Hello, and welcome to the Spotify Q2 2026 Earnings Call. I would now like to turn the conference over to Bryan Goldberg, Head of Investor Relations at Spotify. You may begin.
Great. Thanks, operator, and welcome to Spotify's Second Quarter 2026 Earnings Conference Call. Joining us today will be our Co-CEOs, Alex Norström and Gustav Söderström; and our CFO, Christian Luiga. We'll start with opening comments from the team. And afterwards, we'll be happy to answer your questions. Questions can be submitted by going to slido.com and using the code #SpotifyEarningsQ226. Analysts can ask questions directly in Slido, and all participants can then vote on the questions they find the most relevant. If for some reason you don't have access to Slido, you can e-mail investor relations at ir@spotify.com, and we'll add in your question. Before we begin, let me quickly cover the safe harbor. During this call, we'll be making certain forward-looking statements, including projections or estimates about the future performance of the company. These statements are based on current expectations and assumptions that are subject to risks and uncertainties. Actual results could differ materially because of factors discussed on today's call, in our shareholder deck and in filings with the Securities and Exchange Commission. During this call, we'll also refer to certain non-IFRS financial measures. Reconciliations between our IFRS and non-IFRS financial measures can be found in our shareholder deck, in the financial section of our Investor Relations website and also furnished today on Form 6-K. And with that, I'll turn the call over to Alex.
Thank you, Bryan. Hey, everyone, thank you for joining us. I hope you're having a great summer. Today, we'll pick up where we left off at our Investor Day in May, updating you on what we said, what we've shipped since then and also what the early signals are telling us. So Q2 was another quarter of healthy broad-based growth. Revenue grew 15% year-over-year on a constant currency basis, accelerating from 14% in Q1. Gross margin hit a record of 33.4% and free cash flow continued to strengthen. And we beat our subscriber guidance, crossing 300 million subscribers for the first time, all while continuing to grow engagement with a number of active days for global subscribers increasing. What pleases me the most is the shape of the quarter. More people are choosing Spotify, they're engaging more deeply and they're converting. We've been working on turning our outperformance in MAU into revenue growth. So to capitalize on this opportunity, we are adjusting elements like product optimization and ad load, among other things in select emerging markets. Now this strategy carefully increases friction in our free service with a goal of driving higher user conversion and revenue growth down the line. Yes, this will show itself in our Q3 MAU, but we believe it's well worth it. And as we've shared previously, the free-to-paid conversion cycle in emerging markets grows differently than our established markets. So while a move like this one will take time to play out, the opportunity is vast. This will be additive to our potential over time. And as you will hear soon from Christian, we do not expect it to come at the expense of our subscriber growth. Now we've talked before about rebuilding our ads business and the results are really starting to show. On the supply side, our audience of very attractive global users keeps expanding our inventory. And on the demand side, the enhanced technology we've deployed is making it much easier for advertisers to reach these engaged users. Just one example: our automated channels represent nearly 40% of ad-supported revenue in Q2, up from just over 30% in Q1. And active advertisers grew 60% year-over-year. We're now set up in a way we weren't a year ago, and we will keep building from here. Another development from the quarter worth calling out is our launch of Reserved. At Investor Day, we called it one of the most wonderful improvements to Premium in our history, and the early signs really back that up. Since launching in the U.S. in June with Live Nation, Reserved has supported multiple tours with nearly 100,000 tickets reserved through Spotify. For some tours, we sold through 100% of our allocations. And Live Nation upsized them mid-run. The biggest fans get first access, artists get their most dedicated audience in the room and every seat we fill makes a Spotify subscription just more valuable. I'd like to use the remainder of my time on something that Gustav and I think about a lot: Spotify's position. And these are unique times. Business history has produced just a handful of companies with hundreds of millions of recurring paying customers worldwide. And Spotify is proud to be one of them. And like many of the others, we built that scale on a single product. That is a unique territory to be in. I'm a student of Charlie Munger, who once said that the one structural advantage that matters disproportionately is scale. If you're choosing a music streaming service for the first time, it matters that one brand comes with several hundred million passionate ambassadors. And perhaps most consequentially, the scale and the cash generation we now have lets us innovate and invest in building a much better product, adding more value for every subscriber around the world in the most cost-efficient and competitive way. The point is scale matters. In the past five years, we've added more than 25 million net subscribers every year, growing in developed and emerging markets alike. So with over 300 million subscribers and 777 million users, we have achieved consequential scale. Our financial picture tells the same story. Since our last Investor Day in 2022, revenue has compounded at 18% a year, reaching EUR 17 billion in 2025. Gross margin went from 25% to over 33% this quarter. We became more disciplined with OpEx, which has turned into a growing positive operating margin. And last year, Spotify generated EUR 2.9 billion of free cash flow, and we expect that growth to continue. That strength is why we set the 2030 targets we did in May: a mid-teens revenue CAGR, gross margin of 35% to 40%, operating margin above 20% and strong growth in free cash flow. Gustav and I believe there are opportunities only Spotify is positioned to invest in because of our scale, the health of our business and our focus. So we rigorously explore new premium offerings and new verticals with our customers. We covered these at length at Investor Day, but there are a few worth repeating. The first is AI. Many investors ask about our Large Taste Model, which learns from the 3.4 