PINS - NYSE

Good day, ladies and gentlemen. Thank you for joining today's Pinterest Third Quarter 2025 Earnings Conference Call. My name is Tia, and I will be your moderator for today's call. [Operator Instructions] I would now like to pass the call over to your host, Andrew Somberg, Vice President of Investor Relations and Treasury. Please proceed.

Good afternoon, and thank you for joining us. Welcome to Pinterest's earnings call for the third quarter ended September 30, 2025. My name is Andrew Somberg, and I'm Vice President of Investor Relations and Treasury for Pinterest. Joining me on today's call are Bill Ready, Pinterest's CEO; and Julia Donnelly, our CFO. This conference call is being webcast, and we are also providing a slide presentation to accompany our commentary. Please refer to our Investor Relations website at investor.pinterest.com to find today's presentation, webcast and earnings press release. Some of the statements that we make today regarding our performance, operations and outlook, may be considered forward-looking and such statements involve a number of assumptions, risks and uncertainties that could cause actual results to differ materially. In addition, our results, trends and outlook for Q4 2025 and beyond are preliminary and are not an assurance of future performance. We are making these forward-looking statements based on information available to us as of today, and we expressly disclaim any duty or obligation to update them later unless required by law. For more information about assumptions, risks, uncertainties and other factors that could affect our results, please refer to our most recent Form 10-Q and Form 10-K, each filed with the SEC and available on our Investor Relations website. During this call, we will present both GAAP and non-GAAP financial measures. A reconciliation of non-GAAP measures to the most directly comparable GAAP measures is included in today's earnings press release and presentation, which are distributed and available to the public through our Investor Relations website. Lastly, all growth rates discussed in today's prepared remarks should be considered year-over-year unless otherwise specified. And I'll now turn the call over to Bill.

Thanks, Andrew. Good afternoon, and thank you for joining our third quarter 2025 earnings call. Q3 marks another quarter of strong execution against our multiyear strategy and long-term financial targets. Over the past few years, we've transformed Pinterest from a platform of window shopping, where users often found that all the stores were closed into an AI-powered visual-first shopping assistant. We are digitally replicating the joyful experiences of walking the bazaar or working with a great salesperson on your favorite boutique while seamlessly enabling our users to take action. In an evolving competitive environment, Pinterest continues to distinguish itself as a destination for our users and a vital partner for our advertisers. To illustrate this point, nearly 85% of our users come directly to our mobile app, meaning we're not reliant on search engines or other third parties for traffic. We reached 600 million monthly active users in Q3, marking our ninth straight quarter of record high users with particular strength in Gen

