Friends at Index Exchange for making this conversation happen. I've been covering Spotify now for a decade and it's one of the most interesting companies because they have conviction. They have never given up on their vision and Wall Street is definitely starting to take notice. The rally on Spotify stock has been very impressive. They're a great partner for us at this event because we're trying to talk to innovators and we'll be here all week talking to more. We just had Colin Jost on stage. Tomorrow we're going to have Alex Rodriguez, Maria Sharapova, Lando Norris. So, definitely come back to the yacht and follow all of our coverage on axios.com. So, without further ado, let me introduce the newly branded co-CEO of Spotify, Gustav Söderström.
Hey everyone, it's great to be here.
Gustav, remind everyone, you just got named co-CEO with Alex, but how long have you been with Spotify?
So, I've been with Spotify for almost 18 years since 2008, 2009. I started heading up mobile product development for Spotify then took over all the product development and then all the product and technology. I kind of still have that responsibility so the division of labor between me and Alex is sort of I run product and technology, he runs business and licensing. That's not really true. He's a product guy. He loves product. I'm a business guy so we help each other out but it's a great partnership.
So, having been there now for 18 years, how do you feel now that Wall Street and by the way, Swedish company but you trade publicly. How do you feel that Wall Street's finally kind of getting it and boosting the stock?
Well, I think Spotify went public in 2018, I think. And it is a tricky company for Wall Street to understand because from the surface it's a catalog of content that other people can license and people call us like a DSP and I hate when people call us a DSP.
Because we're something that lives here, right? We're a brand that matters. We do something for you as a consumer. We don't just distribute content. We don't think of it as content. We think of it as art, music, creators. So, I think it's been a brand that when you look at it at a surface level, you kind of ask, well, how could they possibly compete with these giant tech companies that can pre-install their own products on their hardware and so forth. But then like 20 years later, we are still the biggest and we managed to not only compete, but actually gain market share. So, we've managed to compete on building an experience and a product that really matters to people, that is deeply personal and has incredible retention. It's a combination of a couple of different things. We focused on building a freemium product and free tier. We focused on personalization and what is now called AI very, very early. And we also focus on ubiquity, trying to be everywhere and partner with everyone, whereas obviously our biggest competitors focus on their own ecosystems. So, it's taken a while for Wall Street to get their heads around that. And now that our investor day that you very kindly joined a couple of weeks ago, Alex and I laid out our plan and vision for the company. And Wall Street seems to like that.
So, a question for you on those plans. One of the things that you teased is that your big differentiator is going to be that you have a large taste model, not a large language model, a large taste model. Define that. Explain it. What is it?
Yeah, so one question that we get and I'm sure all companies get is AI a headwind or a tailwind and what does AI mean for you? And the way we think about it is that there is general reasoning, the ability to do math or write code that we can buy, just best of breed from Anthropic, OpenAI, Google, etc. So we buy that and we just try to buy the best performance to price that we can. But then when our consumer experience, we also want to do something that is unique. And what is unique to Spotify is all the taste data that we have. So, one way to think about large language models is that when you train them, you train them on sentences, right? You show them sentences of words and you ask it to predict the missing word. That's what a large taste model is. It turns out that you can do that for things that are not English as well. You can do the songs that you play on Spotify and then predict the next song or the podcast you listen to on Spotify and predict the next podcast. So, we take these LLMs and we train them on our proprietary listening data, so sequences of user behavior, literally trillions of them. And it turns out, just as these large language models get very good at writing code, they get very, very good at predicting taste, what you would like to listen to, what you're most likely to listen to. But, what is also very interesting with a large taste model that sort of speaks English, but can speak the language of music and podcast and swipes and saves and clicks, is that when you ask it, like, so here's Gustav's listening history. Now, what is the most likely next thing for him to listen to? We can do that. It's called a recommendation. But, we could also let me, Gustav, put in a few words there saying like, actually I'm a little bit bored with electronic dance music. I would like you to go a little bit left here. So, we can let the user input in plain English. So, this notion of giving people control over the algorithm is something that is very new that we are very passionate about. And I think it's a counter to the very negative narrative about AI, how it's all just going to lead to more and more addictive algorithms and less and less user agency and more control over the user. We've chosen to say, no, let's use this to give back control to the user. So, we launched something called a taste profile, which literally lets you, the consumer, see who we think you are in plain English. This is who we think you are musically, podcast-wise, audiobooks-wise. But then you actually get to disagree and say like, maybe that's who I am, but that's not who I want to be. I want to get back into classical. I want to read sci-fi books from these authors. And you actually program the large taste model yourself. And this is something that is unique to us, very hard to replicate unless you have almost 800 million users.