trillion events our users generate on the platform every day. But the reality is we've been investing in AI-driven personalization for more than seven years. Today, AI-powered experiences like DJ are used by roughly one quarter of our active users. And Prompted Playlists, our newest addition, gives users direct control over the algorithm. Around 14 million of the first 100 million users we have rolled it out to are already using it, and the early retention improvements look promising. The second big area is the power law. As we've grown, we've observed that the usage of our products, features and content follows the power law. At the head of that curve are many millions of people who simply want more and they're willing to pay for it. Audiobooks proved it first in just a handful of initial markets. We see that overall Audiobooks penetration among premium listeners has more than doubled this year. And Audiobooks+, the add-on, has passed $100 million in annual recurring revenue. It's subscriptions on top of subscriptions with more on the way. So combined, these two make for something powerful: inference-driven products carry a marginal cost per use, so they need a compatible way to monetize. And that is exactly the muscle we've built over 20 years of freemium, feature-gated, usage-driven products optimized for the best value-to-price ratio. Now we will price and optimize these features and content just as we have successfully done with our Premium product. So in conclusion, we have a scale that few companies in history have reached, a business that is healthy and compounding and opportunities only we are positioned to pursue. Spotify lives across your whole day: the commute, the workout, studying, gaming, dinner table and sleep. At our scale, that is rare. Most products own one or two contexts. Our position gives us an opportunity space as wide as our users want it to be. And at the Investor Day, we told you where Spotify is going. And this quarter, we're building momentum behind that. With that, let me hand it over to Gustav.
Thanks, Alex. I want to use my time on two things: how we're building and what we're building. And both matter for the durability of our business over the next several years. So a couple of quarters ago, I mentioned our internal tool Honk, which enables our engineers to go on Slack from their phone and ask an AI agent to fix a bug or add a feature. And then they get a testable build back before they even reach the office. That way of working is now ubiquitous and it's unlocking a compounding advantage. So Honk is our coding agent, but there's something new that's called Chirp, which is the engine that we've built underneath it. And you can think of Chirp as something that you use instead of using Claude or other coding models directly; it sits in front of them. So Chirp lets our engineers switch models mid-task and route every job to the best available price performance, including open source models that we host ourselves so that we are never locked in. It also shares the context across not just different models and different developers, but across the company. So we don't pay for the same reasoning twice, and we don't lose our own data. And it also shows us our inference spend down to the individual developers. There is plenty of industry debate about AI investment and costs. But our view is that being an AI beneficiary means winning on the cost side, too. We hold a high bar for every investment that we make in this area. And we invest only where it generates real advantage and sets us up well to drive growth and improve margins over time. If something works, we double down. If it doesn't work, we eventually move on. Over the last three years, we have not increased head count while revenue per employee is on track to double. So as you can see, the operating expense growth is not coming from people. It's coming from compute and marketing. Both of these are variable and entirely in our control. And we will continue to invest in AI on our terms. We're vendor agnostic. We have controls over our usage, and we always undertake these investments relative to the returns that they can support. The inference we give users is also under our control, which means that the cost curve is also under our control. At Investor Day, we laid out four big ideas about where Spotify is heading. One quarter later, we've shipped against all four. We said the world is moving from recommendation to generation with users in control. Today, Talk to Spotify, Personal Podcast and Studio by Spotify are in users' hands. And in the coming weeks, we'll be rolling out Prompted Playlist for Audiobooks to our best-in-class books experience. And Alex mentioned our Large Taste Model, which is a big investment for us. Let me tell you what it's doing so far. In the first month since we deployed our new Autoplay recommendation system powered by the Taste Model, active days have increased, which is very hard to do at our already high engagement levels. Autoplay minutes and track saves have both grown significantly. And Autoplay drop-off has declined. And in our chat experiences, the Large Taste Model is doing the same: minutes are up double digits, more active days and more saves into libraries and playlists. These are powerful inputs to our customer retention rates and lifetime value and among the most challenging metrics to move for us. Now the industry watches this scaling curve and the scaling laws play out with general LLMs. And when we think about it, we are now watching it play out with our Taste LLM on a model that no competitor can buy. We said the engagement follows a power law, and we launched our first one-time credit add-ons to serve the head of that demand curve. From DJ to Daylist, we know these investments take time to scale, and they pay off over time. These items are capped by design. Every credit purchase carries a defined amount of inference behind it. So usage, cost and monetization stay connected. We also said that Spotify is moving from single player to multiplayer. And this quarter, that stopped being a road map and became behavior. Messages is live, and people now react to tracks, reply and share music directly inside Spotify. Listening activity shows you what your friends are playing right now and listening stats is becoming something you share and compare, not just a private report for you. Almost 50 million people are already using Jam every month to listen together in real time. You're no longer alone on Spotify. You're there with your friends, your real-life friends. And we said we would keep optimizing for time well spent and increasing the value of Premium. One