the host, then a lot of Airbnb’s proprietary listings could show up elsewhere and no longer will it be the case that 60% of listings are proprietary. At the extreme, the AI may also create a website for the host’s property with an MCP, so there is a cheaper direct booking option.
Today though the Agentic AI travel experience is very lacking. The default view of booking a stay is a list of links with no pictures, maps, or reviews. The user has to trust that the 5 options they got were the best 5 to pick from. While a user can prompt the AI for more information, pictures, and maps, it is still generally a worse user experience, especially because it is so slow today.
Brian Chesky believes that agents like Muse and Instinct are better at simple routine tasks but fall short on planning a trip. This is in part because a lot of people like the process of planning a trip. For a lot of travelers, it is fun to look at a lot of pictures, browse hotel options, and “day dream” about their trip. Chesky quotes that “80% of travelers say they enjoy spending time researching trips”.
He later goes on to note that “Most of our traffic happens during 9 to 5 hours. These people are supposed to be working. And they’re not. We have evidence. People are actually secretly using Airbnb. Maybe they’re trying to skip their job. But the idea is that it’s highly aspirational. So I don’t think a lot of people want to offload all that to an agent”.
However, the question isn’t whether or not people enjoy researching trips, but where are they going to make that booking? The UI on Airbnb is far superior to any OTA and each listing is rich with photos, descriptions, and reviews. This gives travelers much reason to do their browsing on Airbnb, but what if the AI agent experience improves? Could an AI agent in the future spin up its own mini app for you that includes photos, maps, descriptions, and recommendations all tailored to you? And could it be based off of not just some limited Airbnb browsing history, but a much broader set of your data? And since it is a general AI agent it could not only pull data from Airbnb, but also a slew of other sites. Then it could price check each listing. While, admittedly, this all seems a bit “sci-fi” today, that is where we could be heading.
However, we think it always makes sense to go back to basics and ask what problem is this solving for the customer? For travel with easy to list preferences—like close to an airport or meeting spot, has a gym, under $300 a night—it is easy to see how an AI agent can do well with this. But this isn’t most travel. Most of the time people are picking places based on “vibes” and this is highly personal and ever-changing. Pull up a search for Aspen ski homes and the guest may gravitate toward one design and aesthetic over another. Perhaps they thought they wanted to be close to the slopes, but they loved the design of a further away place and so picked that one. Or maybe one had a sauna, and they didn’t even know they cared about that, but when they saw it they knew they had to stay there. What essentially is happening as a user browses is that they are learning their preferences. Browsing listings is what allows them to understand what they want in the first place. This is why AI giving you a list of options misses for most people. Even if they were the “best” options, you actually don’t know that until you understand what it is you want—which you only understand by learning what you don’t want. This is inherent in the way people understand how to judge and value things. Without getting too philosophical, it is worth pointing out that there is no such thing as an “absolutely best” product or service. Instead, what we do is we compare a set of products and services to see what is the best. Apple’s iPod only became clearly the best by comparing it to other MP3 players like the Zune. Without the process of comparison, we can’t actually answer if something is the best. Now AI may be able to do this comparing, but when we get an output, we don’t actually know if it is the best.
So, AI works when you have clear parameters (lowest price, closest to this destination, has a gym), because whether something is the best can be easily measured and compared. For listings like Airbnb’s a lot of what a customer is picking is more qualitative in nature with trade-offs they themselves didn’t know they were willing to make. This is why we believe the need to browse will always be a part of the travel booking process for most leisure/ fun travel. (Plus there is the aspect that people enjoy browsing—that was Chesky’s point about usage during the work day.) Then the question becomes where is that browsing experience the best—and it seems like Airbnb is in a decent position to continue to answer that question affirmatively that it is on their platform (for travel where you want a private home or apartment).
There is some concern that their high direct traffic is what puts them at greater downside risk for disintermediation, but the flip side of that is that it speaks to how when users think of renting a home, they think “Airbnb”. Overall, there still is AI disintermediation risk, but we believe it is overstated. Having said that, even if Airbnb only loses say 1 in 10 transactions to an AI Agent telling a user to also check out these similar properties on Vrbo or Booking, that could still be a headwind. They’ve also eschewed a loyalty program for years, which further eliminates any incentive to book on Airbnb over Vrbo for instance (which could change soon though).
For Airbnb’s part, what they are trying to do is provide that AI Agent, which will roll out in the first half of next year. Instead of it being yet another chatbot, they are trying to create something that better mimics how people actually browse travel with maps, pictures, comparisons, calendars, and itineraries to help users discover a destination.
Interestingly, Airbnb does not intend to limit its AI agent exclusively to its Airbnb inventory. They plan to integrate other travel providers through APIs and MCPs to eventually allow customers to coordinate additional travel services within the Airbnb platform. This opens up a lot of questions as to whether they will prioritize a booking on their platform versus a 3rd party and how they will monetize 3rd party bookings, but we think it is a smart idea. If you want to create a user’s “go to” travel agent, it can’t be closed to just Airbnb. Just like Booking, they want to be the platform where a user plans and books their entire trip (from flight to experiences).
Brian Chesky believes Airbnb has three advantages in developing an AI travel agent. The first is that Airbnb has the experience and understanding of how guests select properties and how hosts operate. The second is that the company controls the reservation process, including payments, and customer support. Lastly, Airbnb can build an interface specifically designed for travel rather than adapting a general-purpose chatbot to a more complex use case. On the other hand though, perhaps Airbnb just teaches other AI agents what a good travel experience looks like and they merely copy that.
This is all to say that we still believe Agentic AI is a risk to Airbnb, but think they are generally insulated in a way that other OTAs are not because of their better UI, high trust, support, and the fact that most users still think “Airbnb” for private home/ apartment trips.
(Check out our recent podcast with Mostly Borrowed Ideas for much more of this).
Overall, Airbnb, as Brian Chesky says, is in growth mode. They are experiencing growth that they haven’t seen since coming out of Covid. They have a ton of call options between Experiences, Services, a social network, and becoming the go to agentic travel layer. While these are all exciting initiatives, Airbnb also has a history of doing too much at the cost of execution. During Covid the key lesson Chesky said he learned was they were spread too thin and needed to focus more. While it seems that they have reverted back to spreading their focuses, investors have little to complain about with their core business now doing very well (versus the high single digit growth they struggled to move beyond prior to the past couple quarters). This gives them much more leeway to continue to pursue all of these new opportunities in our opinion. However, how much success they need to have with each in order for an investor to make a return ultimately comes down to what is priced in.
Valuation.
Airbnb currently trades at a $191 stock price with a market cap of $114bn and an enterprise value of $104bn with a large $12.6bn cash position.
Looking at them on a free cash flow basis, we need to adjust out their net interest income of ~$484mn to avoid putting a multiple on this low-quality stream of earnings (interest on cash). On a trailing EV/FCF basis, they trade at 39x.
To get a better sense of how much growth an investor needs, we turn to our Reverse DCF.
We then sensitize across different margins as shown below.
At a $191 stock price an investor needs 10-12% growth for the next 5 years to get a historical market-like return of ~10%. If any of their call options come to fruition, growth could be much stronger. On the flip side, if Agentic AI does become even a moderate headwind to growth, achieving these growth rates could become more of a challenge. Ultimately, it is up to an investor to decide what they are comfortable assuming.
*At the time of this writing, one or more contributors to this report may have a position in stocks mentioned. Furthermore, accounts one or more contributors advise on may also have a postion in stocks mentioned.