Results:

Guidance & Valuation:

Balance Sheet:

Results:

Guidance & Valuation:

Balance Sheet:

As a reminder, this company ironically flags emails as spam when you use their name too many times. That’s why I will use “mega-cap” and “Search King” a lot throughout this piece. 

The theme of this event was the firm driving rapid product innovation across every part of its full-stack AI machine… from models, to infrastructure to apps. That will be the focus of this piece.

AI Infrastructure Foundation Facilitating Leading Model Strength:

The Search Giant has broad product offerings for every facet of the AI opportunity, and the foundational hardware arm may be the key enabler of it all. It’s how Alphabet builds what it views as the most efficient data centers on the planet. It’s how the company can create purpose-built compute for specific needs in a less general fashion than GPUs can. It’s highly important.

Without this ironclad foundation, the mega-cap would not be leading the LMArena rankings in every single major benchmark… from reasoning to teaching to math and coding. It would not be doing that while offering world-class performance and output token efficiency. Without this same foundation, it would not have the three fastest models on the planet and would not be driving such rapid upward progress in its Elo scores (popular ranking system for models and other things). Simply put, this foundation is like feet for the body. Without them… we’re not going very far. And with them? This company is able to move forward at the pace needed to keep leading this hyper-competitive field. And this isn’t just a leadership claim; it’s world-renowned, independent research organizations affirming it.

Going forward, the organization will feature a constant, brisk pace of model innovation like the brand new Gemini 2.5 Flash iteration just announced. This is faster and cheaper than the current version… which was already the fastest and most efficient workhorse model out there. I say all of this because sentiment across social media would probably make people very surprised to learn that this enterprise is the best of the best at driving model improvement. And again, that is because of innovation like its 7th generation TPU driving 10x performance gains. 

It pairs all of this model, research and infrastructure strength with 7 products distributed to 2+ billion people and all of the unique data that entails to drive better model training. Simply put, it’s in a far better spot than most investors seem to think.

“There's usually a hard trade-off between price and performance, yet over and over, we've been able to deliver the best models at the most effective price point. Not only are we leading this Pareto frontier, we've fundamentally shifted the frontier itself.”

CEO Sundar Pichai 

AI Adoption Progress:

While it’s nice to hear that Search King has great models and research teams, it’s better to hear that those unique assets are leading to impactful product introductions and meaningful traction. That’s happening.

AI token processing has 50Xed Y/Y to 480 trillion per month. Vertex AI and Google AI Studio now combine for over 7 million developers vs. 2 million less than a year ago.  

Vertex usage specifically is up 40x Y/Y. While this doesn’t dramatically accelerate revenue growth immediately, it’s always good when world-class builders are using the mega-cap’s infrastructure and creating within its footprint. A lot of this work improves its existing products, builds entirely new products to sell on its app store, makes its infrastructure even more efficient and makes software engineers more reliant on its cloud business. It’s a large positive.

Gemini now has over 400 million monthly active users. This compares to 350 million a little less than a month ago, representing great 14% M/M growth. And for users with access to its latest and greatest Gemini 2.5 models, usage gains are directly rising by 45%. Lens (its picture-based search product) now has 1.5 billion monthly active users and is growing 65% Y/Y. All of this makes it crystal clear that its aforementioned model lead can drive more durable growth across its entire business.

AI Overviews now have 1.5 billion monthly active users. Now… it’s worth noting that it’s very easy to rack up massive usage when you’re displaying this at the top of search results that billions are already requesting. Still, it is good to see customer satisfaction, engagement and ad relevance all improving with this introduction. As a vital aside, I got a ton of questions about how well this can monetize vs. traditional search. Recall that last quarter, leadership told us it was already monetizing at par with traditional search – with more room for upside. It’s also worth noting that AI Overviews change the search growth engine. Because the ads and results are far more relevant, conversion rates are much higher and total clicks are lower. So? That’s why search’s click growth was in the low single-digits this quarter while pricing growth was near 10% Y/Y. This is also part of the reason why this product is so popular with users and why volume gains are so compelling. In India and the USA specifically, these overviews are immediately raising engagement by 10%, with that benefit ramping over time.