example that I'm personally very passionate about here is fitness. Last week, we began rolling out Running Mode. You tell Spotify in plain language to build a running playlist for an 8-minute mile or an interval session. It understands not only your taste and the BPM, it serves you your favorite tracks to your exact cadence, mixing them together seamlessly so that you can literally run to the beat. It adds optional audio coaching right in your ear and it stays fresh every week. This is an experience that only Spotify can deliver because of the investments we've made. Looking beyond Running Mode, I'm excited to share that SongDNA has now been used by more than 100 million subscribers, making it one of the fastest-adopted features that we've ever shipped. Finally, our music remix capability. We're excited about the product experience and what it will unlock between artists and fans. Our model is one where artists choose which ones they want to make available for remixing with consent, credit and compensation built in from the start. That model is winning over the industry. Following our agreement with UMG in May, today we're announcing a deal with Merlin, the digital licensing partner for the world's leading independent labels and distributors. The agreement gives artists across 30,000 labels in Merlin's network the opportunity to participate and make their catalogs available for covers and remixes. The enthusiasm from the independent community has been striking. And we hear that same support across the wider market. Deals like these take time, but we built this so that virtually every rights holder who wants can join in. Big picture, there is more work to do before launch, and it will take time to scale into a material revenue driver. But we are very excited about the potential for Spotify for Artists. So if there's one theme I would leave you with today, it is this: Investor Day described where we believe Spotify is going in the future. And this quarter demonstrated that we are already building that future—better engineering, faster shipping, new products and new ways for users to engage. We're still in the very early stages of what is possible and we'll continue to have a high bar for investments. Our margin is a managed outcome, not a byproduct. Our job remains the same: understand the technology early and deeply and turn it into something that people love, creating value for our stakeholders. Now Christian will take you through the numbers.
Thank you, Gustav, and thanks, everyone, for joining us. I will cover the quarter two results and then provide some perspective on our outlook. Unless otherwise noted, as always, our referenced growth metrics are presented on a year-over-year constant currency basis. We're pleased with how the business performed in the quarter. MAU continued to grow healthily at 12% year-over-year with notable outperformance in Europe and North America. Our net additions of 16 million were one million below forecast. We added seven million net subscribers during the quarter, growing quarter-on-quarter across all regions with notable outperformance in Rest of the World and North America. We finished at 300 million, one million ahead of our guidance. Total revenue was EUR 4.8 billion, growing 15% year-over-year, which was an acceleration from the 14% we delivered in quarter one. Premium revenue rose approximately 16% year-over-year versus 15% last quarter, driven by subscriber growth of 9% and ARPU expansion of 7.4% year-over-year. Our ad-supported revenue grew 3% year-over-year, which is consistent with quarter one. Our automated sales channel continued to grow fast and represented nearly 40% of our ad-supported revenue in quarter two, up from just over 30% in quarter one. This strength was largely offset by expected declines in our direct sales channel. Importantly, our price optimization work is now complete, and this channel is stabilizing. We have also completed the migration of our ad inventory to our in-house ad server, allowing us to further streamline our sales process and capture more demand. With respect to our outlook, nothing has changed. We continue to expect our ads business to inflect towards double-digit growth in the second half of 2026. Gross margin came in at 33.4%, surpassing guidance by 30 basis points, a year-over-year expansion of 193 basis points. Favorability versus our guidance was driven primarily by quarterly timing shifts related to our growth investments. We also saw a small one-time benefit from cancellation of the Digital Service Tax in Canada, where we reversed an accrual from previous years. Operating income of EUR 655 million was EUR 25 million above our guidance of EUR 630 million, delivering an operating margin of 13.7%. Social charges contributed EUR 9 million of outperformance versus our forecast due to share price movements in the quarter. Excluding the non-forecasted social charges favorability, we came in EUR 16 million above guidance, driven by the gross margin outperformance. Finally, free cash flow was EUR 797 million in the quarter, up 14% year-over-year. The quarter was slightly lower relative to quarter one, partly as a result of timing of cash tax payments. Our first-half working capital benefit remained consistent with the prior year. On capital allocation, we have repurchased $662 million in shares year-to-date through August 3, representing a 30% increase over 2025 levels. In aggregate, we have bought back nearly 2.2 million shares since we resumed repurchasing activities in 2025, or approximately 1% of shares outstanding. Given the stronger cash flow in the years ahead, even with M&A, we expect that we will also return cash to shareholders. As of close of the quarter, we had EUR 9.4 billion in cash and cash equivalents and no debt other than lease liabilities. Looking ahead to quarter three, we are forecasting MAU of 788 million, an increase of 11 million from quarter two. As Alex discussed earlier in more detail, this guidance includes product optimization activities in emerging markets, while growth rates in developed markets remain stable. On subscribers, we are forecasting 305 million for quarter three or net additions of five million. We continue to see the business as well positioned to drive another full year of healthy MAU and subscriber growth. We're also forecasting total revenue of approximately EUR 5 billion in quarter three or 14% growth. This reflects an ARPU increase consistent with quarter two and improved growth in our advertising business that I mentioned earlier. We anticipate a quarter three gross margin of 32.9%, approximately 130 basis points above the prior year. Our gross