You hear about this problem all the time when Wrapped comes around and parents say, all of my Wrapped is just little kids music and I want to get back to my music. One of the other things you mentioned on Investor Day was that because it's not just about maintaining the ethos of the app, it's also about making sure that you're driving a profitable and growing business is that you want to add more tiered products, right? Now you have free, you have premium. What does that mean? Does that mean there are going to be in between products or are we going to do ultra lux premium?
Yeah, so what we talked about at our Investor Day was that we said that one of the principles we use is that the world is a power law. What do I mean with that? Well, you always have this price demand curve, right? Where you have users on one side and willingness to pay here. And people imagine it's a line like this. If you raise the price, you get fewer users. It's not actually a line. It's a logarithmic scale. It's a power law like this, which means the tail of those users, they don't want to pay that much and we cover them with our advertising tier. You can pay somewhere between zero up to the 9.99 of premium with using ads. Then when you feel it's worth 9.99, you can pay for premium and we have a bunch of offerings there: student, duo, family plan. They have Spotify premium, but that used to be it. If you actually wanted more out of Spotify, that wasn't possible. So the head of this power law, we never covered and there is a lot of willingness to pay there. So the first example where we took this from a theory to reality was with audiobooks. So with audiobooks, we can include some audiobooks in the free tier. We include 15 hours of audiobooks in the premium tier. But what if you want more than 15 hours? Well, now you can add a subscription called audiobooks plus to Spotify. And it turns out that some people, including my wife here, listen to like a hundred hours of audiobooks per month, right?
So now we found a way to capture not just the long tail, not just the shoulder, but actually the head of this power law. And we think this is repeatable for music interactivity. For example, if you want to remix and make a cover of a song, we think that's an add-on tier. We talked about that at Investor Day, how we're now training the world's first truly legal generative music model on legally licensed content, artists can obtain their catalog and allow their songs to be remixed. So for the first time ever, fans can actually play with their favorite artists' music. And when someone listens to that, so I can make a remix, I can send it out. If it gets popular, that artist actually gets the royalties. So now existing artists can participate in AI instead of just being replaced by AI.
What about a premium or like a tiered offering in podcasts? I know you were talking about AI-generated podcasts. Like what is that going to look like?
So we announced a couple of things at Investor Day. One was creator memberships, which is something podcasters have asked us for. So we help them with advertising if they want to. And if you're in premium, we actually pay out from our premium tier called Spotify Partnership Program. We just announced dynamic sponsorships in premium. But there are creators who want to work with their super fans, right? This power law again. So we just announced that we're rolling out a creator memberships where creator can have a club of Spotify users that they can charge for access to whatever they want, extra content for example. What is a little bit different with Spotify though is normally if you're a platform, you try to disintermediate the creator and keep these users secret, who they actually are. We're actually allowing the creator to know who these people are, their emails. So it really becomes their super fans, which I think is quite unique and much more creator-friendly than what other platforms do.
You have made a lot of news in the past few weeks. Most notably, you announced Reserved, which is a ticketing opportunity. Basically, you are going to allow premium users to get access to tickets on your platform. There were reports that you outbid Apple and Amazon many millions of dollars for access to the Live Nation rights to do that. You're obviously putting resources in to get this off the ground and make it work. What kind of adoption do you need to make this worth it? And when are you going to make that decision of like, all right, this is worth it to keep moving forward or maybe this isn't going to work?
So, the way we think about Reserved is we looked at our creators and musicians specifically in this case. And for many of them, touring is actually as big or for some even bigger part of the revenue than streaming royalties, right? Especially early in their careers. It's very, very important. So we decided to help them sell tickets a long time ago. And we made this deal, so if you have Spotify, you've probably gotten a notification saying there's an upcoming concert close to you. And we saw that we could do a very good job of this, which is not surprising because we know literally there is a person that is the biggest Taylor Swift fan in the world, and we actually know who it is. Right?
Because it's a stream count.
So, wow. And you've never revealed that person to any privacy? You can't.
But we actually know who the biggest fan is. And it you know,