Project Starline Practical Innovation:

Project Starline is its 3D video research arm. During the event, the company unveiled “Google Beam.” This pulls from Gemini models to turn 2D images into hyper-realistic, “near-perfect” head-tracking. It unlocks real-time 3D rendering of the person you’re chatting with. Through a partnership with HP, this will become available to some users by the end of the year.

The other exciting announcement from Starline was placing real-time language translation right into Google Meet. Customers can now have natural conversations between English and Spanish.

Project Mariner Practical Innovation:

Project Mariner is where it works on agentic web browser innovation. It’s where the tech mammoth enables users to offer goal-oriented tasks, while allowing actionable agents to perform those tasks in the best way possible across the web. This debuted last December, while the company announced multi-tasking and up to 10 simultaneous workflows this past week. It can also now store previous tasks in memory and learn plans for similar assignments in the future.

Project Astra:

Project Astra is its universal AI assistant. This is where the Gemini Live API, which fosters ultra-low latency audio and visual exchanges, comes in handy. It’s what empowers Astra to be a conversational companion that can highlight inaccurate statements in real-time. In the demo, a user was kindly corrected when calling a garbage truck a convertible and their own shadow a stalker. All in real-time and with updated voice control to make these conversations more natural and expressive. A new text-to-speech model with “native audio output” (more natural conversation) brings that to life.

World models are also now a company focus and a key enabler of bringing Project Astra to its full potential. These models have an intimate understanding of the physical and digital worlds. They know how an AI agent’s action will alter a web environment. They know how skipping a stone across a pond will lead to how many ripples and splashes. They understand physics and the cause and effect of physical actions. Pretty cool and pretty useful in areas like robotics. World models can also essentially build digital worlds and experiment with them in a low-stakes environment. This is similar to Palantir’s digital twin engine and Nvidia’s Omniverse offering.

Agentic AI Experience Personalization:

The firm is leveraging its dominant positioning in email, productivity, video streaming, maps and so much more to drive unique agentic experiences. This is where it can shine. And with a user’s stated permission, Alphabet does exactly that. They call this “personal context,” which is really just a way of saying they’re scraping not just public web data, but also our own. This means smart replies to emails can now emulate our own voice and personality. It means my inbox will eventually turn into a queryable destination for me to recall any important notes I forgot. That’s how you use world-class product distribution to effectively drive incremental AI utility.

Agent and AI Mode:

More Agentic AI is coming to search. Soon, it will debut “Agent Mode” in Gemini and core search. This will embed actionable task completion right into the search process. Separately, it’s also extending AI Overviews into a new product called AI mode. In this mode, the overview becomes the full page, with more links (to monetize) and the ability to ask follow-ups. In early testing, it’s leading to 2-3x query length vs. legacy search… and that’s before Gemini 2.5 was added to this product. That’s now happening. 

AI mode uses something called a Query Fanout technique. This matches input tokens with the appropriate model based on complexity on response or task. This means more expensive models are only used when needed, which boosts efficiency per output token. It’s similar to the “mixture of experts” approach many model builders are now using. Mixture of experts means only certain needed parameters are activated per input token based on what’s needed. Query Fanout does that at the model level, rather than at the parameter level within a specific model. 

Both of these products scream monetization potential to me. If the company can use top-of-funnel search to guide a customer all the way down to a restaurant reservation or t-shirt purchase or concert splurge or haircut appointment etc., they should (and will) be very handsomely compensated for that. That’s the idea and is why shopping use cases were such a prevalent part of this talk.

Hardware:

The company introduced Android XR as its first operating system natively built with Gemini to support XR devices. They did this in partnership with Samsung and Qualcomm, with Samsung planning to release its first headset later this year. In partnership with Gentle Monster and Warby Parker, they will eventually debut smart glasses with cameras, microphones, speakers, music, calls and translation. This will directly compete with Meta’s Quest and Ray-Ban businesses.