margin outlook incorporates continued strengthening in our core business, reinvestment into new products and the typical charge we take in the third quarter to account for our annual exposure to regulatory fees in one of our markets. Moving to operating income, we're guiding to EUR 670 million in quarter three. This reflects the above, along with a temporarily elevated operating expense in quarter two and quarter three. We continue to expect these marketing and AI-related investments to drive approximately EUR 200 million in incremental operating expense for the full year. Quarter four is well positioned to see a moderation in the rate of year-over-year operating expense growth as we move past the heavy concentration of this year's marketing activities and calibrate on the other costs. In 2026, we continue to be flat on the number of employees. So this year's investment cycle is not driven by structural headcount additions. It's marketing and AI related; both are investments that are entirely in our control. As we stated at our recent Investor Day, we remain focused on striking a clear balance between LTV-enhancing investments and sustainable margin expansion. Although we do not provide full year guidance for gross margin and operating margin, we continue to expect both to improve in 2026 on a full-year basis. As we have said, the quarterly progression of our margins is variable and dependent on the timing of investments. As you've seen from us historically, the rollout of new features and products can temporarily moderate margin expansion while setting the business up well for future monetization potential. We continue to expect meaningful year-over-year growth in free cash flow in 2026. So in conclusion, we had a solid quarter two. We remain well positioned to continue compounding growth, profitability and free cash flow. And with that, I hand it back to you, Bryan, for Q&A.
Great. Thank you, Christian. Again, if you've got any questions, please go to Slido.com, #SpotifyEarningsQ226. We'll be reading the questions in the order they appear in the queue with respect to how people vote up their preference for questions. And our first question today is going to come from Jessica Reif Ehrlich on the product road map. You have a unique global platform for both audio and video products. You provided a robust product road map in May at your analyst event. What products are you most excited about in the near term? And where do you think you will get most traction over the next three to five years?
So thanks for the question, Jessica. This is Gustav. I'll start and then maybe Alex wants to jump in because I know he's also very excited about product. This is kind of like asking me to choose between my children. So it's a tough question, but I'll try. But first, I kind of want to level a little bit because what I am most excited about is the system that we've built and what we presented at Investor Day. So the way to think about it is some time ago, over a year ago, Alex spoke and we realized that the world that we were in where software development was mostly an amortization game— you developed once and then you amortize your developer investment over more and more users— that world was going to change because inference adds a variable cost per user. So we looked at this and we said the game is going to change. There's going to be more variable cost per user. This could be a headwind or it could be a tailwind. So we decided to start working on changing the model with this power law that Alex and I talked about where some users use the product a lot more than others, which didn't really matter so much in the amortization world because it didn't have a large variable cost, but it will matter a lot in this new world. We decided to change the model. So we've built this platform where we can decide how much inference we want to put into the free tier, how much we want to put into the premium tier. And then instead of saying that's where the party ends, you have to go somewhere else. We say to people, if you want to live in the future, you can. You just pay more. So we let some people who are prepared to pay for it run ahead of other users. And we demonstrated this model with Audiobooks, which works exactly like this. We give about 15 hours of Audiobooks in the premium tier. Some people want much, much more than that. They're allowed to run ahead of everyone else and pay for it. So we spent a lot of time building that infrastructure. And as we don't like to ship ideas, we didn't really talk about it until Investor Day. That's kind of what we revealed this system to be able to have different types of users with different levels of monetization and different needs on the same platform instead of being sort of kept back by the average or what the least engaged user can afford or we can afford to give them. So that is what I'm excited about, having built this platform. Now we're launching a long range of products on top of this. And I can tell you sort of what I'm excited about there and which ones in the near term and longer term. But it's important to think about the structure. I'm a systems type of guy. That's what really excites me. In terms of actual products, I would say right now, the thing that excites me the most is Reserved. It is probably the feature that the most people ever have said, 'This is the best thing you ever did at Spotify.' And we've seen, as Alex said in his introductory remark, really exciting numbers there. So I think this is one of the biggest values that we put in Premium ever. As we said before, this is actually unique to us because of the structure of these deals. So that's very exciting to me. The second thing I would say, more in terms of AI, which I mentioned in my comments, is the LTM. The LTM, we talked a lot about it at Investor Day, pretty big investment for us, both in terms of personnel but also training costs. So it's very good to see it paying off. The bet we made there was that the old type of machine learning had capped out. More data and larger models did not produce better results. The sequence-based LLM follow different laws, the laws that are called scaling laws. Where more compute, more parameters and more data actually produces better results. And now we're seeing those. We're seeing these scaling laws play out on the inside in terms of taste and recommendations. So that's very exciting to me. The third thing that I'm really excited about, which I think Alex mentioned as well, is SongDNA, which is a feature that we put a lot of love into. We acquired a company called WhoSampled to power this feature. So we invested something there. We took some costs and now it's paying off with over 100 million users using it and loving it. It's also very unique to us. So that's maybe my third if I would rank them. Another one, which we don't talk so much about, but that I'm personally very excited about is music videos. We invested in music videos and a great music video experience for a very long time. Now we have both a good experience and the catalog. And we are seeing that songs with music videos, specifically new releases, are performing much better, which was the bet, meaning that music videos are of outsized importance when you're discovering a new artist because you're wondering who the artist is, you want to see them. It's not that important the tenth time you are listening to the song, but it's very important the first time. So that's quite exciting to me. Obviously, I think it's a tie between maybe GenPods and Running. Running just because fitness is so close to my heart. I think we have a very unique experience there, which no one else can really do. There is no other service that can give you a playlist at the cadence you're running in, with your favorite songs and beat-match them and mix them together. GenPod excites me because it's a podcast that didn't exist before. It's a podcast about you, both your taste, the new releases that came out in terms of music, the podcasts that you missed last week, the books that you should have read or that are coming out. But also if you ask for it, exactly what happened in your neighborhood. There's no podcast about your actual neighborhood, but now there is. So that excites me and what we see in the very early signs are encouraging. This is something that is new to people. I think those are the ones that are near term and exciting. Longer term, I was also, obviously, I'd be remiss if I didn't say that our remix and covers, I think, is an incredibly exciting product, again, because there is no one else that can really do this. Normal generative music will happen with or without us. This product will not happen without us and it needs to exist so that existing artists can participate in this. So that's very exciting to me. It's a lot of work. It's going to take more time, but I'm very excited about what we're seeing internally. Lastly, I would say the longer term, back to systems thinking is really what you can see in 'Talk to Spotify.' So if you've been lucky enough to be in the U.S., we started rolling this out and try it; you can literally talk to Spotify now and you get very good answers about the music, about who played in what band, when they're touring, what the song is about. These are better answers than you get from any general-purpose LLM, certainly within our domains, about podcasts and books and music. So without stretching it too far, I feel like Spotify is coming alive and you can literally start talking to it. It was a long answer, but that's what I'm excited about.
All right. We'll go on to the next question. And another one from Jessica Ehrlich on advertising. Advertising growth in the past year has been subdued despite programmatic now comprising over 25% of ad-supported revenue. Can you provide an outlook for the coming year or years? And what can you do to drive momentum in what is a very high-margin business that monetizes your engaged and growing base of users?
Jessica, this is Alex. I'll start. I actually take the opportunity to organize this a little bit differently because I do think that a similar question is coming further down the line from Rich Greenfield. So I'll try to do both at the same time. I'll speak mostly to why investors should have confidence in us when it comes to the ad sales business of Spotify and then maybe Christian can talk about the margin side of things. I want to pull the lens back a little bit and talk about the past two years here. I've said many times now that we have been in a transitional phase with ad sales. And at the beginning of this year, we completed that transition. We now are entirely on the new ad stack that is proprietary and built inside the walls of Spotify. I think 99% of all the impressions that we serve are now on our own ad stack. But the big thing is the addition of the biddable exchange that we put in place and the automated sales channels. These are now almost 40% in Q2, up from 30%. And if you ask me, it's even going up further from here. So the two things to think about are supply and demand. We'll start with supply. The supply picture has actually never been stronger. It's not only about user growth and reach, which is obvious because we grow and then reach sort of follows, but it's also about the new placements we have in the free tier. It's also about us launching personalized ad load, and it's also about the depth of the engagement that we have. So this, in turn, drives a lot of supply for us to match the demand against. Now on the demand side, for the longest time, since the inception of Spotify really up until two years ago, the way to buy ads on Spotify was the brand would call us, literally call us or contact us via e-mail, and they would buy fixed and guaranteed campaigns from Spotify. Now that's all good, but it's also capping us in several ways. One is it's really capping us in terms of pricing and sell-through. So when the inventory is bought, it's bought. And it's also capping us, obviously, in the ways that people want to buy. Not everyone wants to write e-mails, sign I/Os and make calls. They further wanted to automate the buying as well. So what we have put in place now actually uncaps this, unlocks both of these things. One, a lot of it is self-serve and automated. The other thing is that now people that find our ad inventory valuable can go in and bid up the pricing. So it unlocks really both of these things. And the bottom line is really that this has led to not just existing advertisers moving over to this new way of buying, but it's also that we have increased the active advertisers. We are now at 33,000 active advertisers. This is 60% up from last year, which is great. So it's really the right plan, and we need to just keep working on it. A funny tidbit is that we recently launched plug-ins and MCP towards Claude, ChatGPT as well as Gemini and people are now actually prompting to create campaigns and audio assets. I think out of the 33,000 active advertisers, we now have 7,000 of them using our AI audio asset creation tool, which just makes it easier for brands to buy on Spotify. So there's a lot of change going on, but I'll just reiterate: there are three reasons that people come to Spotify to buy ads. It's a beloved brand. It's high-quality content, and it's high user engagement.
So Christian here, just on the margin side to fill in on that. One of the things Alex brought up is very important to understand. When we move to an automated sales channel and we also have self-serve, it becomes a scale business different from before. So as we scale the ad monetization, both from music and podcast, we will also be able to improve our profitability. And as we said at our Investor Day, we believe that we can move from the roughly 20% range we have today towards 40% over time. And music, as we scale in emerging markets, should see healthy margin flow-through on ad sales. So that's what I wanted to add on the margin side.
Okay. Our next question is going to come from Rich Greenfield on music add-ons. Are you surprised more artists have not signed on to the AI music tier to enable you to launch? What is stopping artists from participating?
I'll start and maybe Gustav, you can jump in. Following our agreement with UMG in May, we announced today the deal with Merlin, which adds 30,000 labels in Merlin's network the opportunity to partake in this new product that we're building around covers and remixes. I think it's worth repeating what we said during Investor Day. What we're trying to do is very considerate and planned out. So we call it the three C's at Spotify. First of all, we are looking for consent. We want artists to consent their work into this catalog so people can play around with covers and remixes based on their art. We also obviously want to give them credit. And last but not least, this is about compensation to the labels and publishers and artists and songwriters. We not only have the consent and give credit, but we also drive compensation for this. So really, we're talking about the first legal way to partake in this AI tailwind that we see coming for interactive music basically.
And I would just say we have really strong momentum there. One thing that I think is important to remember is that while you can see skepticism around net-new artificial music by many people out there, what we're doing is different and artists see that. Our products are about real artists, not fake artists. And in the case of remix, real artists with real voices. So you're listening to real people. That's a very different proposition, which is why we're focusing on this. This is the thing that doesn't exist today, and artists remain excited about that and consumers remain excited about that. But it is an ever-changing landscape. I would say what to expect from us is a research preview coming out of our model where we start to give this to some users. And for that, we do not require a full catalog. Why is this important for us? This is not a test of the product. It is the thing that guarantees the product will be very good. For those of you who know about machine learning, one of the most important phases is the post-training using reinforcement learning. So what we will do is we will allow people who are fans of a certain artist to start making remixes with songs from that artist. And what they do is say, 'I think this remix was better than that remix.' And then we get the preference data that actually makes the model better. And this is our unique advantage in this business. We have 777 million people and music fans to do reinforcement learning with, which is why we think we're very well positioned in this business. So that's kind of what you should expect as the next step. We're going to launch the product when it's ready, and we think it's got the right appeal to consumers.
Okay. Our next question is going to come from Justin Patterson on AI tooling. Over the course of 2026, we've seen meaningful momentum in open source. How are you thinking about the costs and benefits of deploying open source more broadly versus how you're using Anthropic, Claude and other models today?
Yes, I'll go here as well. Of course, the open source movement is very helpful for us, and we talked about this: the Large Taste Model is based on open source. It is using what is called continuous pretraining, where you take a model that was trained as an open source model and you continue to train it on your proprietary data. This is what I talked about that is performing really well for us. So the fact that these models appear more often and are much better and cheaper at the same time is obviously very helpful for the consumer product. But it's also helpful for our development environment and our developer costs, as I said. We've built this internal product called Chirp that we will also offer to other companies because they're asking for the same thing, where you can seamlessly switch between one of the paid models and a hosted open source model, keeping the context of the project that you are in. This is why it's important for us that we control the context of the developments we're doing so that we can always go towards the best price performance. Also, I think you can all see that this is putting pressure on pricing. And there are lots of announcements of prices coming down per token, which is obviously very helpful for us. But on a constant per-feature quality level, it's pretty clear that the costs are coming down for a certain level of quality. That doesn't mean that we won't use more advanced models, but for a certain feature level of quality, the costs are coming down quite fast.
Okay. Our next question is going to come from Batya Levi on MAU. What's informing your guidance for slower MAU growth in the third quarter? Do you see a change in the competitive environment or the general intake for your campaigns? And if I may, I see Benjamin Black had a question about some initiatives to drive MAU growth in the future as well, so you might want to address that also.