Thanks, Bill, and good afternoon, everyone. Today, I'll be discussing our third quarter 2025 financial results and provide an update on our preliminary fourth quarter 2025 outlook. All financial metrics, except for revenue, will be discussed in non-GAAP terms unless otherwise specified, and all comparisons will be discussed on a year-over-year basis unless otherwise noted. Now let's start with our third quarter results. We ended the quarter with 600 million global monthly active users, or MAUs, growing 12%, our ninth consecutive quarter of record high users. We continue to demonstrate user growth across all of our geographic regions. In Q3, our U.S. and Canada region had 103 million MAUs growing 4%, our Europe region had 150 million MAUs growing 8%, and then the Rest of World markets, we had 347 million MAUs growing 16%. Shifting to revenue. In Q3, our global revenue was $1.049 billion, up 17% on a reported basis and 16% on a constant currency basis. We saw strength across our conversion and awareness objectives. Across verticals, we continue to see strength led by retail as well as by smaller, faster-growing categories on our platform, including telecom and entertainment. We also continue to see a normalization within CPG, driven largely by our food and beverage subvertical. Turning to our geographical breakouts for Q3. In the U.S. and Canada, we generated $786 million in revenue, growing 9%. Strength came from retail, CPG, Telecom and entertainment. In Europe, revenue was $193 million, growing 41% on a reported basis or 34% on a constant currency basis. Strength in Europe was driven by retail. Revenue from Rest of World was $70 million, growing 66% on a reported basis or 65% on a constant currency basis. We're pleased to deliver the strong 17% third quarter revenue growth which exemplifies our multiple ways to win that I've spoken about for many quarters. We continue to diversify our business across geographies, grow long-standing as well as new advertiser verticals and begin to deepen our share with mid-market and smaller advertisers. We did face pockets of moderating ad spend in UCAN in Q3 as larger U.S. retailers navigate tariff-related margin pressure in the current environment. However, as Bill noted, we also saw accelerating strength across our international geographies in Q3 as we have begun to successfully export our lower funnel playbook around shopping. In Q3, overall ad impressions grew 54% while ad pricing declined 24% year-over-year. The primary driver of the continued strong growth in ad impressions and corresponding decline in ad pricing continues to be the growing mix shift from ad impressions in previously unmonetized or undermonetized international markets, which carry lower ad pricing than our more mature markets. Moving to expenses. In Q3, cost of revenue was $206 million, up 13% year-over-year and up 5% versus Q2 due to increased infrastructure spend related to our user and engagement growth. Our non-GAAP operating expense was $543 million, up 15%. The increase was due to investments in sales and marketing and R&D as we continue to invest in headcount to support our AI and other product initiatives as well as our sales force. Our revenue growth, combined with our disciplined approach to cost led to another strong quarter of adjusted EBITDA coming in at $306 million, a margin of 29%. Adjusted EBITDA margin expanded 170 basis points versus Q3 last year and helped to deliver Q3 free cash flow of $318 million. This speaks to the inherent profitability of our business and highly cash-generative nature of our model with over 90% of our adjusted EBITDA converting to free cash flow over the trailing 12 months. We ended the quarter with cash, cash equivalents and marketable securities of $2.7 billion. As a reminder, we've previously discussed the 4 pillars of our capital allocation framework, which remain unchanged. First, investing in product and technology innovation; second, balance sheet optimization; third, preserving flexibility for opportunistic and disciplined M&A; and fourth, dilution management. To that end, as part of our ongoing efforts to mitigate dilution from employee stock-based compensation, in Q3, we allocated $199 million towards share repurchases and $115 million toward net share settlement of equity awards, thus bringing fully diluted share count roughly flat year-over-year. Now I'll discuss our preliminary guidance for the fourth quarter. We expect Q4 revenue to be in the range of $1.313 billion to $1.338 billion, representing 14% to 16% growth year-over-year. Our guidance assumes the impact of foreign exchange to be approximately 1 point of tailwind based on current spot rates. Moving down the P&L. We expect Q4 2025 adjusted EBITDA to be in the range of $533 million to $558 million. We anticipate Q4 2025 non-GAAP cost of revenue to grow sequentially from Q3 2025 by high single-digits percent. Within Q4 non-GAAP operating expense, our primary area of investment will continue to be headcount growth within R&D to support our efforts in AI and other product initiatives as well as our global sales team. Our Q4 adjusted EBITDA guidance confirms that we will continue to expect adjusted EBITDA margin expansion in the second half of 2025. Consistent with our commentary on our last earnings call, the level of expansion in the second half will be lower than the more elevated expansion we delivered in the first half of 2025 as we continue to invest in revenue-driving initiatives. Overall, we are pleased with our progress in 2025 towards our long-term adjusted EBITDA margin targets and our ability to continue generating significant free cash flow. In closing, I'm proud of our team for another strong quarter as we continue to deliver for our users and advertisers. With that, I'll hand it over to Bill for some final words.

Thanks, Julia. I want to thank our teams at Pinterest, our advertising partners and all the people that come to Pinterest to find inspiration and take action. And with that, we can open the call up for questions.

[Operator Instructions] The first question comes from the line of Ron Josey with Citigroup.

Great. Bill, a bigger picture question for you, and then Julia, I had one for you. Bill, on the future of e-commerce, I would love to get your thoughts on agentic commerce, agenetic search and how everything is evolving here. And specifically Pinterest's opportunity and strategy given its evolving landscape and then clearly with the launch of Pinterest Assist? And then, Julia, I think you mentioned some pockets to growth in UCAN given tariffs. Wondering if this continued or if things have normalized since then.