More Notes:

The company is adding a $19.99/month subscription for access to a bundle of its latest AI products and a $249/month tier with its highest rate limits and the earliest access to all new products.

Some Thoughts:

I don’t think the people calling this firm a dinosaur watched this event or read their latest earnings materials. To think they’re falling behind in AI is to stick your head in the sand and ignore reality. They are leading the pack.

More:

In a research note, BMO talked up search resilient for Alphabet. Their data shows rising Y/Y prevalence of using Google as a go-to search tool and falling willingness to switch.

As a reminder, February was a bad month for the travel industry. Several companies revised or removed guidance. Airbnb fell behind on its Q1 guidance, but strength through March allowed it to catch back up and beat its target.

Global Demand & Macro:

In North America, slower NEB growth was in the low-single digit range for the quarter. Excluding this region, NEB growth would have been 11% instead of 8%. More on this later in the section. In Europe, the Middle East and Africa (EMEA), NEB growth was around 5% Y/Y and ADR rose by 2% (4% FXN). In Asia-Pacific (APAC), NEB growth was around 15% and ADR growth was -1% (+3% FXN). Japan is quickly joining Brazil and Korea as underpenetrated expansion markets enjoying strong success. Domestic nights booked in that nation rose 20% Y/Y following a successful marketing campaign.

In Latin America, NEB growth was just over 20% Y/Y and ADR fell by 7% (+2% FXN). The continued to accelerate Y/Y growth in Q1 2025 vs. Q1 2024 despite lapping a great year for that geography. Brazil continues to be a demand standout for the company, as NEB growth was 27% Y/Y and kept accelerating. First-time booker growth accelerated again, as ABNB continues to prove its marketing and localization playbook can work extremely well outside of its 5 core markets. These “expansion markets” enjoyed double the Y/Y bookings growth of its business overall.

From a macro point of view, for the most part, Airbnb’s business has been relatively macro-resistant. Its geographic diversity and wide price ranges cater to a plethora of interests, and the worsening macro also has a way of giving it easier access to supply. It’s similar to Uber in that regard, although vacations are certainly more discretionary than ride-sharing and food delivery. And notably, average booking lead times have not shrunk in most of its markets over the last few months, which shows consumer anxiety not really flowing into its business… but there’s a notable exception.

Now back to North America for a moment. The main area of pronounced weakness is Canadian travel to the United States and overall international travel to the USA. They’re picking other destinations like Mexico and Japan more frequently. Fortunately for Airbnb, that’s just 2.5% of its North American business, as domestic dominates the volume pool. Still, this is slowing down overall growth a bit.

For domestic USA travel demand, long lead time bookings are also a bit challenged for the market overall; consumers are waiting for trade policy and economic clarity. As you’d expect, this has impacted the affluent U.S. traveler less than the mainstream U.S. traveler. All of this is why North America is currently its slowest growth market. It’s macro and geopolitics… not Airbnb. For evidence, it continues to take share in Canada and the United States. Airbnb sees a real opportunity to lean into lower-price listings as American consumers seek out less expensive trips such as “staycations” this summer. That’s the luxury of its massive supply base. It can cater to a wide variety of tastes.

Core Product Improvements:

Airbnb made a series of tweaks to the core product this quarter. It upgraded algorithms to improve search filtering and added dynamic tips to alert travelers when discounts are available or booking windows are closing. It further streamlined checkout to eliminate more friction and optimize shopper conversion rates. The improvements it has made to app usability also continue to work. NEB growth through the channel was roughly double overall growth at 17% Y/Y.

Its work to remove low-quality listings is also going well. It has now cut 450,000 of them vs. a couple hundred thousand as of last year. Customer service scores are rising while chargeback rates are falling. Both are strong pieces of evidence that this is a good strategy. Its guest favorites tool continues to grow in popularity and its co-hosting tool (allow hosts to connect with a Superhost) enjoyed 15% Y/Y bookings growth. Finally, it has now added travel insurance to 12 countries, with plans to expand from there. As we’ll see later in this piece, there is a sharp focus on extending beyond the core. 