Yes, I will. Thanks, Batya and Benjamin. Just to hijack this for a second, you don't see us here, but Gustav and I are in Stockholm. The sun is shining, and we're happy about the fundamentals of Spotify. It's really in a good place. We just hit 300 million subscribers, and we're super elated to be in this sort of rarefied air, and it just keeps growing. On the question of MAU, we've had a few years of outperformance in MAU. And more recently, in the past four or five quarters, we've had outperformance relating specifically to emerging markets. I think we've pointed that out in past earnings calls. And of course, emerging markets includes countries like India and Indonesia and so on. And as a reminder, these are very populous nations. So they're a very lucrative opportunity for us. What you're seeing us do now is that we're making changes to the product and the value proposition and strategy in these markets. We've made changes like tweaking the sign-up to get higher-quality MAU throughput. We've deprecated lower-end Android device support, which builds the business and makes it more efficient for us. We've carefully introduced some friction in both ad load and some limitations in our free tier. All of these things are positioning us for more monetization. Gustav and I have a saying where we say sometimes we pull the growth lever and sometimes we pull the monetization lever. Now here, we're starting to pull the monetization lever. So our belief is that this is going to take some time, but the shape of the growth curve follows other markets. The way to think about it is really when you start out in a market, you work on getting product-market fit and get some MAUs. Slowly and surely, that MAU growth will increase. Then you have a base to convert from. Then you have some subscribers come in, there's some conversion coming. And then as you calibrate the product and value proposition, that growth then continues to become something like Latin America, which also started out with very low conversion but massive MAU growth. So our planning here is very considerate. And a consequential point here is that this will not affect subscriber growth in the near term.
All right. Our next question is from Rich Greenfield on Reserved. You started a meaningful ad campaign for Reserved ticketing after Role Model. How many concert tours have you worked with since? Are you starting to see an uptick in conversion to paid tiers to access Reserved? And how do you see Spotify's positioning in the broader live event ecosystem evolving?
That's a good question. Thank you, Richard. Whenever I get to talk to users about Reserved, we get so much praise. We get praised, Live Nation gets praised for the partnership with us and the way we're shaping this feature. Just as a reminder, the big idea with Reserved is that the biggest fans get access to tickets. Secondly, artists get to have the biggest fans in the room. And third, Spotify gets to have unique value for our premium subscribers, which obviously gives us differentiation and increases value-to-price perception for Spotify. It really is like a triple win. It's still early, to your point on how many tours we've done. It's been a few and it's U.S.-only for Premium right now. So the rough numbers are that we've had about 100,000 tickets reserved. In some instances, we've blown through the allocation and Live Nation has even upsized them mid-run. Obviously, I'm looking forward to many more tours and more markets. As far as monetization goes, this is currently about increasing the value-to-price ratio on Premium.
All right. Our next question is going to come from Jason Bazinet on music add-ons. You now have AI music deals with UMG and Merlin. Do you need deals with all the majors before you launch an AI service? Why or why not?
So I can start there. The answer is we do not need a deal with all the majors. We would like to have as many artists as possible, obviously, but we don't expect to have all artists. If you go back to the beginning of Spotify, Spotify started without significant parts of the catalog. It took many years before big acts were on the service. So you don't need a full catalog to start. Of course, we want as many as possible because that's better for consumers and for creators. But as I said, what you should expect is that we start improving this product in public as a research preview so that we can start getting the preference data that automatically improves our model. That would be the next step. And then we'll decide when we launch based on where we are. There's more to come, but very strong momentum.
All right. Our next question comes from Eric Sheridan on monetization. On the heels of Q2 '26 Premium subscriber growth reaching 300 million, how should investors think about the scope for monetization efforts as a result of more interactive tools and live event integrations across the subscriber base?
I'll take that. You're basically baking in an opportunity and TAM question and also sort of a pricing and product question. We've reached this 300 million mark, which is rarefied air. There aren't many other companies that have built one product and have 300 million recurring customers come back again and again every month. So we're now close to four percent of the world's population. We've said before that maybe we'll not reach 90% of the world, but it's not impossible that we'll get to 15% penetration of the world. Our most important line of work here for Gustav and I is actually to drive the value-to-price ratio upwards. Meaning we need to keep building interactive tools like you talked about, including covers and remixes, but also some of the more recent ones like taste profile, personalized podcasts and so on. Audiobooks is a great example of an add-on on Spotify that generates more structural ARPU increase than just price increases. ARPU can change by way of price increases, but when we have success with an add-on like this, it will structurally increase ARPU in a different way. So in the price-to-quantity equation, we're providing a different type of price increase than just moving the price of Premium upwards. We do think the opportunity ahead of us is still immense.
Okay. Our next question comes from Justin Patterson on time spent. In 2025, you streamed 211 billion hours of content versus Netflix's 191 billion. As you expand new formats like fitness and deepen personalization capabilities across the platform, how do you believe time spent can evolve on Spotify?
It's a good question, Justin. You're pointing to Netflix and sometimes we also get a similar comparison with YouTube. I think the important comparison and observation to make here is that we're very different from these other streamers. When it comes to engagement, our engagement typically spans many more devices than our friends here in the industry, whereas most of them are big screen and maybe small screen. We are speakers, we're gaming consoles, we are cars, we are contexts like sleep, we are contexts like studying, where you actually find Spotify much more compatible. So you have the context universe spread out over our billions of hours. The one metric that we pay special attention to when it comes to engagement is active days. We've talked before that we have over 100 million of our subscribers spending more than 20 days in a month with us. The reason we track active days is that we find it to correlate very well with lifetime value. So we've seen that increase again and again. Even this quarter, active days in a month has increased overall for premium subscribers on Spotify.
All right. Our next question is from Deepak Mathivanan on AI products. You've launched several AI-powered products in the last six months, including Prompted Playlists. Can you discuss what types of benefits you're seeing in the KPIs such as listening hours, conversion rates or churn?
Sure. Thanks, Deepak. In general, AI-powered experiences now reach about one quarter of our active users. This year's launches—Talk to Spotify, Studio, Personal Podcast, Prompted Playlists—are scaling really fast. So they've already reached more than 25% adoption among active users. Prompted Playlist specifically has reached 14 million of the 100 million actives we've rolled it out to. In terms of what we're seeing, I shared a bunch of metrics in my prepared remarks around the Large Taste Model, where we are moving some of the metrics that are absolutely the hardest to move, which are active days. Related to viewing hours and engagement, we focus on active days as the most important metric to try to drive rather than only engagement in the moment. We are seeing those effects from the Taste Model. Overall, since 2010 when we started investing in personalization, the correlation between personalization and retention has been very clear for us.
Okay. Our next question comes from Doug Anmuth on third quarter operating expenses. Can you talk about the drivers of the implied acceleration in operating expenses in Q3?
Thank you, Doug. I understand there could be a little confusion about our expense guidance for this year. We have said we're going to elevate operating expense by approximately EUR 200 million for this year—something that is very much in control and non-structural. Just to give a little better guidance on it, excluding currency and social charges, we see our expense growth in quarter three to be roughly consistent with quarter two. So it's going to be pretty much the same growth. I think it's important to be careful with the rounding nature of our guidance. We talk about EUR 5 billion and add growth percentages, and we also assume ARPU consistency between the quarters. We're well positioned to moderate this into quarter four and to keep it consistent into quarter three. This is a nonstructural cost change; personnel levels are the same. We are doing this investment to increase engagement and LTV and in marketing. We have a lot of new features coming out, and we are also boosting a bit on our AI, as we talked about at Investor Day. So nothing strange, nothing new, and we are keeping this consistent with what we have been thinking all along from the beginning of the year. We are in full control. So excluding currency and social charges, we see our expense growth to be roughly consistent in quarter three with quarter two growth.
And I would just chime in: we started this year by investing to make sure that we were leading in this wave of using AI. We are leading according to other companies. Now we're starting to focus on cost and efficiency, which is always the second step. This is what I talked about when we talked about Chirp, where we have control over our spend and can choose models. So we are sticking to our cost moderation in Q4. It's also important to remind everyone that these costs are compute and marketing. They are within our control. These are not long-term capital commitments that need years to roll back.
All right. We are coming up on the hour, so we've got time for a few more questions, and we've got a follow-up from Doug Anmuth on product tiers and ARPU. Can you help us understand the timing of add-on tier rollouts across verticals? And how should we think about their impact on Premium ARPU in the second half of this year and into 2027?
Doug, Alex here. We don't comment on timing for launches and we don't give guidance on ARPU either. But I can comment qualitatively. You heard Gustav talk about the different products we've been rolling out and how the usage-driven paradigm is an important approach for us and how we convert from free to premium and from premium to add-ons. The one example I want to share is that Audiobooks+, since we last shared a number on that, has doubled, which is great. It keeps growing and this is just in a select few markets. So obviously, that introduces another type of change to ARPU. ARPU can change by price increases, but when we have success with an add-on like this, it will structurally increase ARPU in a different way.
Okay. And our next question is going to come from Jason Helfstein on the pricing environment. Is there any read-through from the Apple Music price increase that was announced in July? And how does this impact your thinking about price increases going forward?
We don't comment on other companies' price increases. But what it does show is that this is an example of continued value in music streaming services, and we're happy about that for the ecosystem. Our category leadership in not just user scale and subscriber scale, but also our leadership in engagement keeps being our focus. Of course, the more value we deliver, the more pricing power we will have. As of right now, we're happy to be the price leader in our category.
Okay. And our last question is going to come from Jason Bazinet on the subscriber opportunity. You had solid Premium net adds this quarter, but the record labels suggested the industry's Premium growth slowed in the second quarter. How confident are you in the long-term growth of Premium subscribers for the industry and Spotify?
I'm very confident. We don't give long-term guidance, but what we're optimizing for is a healthy funnel. This not only builds our business, it's, of course, aimed at helping our subscriber growth and not slowing it down. So very confident about the long-term growth.
All right. Great. So that concludes our Q&A session. Thank you, everyone, for the questions. And it also concludes today's call. A replay of the call will be available on our website and also on the Spotify app under Spotify Earnings Call Replays. Thanks again, everyone, for joining.
This concludes today's conference call. Thank you for joining. You may now disconnect.