Thanks for the question, Ron. So one of the things I'm most proud of when I look at our results, particularly over the last 3 years is the strength that we have had with users, 9 straight quarters of record high users, in the fact that shopping has been at the very center of the resurgence of our platform. And the core of that is that we've effectively become an AI-driven shopping assistant, as I discussed in my prepared remarks. And to go a little further on that related to your question on agentic. Pinterest is proactive. It anticipates what users love without the user having to ask. Effectively, that is the promise of agentic that AI is working for you without you having to tell it what to go do. And that's exactly what users experience on Pinterest every day and what is leading to that deepening engagement that we understand their style and taste and preferences so well that every time they open the app, they're getting great new recommendations from our AI-driven systems, and we're making those more and more helpful as evidenced by the Pinterest Assistant that we just announced. We're clearly not standing still. We're going to continue advancing that, and we're going to continue to be centered on our strength in a visual first experience. But adding voice, bringing more of the AI to the foreground, we think will take us further into that AI-driven journey for the user. And our focus, to be clear, is in guiding the user through the decision-making journey. We believe that is the most impactful part of the promise of agentic. From a buying perspective, we deliver great buying experiences for our users. For example, we have push button buying with Amazon linked accounts. That's a great experience. We have millions of users using that today where they can buy right within our platform. And if we see users saying they want the AI to push the button for them, that's not a technically complex thing for us to do, but we think actually the more differentiated thing is how we're guiding the user through that journey helping them go further down the commercial journeys every time they come back to our app. And it's also worth noting, as you've seen other AI platforms really burst onto the scene AI, open AI at 800 million users, that's branded over the last few years. But even if that has happened, we've delivered 9 consecutive quarters of record users while deepening engagement across the metrics that we want, including on search. And within that, a 44% increase in queries in our latest visual search features. So we think that is really clearly demonstrating that we've carved out a unique space for ourselves there. And where -- when you think about that broader promise of agentic, the AI working on your behalf to guide you through those journeys, I think we are at 600 million-plus monthly active users, I think we are one of the most popular places for us to go in that and doing something unique and distinct from others. And then on the cost side of that, I think there's -- it's really important to understand that just as we've been talking about, our ability to go align the AI with great monetization continues. The cost implications there it's not only about us aligning the assistant with our ability to monetize and how that is running great search results that are highly commercial in nature. We have our own proprietary and compact fit-for-purpose models that perform really well. And every pen we serve today is driven by that, and that's really already in our cost structure. And when there are things that we need to do with broader LLM capabilities, we are constantly doing side-by-side testing between both the leading off-the-shelf proprietary models and open source models. And one of the really, really interesting things that we're seeing is that we are just getting tremendous performance from open source models, specifically for Pinterest use cases on visual AI given current market rates and per token costs in early testing, we're seeing orders of magnitude reduction in costs with comparable performance using fine-tuned open-source models versus leading off-the-shelf proprietary models. So going forward, we think open source can be applied to many more of our use cases and at a fraction of the cost of the larger model providers using open source. So again, we feel really good about the value that we're bringing to the user there. Our ability to align that with monetization and our ability to control those costs and deliver that effectively.

And then, Ron, the second part of your question, I'd say overall, with respect to Q3, the quarter played out largely as we expected. In addition, we saw some of the pullback from some U.S. retailers spend from Asia-based e-commerce players in the U.S. was down year-over-year again in Q3, though relative to Q2, we did see a partial recovery there. As we think about guidance for Q4, our Q4 guidance range is 1 point lower than our guidance range was for Q3 as we see these broader trends and market uncertainty continuing with the addition of a new tariff in Q4 impacting the home furnishings category. So I think overall, we still feel really good about our mid- to high teens kind of revenue growth targets over the medium and long term and the durability of our revenue growth. There are several areas of momentum in our UCAN business that continue that you've seen over the last several quarters. So one of those has been momentum in emerging verticals, also momentum in smaller and mid-market advertisers and then some of the international opportunity that Bill touched on in his prepared remarks. To put this into perspective, some of these emerging verticals in the U.S. like financial services is nearly a $40 billion digital ad category in the U.S. We estimate that we have less than 0.5 point of market share there, and that category has been growing really nicely for us for some time. We expect this to translate into further share gains in this and other emerging verticals like travel, entertainment and telecom Likewise, smaller and mid-market advertisers today represent only 15% of our revenue as our priority has been to solve the needs of larger enterprise advertisers first. But we're seeing nice tailwinds as these smaller and mid-market advertisers adopt Performance+ campaigns, and we plan to continue to invest more into growing this segment. So while all these initiatives continue to play out over time, and we expect them to take time to play out, we're confident we have the right playbook to drive further growth, including an UCAN moving forward.

The next question comes from the line of Eric Sheridan with Goldman Sachs.

Maybe building on that answer, just Bill, can you characterize more broadly the digital ad environment that you find yourself operating in representing Q3 what you just reported and sort of the building back of Q4? And Julia gave some really good color there with respect to UCAN. Can you characterize what you're seeing in UCAN relative to the rest of your operations globally against that broader ad environment?

Thanks, Eric. We're pleased with another strong quarter in Q3 at 17% revenue growth. And as I'm sure you've noted, we've been quite consistent in our growth and in line with the long-term targets for revenue and margin that we laid out at our Investor Day 2 years ago. So we continue to feel good about that mid- to high teens revenue growth target over the long term that we laid out at Investor Day and our ability to consistently deliver. It's also really important to note, and this gets to your sort of question on sort of the broader environment. It's also important to note that we grew 17% despite operating in an environment where some of our largest retailers in UCAN pull back spend across the industry, not specific to us, a pullback across the industry as a navigated tariff-related margin pressure. And we think that's disproportionately impacting large retailers, but that is a segment that we have more exposure to than other platforms given our focus on shopping though we continue to grow in other verticals and segments of the market in addition to those. Additionally, as advertisers are adopting AI-driven platforms, there's a next level of optimization in the AI ad platforms that's taking place right now, where bidding is getting further aligned to advertisers measurement sources of truth and more events that lead up to a conversion or being incorporated. We think that presents an upside opportunity moving forward. We began that journey earlier this year with ROAS-based bidding, and we've seen good results there. For example, after launching ROAS bidding in Q1, we saw a 100% increase in shopping SKUs with paid ad impressions across the platform in Q3. And has driven entirely by ROAS bidding adopters, as I noted in my remarks, and more of those advertisers are uploading greater portions of their product catalog. We've talked about that opportunity to get deeper into the catalog even of our largest advertisers. We think this will continue to help us do more of that. Going forward, we see meaningful potential to expand further into AI-based optimization of other events that are valued in advertisers' measurement sources of truth. And while many advertisers have adopted these solutions first with the larger platforms as typical of the adoption cycle, we are testing this with some of our largest partners, and we're seeing really good early results. Certainly, more of that is in front of us than behind us. But the good news is that the alignment of AI bidding with the advertisers measurement source of truth is market expanding. I think we've already seen that reflected in some of the larger platforms and what they have been out in the market with. And that tended to give more credit to events leading up to a conversion, such as view through attribution, which should be good for a full funnel platform like Pinterest. So this should accrue to our advantage in future quarters as we continue to roll out those features are very early on our platform now, but we are seeing good early results. And through our broader deployment of these additional Performance+ solutions in 2026, we think there's continued opportunity there. So we know we're driving performance for advertisers. Clicks to advertisers increased 40% in Q3 and clicks to advertisers outpaced revenue in all of our reported geographies. In fact, clicks to advertisers were up over 5x over the last 3 years. But clearly, we have more to do to get proper credit for that performance.

The next question comes from the line of Rich Greenfield with LightShed Partners.

[indiscernible] seen that's allowing users to actually remove AI content. I guess the big question is, how do you know what's AI generated? What's not? And why did you make that decision? It seems like a lot of your peers are actually encouraging and want AI content because they want more content to keep people even more engaged, to sell more ads or to build the business. And so it seems like you're sort of doing the opposite, and I'm curious or maybe allowing the opposite versus doing the opposite. Why and help us understand how this all works and why you're doing it?

Yes. Thanks for the question, Rich. So it's a really great clarification. And overall, we see Gen AI created content as a big tailwind for our platform, and we are very much embracing it. Part of embracing it is addressing content quality. And we address content quality primarily through our recommender systems, but then also by giving users choice and saying what they want to see. And so I think you've heard this across pretty much every platform out there, that there are some users and some use cases that would like to see less AI-driven content. And so we're simply giving the user choice, but we're also leveraging all the really great AI-generated content that is available out there. And I think this is not dissimilar to what's happened in prior expansionary moments with content, whether you think about what happened with online video or short form video or the ability for everyone to take a photo with their phone, initially, people react to changes in the volume of quality leading to some examples of lower content quality. But if you have recommender systems that can really parse what's great quality versus not great quality, then you can give users the best of that expanding content corpus. And when you look at the engagement on our platform, I think that we -- I think that demonstrates that we're doing a pretty good job of parsing all that great influx of AI-generated content that's out there and figuring out what's going to be relevant for which user at which moment in time. And then what we're doing is augmenting what our recommender systems can do with the user's ability to give us direct feedback if they want to see less AI-driven content. You do have some use cases where someone might say, well, hey, I'm looking for a specific thing to provide architecture on my home. And so I need to know it's a real picture of something that could actually be built versus something that is AI generated. I think over time, users will accept more and more of that as the quality filters get better and better. But again, overall, we see Gen AI content as a significant tail on the platform already. And it's really about getting the content quality right, which primarily we do behind the scenes on our users' behalf, but we're giving them tools to say when they want to see more or less. And then specifically to your question of how do you tell. I would say there's not a precise ability for any platform to catch a 100% of what is AI generated. There are some industry-level tags that we would act on. There are also things where we're looking at metadata and other indicators that give us indication of that. But that's why we say see less, not see none of because the ability to precisely spot that is not perfect for any platform. And I would just say, over time, I think it will get to a place where almost every piece of content you see will have been at a minimum edited by AI in some form or other. And there are analogs this. If you think back to -- it used to be that people would get really sort of insensed or some people get insensed if photos had been photoshopped. Well it has long since been the case that nearly everything you would see out there would have been not just Photoshop, but you have filters and all these things that even the average person can do. We think that over time, almost every piece of content you see will have been AI modified in some way, and it really will come down to content recommendation and content quality. But in the near term, we're giving users choice and allowing them to express what they want to see in honoring that choice.

The next question comes from the line of Shweta Khajuria with Wolfe Research.

I was wondering if you could please talk about your relationship with Magnite and more broadly about your efforts to add new sources of demand and perhaps time line around that?

Thanks for the question, Shweta. We've been very consistent from the beginning as we think about our programmatic and third-party strategy. Our first-party ad demand continues to be the primary driver of our growth with third-party demand really complementing and rounding out our auction when there may be gaps in the auction. With respect to Magnite, when we announced that, we said it would take time to integrate test and doing more fulsome go-to-market. We're still working through that testing now, and we're in the early days. Today, most of our efforts have been with respect to 3P have been focused on bringing on new sources of demand on the platform as many of these programmatic budget pools are large and new to Pinterest today. So we continue to see that consistently with how we have before. But I would also call out that there's a next potentially meaningful opportunity that we're also starting to look at that we think we bring a unique audience to these budget pools given the high intent nature of our audience and the visual discovery that uniquely occurs on Pinterest. So as a result, we're also starting to explore very early testing how valuable our audience could be even beyond our own platform. When you think about the very strong commercial intent that we see with our users, obviously, we satisfy a lot of that commercial intent on our platform, but we think that could be helpful beyond our platform. So we're in very early days of testing that.

The next question comes from the line of Mark Kelley with Stifel.

Bill, I appreciate the extra color on Performance+ tonight. I was hoping maybe we could drill into the SMB and kind of mid-market opportunity, just a little bit more. I know you threw out that -- that's that, the 24% higher conversion rate. Is that pretty like standard, whether you're a smaller advertiser versus a more sophisticated and larger brand? And then second, maybe just to clarify, the 10% to 15% of revenue running through P+, was that just an SMB and mid-market stat? And where might that go over time?

Thanks for the question, Mark. So first of all, we feel really great about Performance+ and how well that's working. We're not even a year end to the deployment of that. And as a reminder, the larger platforms are many years into their rollout of this. So we're at year-end, and we feel great about the progress that we've made. People's campaigns or bundled suite of AI-enabled automated features. It's really driving much better performance and reducing the inputs, 50% fewer inputs. So I called a little bit of this out, but for example, retail advertisers using Performance+ campaigns are seeing a 24% higher conversion lift versus traditional campaigns. It is also helping significantly as we expand into the smaller and mid-sized advertisers, those tens of millions or hundreds of million -- $10 million to $100 million of GMV, who really value that automation. And for that segment, we see a 12% higher monthly revenue growth rate of adopters of P+ campaigns versus non-adopters and approximately 15% of our current revenue the segment is growing nicely. So there's a lot more to do to deepen share of wallet there. And again, as a reminder, this is a segment of the market that larger platforms have had much more exposure to. We are earlier on this segment of the market. But P+, we see as a significant unlock there and it's having exactly the effect of intended, but we still have a lot more of that opportunity in front of us than behind us. So we're quite excited about what we're seeing from Performance+ there. And then on the larger segment of this, ROAS bidding, as we've talked about, is really helping us get much deeper into the catalog of our -- of the largest retailers out there. So I've shared 22% of retail lower funnel revenue, both enterprise and SMB. So -- and that's just launched in 2 quarters on ROAS bidding. So road bidding is really having quite an effect. And I shared how much is getting us into many more SKUs. So when we think about ads as relevant content, we want to make sure that we have ads that line what you are actually looking for, getting a much larger portion of large retailers catalog on our platform is super helpful to that and ROAS bidding as a subset of Performance+, as we rolled out over the last couple of quarters, effectively has doubled the number of SKUs with a -- that are delivering a paid impression on our platform. And it's because that ROAS bidding allows advertisers to more precisely tune and really the AI tunes for them across catalogs with high variation in price points and margin points and those kinds of things. And so we see really great progress there, but we think there's a lot more of that to go. So when we think about, I would just say our opportunity globally, but specifically UCAN, even as we think about growth in UCAN it's really three things: the expansion into the full catalog with the largest retailers, driven by more granular bidding like ROAS bidding from Performance+. And so -- we've got good results of that. We're just a couple of quarters in, a lot more to do there. They can get us deeper into the catalog to the largest retailers. And then secondly, it's mid-market and SMB driven by Performance+ where we're just getting started in that segment of the market, but seeing really good promising early results. And then the third point that I alluded to in one of my prior answers, the AI-driven alignment of bidding in -- bidding systems to the advertiser measurement sources of truth. That's giving a clear view of full-funnel attribution and events across the funnel. And other larger platforms have fully roll that out. We are just getting started testing on some of our legist advertisers. We're seeing really promising early results there. So we think that will -- that's having an expansionary effect that you're seeing with some of the largest platforms out there. We think that's a real opportunity for us as we look into next year. So hopefully that helps give a little more color on the broader sort of things we're thinking about with Performance+ there.

The next question comes from the line of Ross Sandler with Barclays.

Great. Just want to bring Julia in on key investment priorities for next year. And how do we think about the pace of EBITDA margin expansion in '26 compared to the pace we're seeing in second half '25? And then, Bill, just a follow-up on the agentic question. So Walmart's integrating its catalog into ChatGPT for this checkout service. I know it's early days, but for something like that, do you view that as neutral, positive, negative in terms of where that kind of a marketer might move their Pinterest ad budget in the future? Is this going to help you guys or potentially create a new headwind? Any thoughts there?

Thanks, Ross. I'll take the first question. So it's still a bit too early to talk specifically about 2026 as we're still reviewing those plans internally. But stepping back, we still feel confident in the long-term margin targets that we provided in 2023. And at that time, just as a reminder, we said we would target a 30% to 34% adjusted EBITDA margin over a 3- to 5-year time horizon. So now here in 2025, we're already approaching 30% for the full year. So we've made a lot of progress already by growing the top line while investing thoughtfully primarily in R&D and to a lesser extent, the sales area as well. So looking forward, we continue to see many investment opportunities with high ROI, particularly across AI, that power user experiences, including our new Pinterest Assistant, which we believe will help keep us on the leading edge in visual search and discovery as well as ongoing investment in our performance ads platform. On the gross margin line, in aggregate, we expect cost of revenue next year to grow more in line with the business going forward, so there can be some variations quarter-to-quarter, and there are a few factors at play there that I'll call out. Cost of revenue, which is mostly our infrastructure costs will naturally rise, of course, due to kind of ongoing user and engagement growth. Additionally, we've previewed for multiple quarters now on several calls that we expect to see diminishing returns from the infrastructure cost optimization work that we've undertaken for the last 2-plus years. Now partially offsetting this kind of natural upward pressure on cost of revenue is the fact that we're able to apply AI use cases sort of directly in service to monetization, where we see immediate revenue lift, right? So we're not rolling out new features and then planning to monetize them years later. They monetize right away as sort of our general philosophy. We're also being cost efficient with our model usage, including, as Bill alluded to, utilizing open source models where applicable that come at a meaningful reduction in costs. So -- in summary, we continue to stand by our long-term adjusted EBITDA margin targets that we've always said, though, that the rate of adjusted EBITDA margin expansion would vary year-to-year. We've made a lot of progress towards these goals, and we'll continue to be thoughtful about how we invest moving forward as well.

And then following up on your agentic commerce part of the question, Ross. A couple of things I'd say. One is that for the largest retailers, we have catalog integrations. I shared how we're getting deeper into their catalog with our Performance+ capabilities and ROAS level bidding and things like us getting a much broader part of the catalog available for ads, but we have those catalogs for organic shopping for push button type buying and linked accounts. We have that with Amazon. We've had that for some time in millions of users on Pinterest take advantage of that and have a great experience. So we feel really great about the shopping experience that we're providing. And I think overall, I would just say it's worth noting that there's an expansionary moment happening with search generally. And I think users just as I did in an app-driven world, are thinking about different places to go for different types of experiences. Traditional search was always great for things like product research. And I think broad-based chatbots are sort of the next evolution of that sort of research type of shopping behavior if you're trying to figure out like every attribute of the latest 4K Ultra TV or things like that. Search was -- traditional search was always great for that, and then chatbots are even better for that. But we're solving a different type of shopping journey on Pinterest, really the more the -- I'll know when I see it a tight problem. And the best evidence I can share of users think about these things as distinct and separate is that over the last couple of years as you've had an explosion of usage in AI chatbots, we've put up 9 straight quarters of record high user growth, shopping being at the very center of that. And it's because we're taking a visual first approach driven off the human curation that happens on our platform where we just understand user style and taste and preference. We just have a unique signal is completely distinct from any place else in the western world, and that's why we're able to do things like what we've showed on our -- the relevancy of our latest multimodal visual search models, where we're able to outperform the off-the-shelf models by over 34 percentage points on the relevance of shopping recommendations on our platform. That just gets you to a bit of the kind of unique things we're able to do with users. And so I think, again, it's a market-expanding moment. Multiple players are growing simultaneously. And very clearly, we are one of those players that is growing, delivering a lot of value for users, a tremendous amount of shopping occurring on the platform. And a tremendous amount of searching happening on the platform, I shared that in my prepared remarks, the 80 billion-plus monthly queries, the vast majority of which are visual in nature, which just, again, gets to how we're doing something that's very different than traditional search or even what chatbots are doing. So hopefully, that helps contextualize that.

The next question comes from the line of John Blackledge with TD Securities (sic) [ Cowen ].

Great. PINS historically has had all of its infrastructure running through AWS. Given the move in recent years of companies shifting to multiple cloud vendors, how should we think about PINS potentially diversifying to other cloud platforms?

Thanks, John. So first, I'd say we view our infrastructure and platform as strategic assets that support our performance, reliability and our AI road map. And Amazon is a fantastic partner for us, to be very clear. But in the same way that we are constantly testing all the various LLMs and benchmarking different LLMs across one another. We're also constantly assessing the best infrastructure options for us as we move forward, especially in an AI-driven world. And that infrastructure includes LLM, chip providers and hyperscalers. So again, if you look at what we've done over the last few years, you've seen us put AI at the center of our business, really effectively align that with the ability to monetize for users, deliver great results for users and do so cost effectively. And again, we've had great partnerships there. But the space is evolving rapidly, and we continue to pay really close attention to that and benchmark across multiple providers on each of those sort of layers of the stack. Hopefully, that's helpful.

This will be the last question from Michael Morris with Guggenheim.

Bill, you referenced a couple of times deepening engagement per user. And I'm hoping you can add some context to that. Are you talking about time spent per visit or frequency or some other metric? And do you see that as a leading indicator of reaccelerated growth in the UCAN market? And then secondly, on international, the growth there is significant. That has accelerated. How much runway do you see to continue to grow internationally at that elevated level? And maybe -- you gave us a few drivers, but what do you see as the 1 or 2 key drivers as we look forward?

Yes. Thanks for the question, Mike. On the deepening engagement per user, we talked about this very consistently over many quarters now is that we are deepening engagement per user across the areas that we want, which are around search, curation, clicks and actions. And I share a little more color on this call around the search behavior where we are getting more searches per user. So it's not just our searches are growing, we are getting more searches per user. We are getting it in the ways that we won't particularly on where we are very highly differentiated around our visual searches related items and other forms of visual searches drive the vast majority of that search behavior, but it is more searches per user happening on our platform even as we put up record high levels of users and also significantly growing actionability. I shared just how much the clicks to advertisers have grown 5x over the last 3 years, we're driving a tremendous amount of clicks to advertisers. So it is that great relevancy driving great recommendations that leads to actionability that leads to more users coming back for more searches and more actions like that is that flywheel is spending on the deepening user engagement. And to your question around is that a leading into monetization, I think absolutely yes. What I would say is that 3 years ago, Pinterest was pretty much upper funnel only. We've had a major transformation of the business, both in terms of user engagement, which, again, I think, continues to be the brightest spot in the business. And advertisers will always follow where users and commercial intent are. We stood up a performance ad platform pretty much from scratch over the last couple of years. And we are still a long way from having the capabilities of the very largest platforms. But even as we've made the basic capabilities of that available, we've really broken into those always on performance budgets. We have a lot more to do there. And with UCAN specifically, I shared those sort of 3 areas where I think there is significant opportunity around getting deeper into the full catalog of the largest retailers, mid-market and SMB that are really driving a lot of the growth across the broader market, but it's a newer area for us. And the alignment of AI bidding systems and advertiser and [indiscernible] giving a clear view of full funnel attribution. I think those are things that will help us capture more of that value that we are driving. But I do think absolutely yes. Is generally true that user behavior is a leading indicator of where the advertiser dollars are going to follow. And we have the largest platforms are many years into their AI-driven ad systems, we are only a couple of years into ours. But because the user behavior is so strong, that's what's really letting us make progress. And on the international side, as I shared in my prepared remarks, the playbook that we have used in UCAN, we are now exporting and we're seeing that really take hold internationally. We are still early on in our work there. We have a lot more of that opportunity in front of us, but we're really pleased with the progress we've made and how it's starting to show results. As I noted in Q3, Europe ARPU grew 31%, while Rest of World ARPU grew 44%. We have a lot more of that to do. And I shared in my remarks, how shopping ads are really at the center of what's driving that. So it is that commercial intent commercial behavior. So a lot more of that to go, but it's working in UCAN and is exporting well. And again, to put it in perspective, 2 years ago at our Investor Day in September of '23, shopping ads represented just 9% of international revenue. In Q3 of 2025, it reached 30%. So Q3 shopping on revenue in both Europe and Rest of World grew over 2x faster than revenue growth of their respective regions. So we believe there's many years of runway of continuing to grow ARPU as we increase product catalogs, add additional ad demand, drive up relevance and thus, our ability to take a bad load, certainly for international, but again this year, we think there's a lot more to do in UCAN as well.

Thank you. I will now hand the call over to Bill Ready, CEO, for any closing remarks.

Thanks again to all of you for joining the call and for your questions. We look forward to keeping this dialogue going, and we hope you enjoy the rest of your day.