There was understandably considerable focus on product expansions and newness in this report. We’ll dig into all of that, but we cannot ignore the importance of core business health. This will remain the main revenue driver for a long time and will be the cash cow that funds the rest of its product investments. It is much easier to effectively focus on impactful product launches when your existing suite is in healthy shape and not in need of so much maintenance or time.

New Products & the 2025 Summer Release Event – New Offerings:

The decision to delay this earnings review instead of other ones during the peak season was intentional. All of this news came about 12 days before its highly-anticipated summer release event. This is where it re-introduced Airbnb Experiences, more new products and planted a stake in the ground in its pursuit to expand beyond the product core.

The first announcement was the expected relaunching of Airbnb Experiences. As the investor materials tell us, this lets people “explore a city with a local who knows it best.” It distances travelers from the typical, cookie-cutter tourist attractions and city tours by letting them plug into local experts who can show them their cities and culture from a unique point of view. Experiences out of the gate will include museum and “cultural” experiences, food tours, cooking classes, outdoor adventures, art tours, private workouts and wellness sessions. It will also include “extraordinary experiences” called “ Airbnb Originals. This seems like a re-branding of its “Icons” product, as it similarly matches guests with world-class celebrity chefs and more options like learning to throw a spiral with Patrick Mahomes. Non-Airbnb customers are also welcome to use this product, as leadership thinks these activities will be interesting for locals too. Experiences are constantly vetted and re-evaluated, with lower-quality options habitually removed.

Next, the firm is creating a new “Airbnb Services” offering to book add-ons to travel itineraries. As the letter stated, “people often choose hotels for the services they offer.” This is trying to emulate that reality, while offering Airbnb a potentially promising growth lever. Categories include private chefs, photography sessions, massages, personal trainers, hair and beauty appointments. Hosts will have an average of a decade of experience, with guests getting access to easily book on “detailed listing pages.”

Airbnb has always come with a risk of guests and hosts supplanting the marketplace and booking directly. That risk has been rather modest, as doing this eliminates all travel protections for the guest. They are out of luck if that host decided to double-book, misrepresent a listing, cancel last-minute or anything else. When you’re traveling with your family across the globe, I don’t think that’s a risk most are willing to take for a few hundred dollars on top of an already expensive vacation. I think disintermediating Airbnb within the new experiences offering will also be tough, as most of these offerings are unique and special. For Airbnb Services, however, I don’t see why Airbnb would get much repeat business. If you’re in a brand new city and just don’t know who gives a good haircut, I get using it. The second time you visit that city, however, why would you not just text that person directly and side-step the added fee? It’s far lower risk than doing it with Airbnb’s core offering and is probably a pattern that will be prevalent.

Airbnb’s overhauled tech stack is enabling all of these new launches. It’s how the company has grown flexible and nimble enough to make needed changes to products on the fly; it’s how it maintains the neatness of its user interface as offerings grow. Simply put, Airbnb was not ready to sprint with its old, somewhat archaic back-end architecture. Now it is. The new app comes with dedicated tabs for experiences and services, with personalized suggestions, itinerary, upgraded messaging and a listing process upgrade to make adding new services and experiences seamless for hosts.

Supply Growth:

As the team has talked about in the past, they’re focused on adding more hotel supply to the marketplace. Especially in its popular urban areas, listings have sky-high occupancy rates and Airbnb doesn’t always have compelling availability. Hotels can be a great supplement, while also offering a more consistent experience for those preferring it. They’ve just started offering 10% Airbnb credits for HotelTonight bookings (company it bought in 2019). They plan to do more things like that to nurture cross-selling, while also simply adding more supply to raise overall GBV.

Fine quarter. I go back and forth on this name a lot. Part of me thinks the new product launches are perfect complements to its massive core business and can help durably reaccelerate the growth engine. Easier comps and a better macro backdrop for travel should help too. At the same time, its core business is gigantic. It will take a ton of traction for Experiences and Services to become material, and I’m not sure how long that will take. I also worry about being supplanted as the middle-man for several use cases within the Services business specifically.

It’s certainly not expensive, but is also just not giving investors much newness to be excited about. We’re left with a core business that prints cash, enjoys ubiquitous, verb-fueled brand awareness and keeps taking market share. We’re just also left with a steadily slowing growth engine and a need to accept some near-term margin weakness to hopefully find new growth. I like the investment. I don’t love the investment. I only pick investments that I love. If it becomes apparent that these launches can drag revenue growth up into the low-teens range in the coming quarters, I’d be happier to start a position at higher prices than I am today.

Anthropic’s Claude 4 model debuted this week to a warm reception. I’m sure Amazon and Alphabet (both large investors) are both pleased… but Amazon should be even more so. It’s assumed that a large chunk of this world-class foundational model was built with Amazon’s Trainium 4 chips. These aren’t as powerful as Nvidia’s Blackwell GPUs, but apparently they’re good enough in some places for a model ranked near the top in key areas like coding and reasoning. That’s highly encouraging for Trainium demand and Amazon.

Amazon’s shareholder meeting was uneventful, as most of them always are. There was, however, an interesting tidbit in there. Amazon has seen no impact on consumer spending behavior from the trade war. Merchants also haven’t raised prices for the most part. Really good to hear on both accounts.

There were some rumors about Founder/CEO Georgue Kurtz rapidly liquidating his shares and relinquishing control over the company. He took to LinkedIn to call this bogus. He hasn’t had a super majority of voting shares since the IPO and decided to move “some stock into trusts for his family and causes they care about.” He’s “all in and not going anywhere.” Great to hear from one of the best CEOs out there.

CrowdStrike also hired Snowflake’s Chief Communications Officer for the same role and joined Nvidia’s AI Factory program to help GenAI and Agentic AI proliferate more securely.

Youssef Lahrech is stepping down as Nu’s COO and President. He has a phenomenal reputation, as he came to the firm after two decades of being a VP, Managing VP and Senior VP at Capital One. His resume is fantastic and his role with Nu was large. Fortunately, this seems to be an amicable split. He will be a “permanent observer” for the board’s audit and risk committees and an “expert advisor” on its credit products. Founder/CEO David Vélez will absorb his duties while (I think) they likely look for a replacement. The Nu talent bench is very deep and the culture is world-class. Losing Youssef is tough, but I’m confident they attract a great replacement.

Mercado Libre’s legendary Founder and CEO Marcos Galperin will step down and transition to Executive Chairman at the beginning of next year. The firm’s highly regarded commerce president, Ariel Szarfsztejn, will step in as the new CEO. Just like for Nu, Mercado Libre has a deep talent bench and a phenomenal culture. It’s great to see Galperin staying on in a formal capacity and I expect this transition to be a smooth one.

PayPal added OpenAI support for its model context protocol offering.

SoFi’s Galileo launched Payment Switch in partnership with Atomic. This makes it easier for financial service vendors to make it easier for customers to switch default bill pay methods. It unlocks easier access to using Venmo, PayPal and more digital wallets as methods. This also helps customers nudge consumers to their own payment options to raise market share.

Meta got a legal win in Germany this week to utilize user data in its AI training.

Shopify debuted a new AI store-builder for its merchants. They talked about this on the earnings call, but now it’s ready to enable businesses to easily, conversationally build sites. Whatever Shopify can do to “remove more headaches” from maintaining a business will be good for adoption and volume.

The Trump Administration is threatening to raise EU tariffs to 50%. They say the EU isn’t negotiating in good faith, but most other places are. Bessent also plans to continue talks with China in the near future and teased more trade deals coming soon.

Consumer & Employment Data:

Output Data: