The Company Apple Couldn't Kill | Spotify Co-CEO Gustav Söderström

Summary

• Spotify solved CEO succession with a three-year live rehearsal: Daniel Ek made Söderström and Alex Norström co-presidents, gradually handing them the day-to-day, week-to-week, and month-to-month business with the full P&L and balance sheet before handing over the CEO roles. Only PR, government, and being the public face were new. Söderström’s read on Ek: Americans “can mistake humbleness for weakness. It’s the opposite. He’s insanely tenacious.”
• The new co-CEOs deliberately reversed Ek’s operating style, replacing his one-on-one “star pattern” with a single weekly three-hour E-Team meeting of ~14 SVPs where “let’s take it offline” is banned. It burns expensive leadership time and contrasts with Elon’s leave-irrelevant-meetings rule, but produces 14 people with “the full CEO perspective” — insurance against the super-app’s biggest risk, which is “to ship the org chart.”
• Spotify’s survival against Apple rested on three deliberate counter-positions: freemium, because Apple struggled to make advertising work; personalization, because Apple was “almost” against data; and ubiquity on non-Apple hardware. When Jobs bought Beats and brought in Iovine and Dre, “they gave us six months” was said in Apple’s internal meetings. Söderström says Apple still struggles with free offerings and non-Apple hardware, and cautiously says Spotify may be better at personalization.
• The strategic thesis behind the super app is that distribution, not product quality, is the scarce resource. When Spotify entered podcasts there were at least ~7 good podcast apps in the App Store, yet Apple Podcasts still had ~98.5% of usage — “the problem was that they didn’t get any distribution.” Spotify took the pain of one complex app to reach 300-something million existing users: “it’s just software.”
• “No Regrets” became Spotify’s strategy, backed by anonymous third-party surveys: Gen Z valued almost 90% of its Spotify time, while on many big engagement platforms young users regretted almost 60% or more of their time — “No, I’m trapped.” Because almost 90% of revenue is subscription, Söderström argues Spotify is rewarded for felt value rather than engagement, which is why it shipped the anti-engagement video-off option and is investing in fitness, including AI running playlists previewed for Investor Day.
• Söderström’s AI bet is contrarian: give users control over the algorithm rather than build “the most addictive algorithm you have ever heard of.” Generative AI means “finally, computers understand English” — potentially enabling deep user research with all 761 million users instead of ten. His own agent-built private podcast filters “rage bait, clickbait, politics”; he calls this “premeditated media,” and says the 1/9/90 power law lets highly engaged users improve the experience for less-engaged people with similar tastes.
• Long tenure is his answer to why Apple’s functional org avoided Yahoo-style politics — and Spotify adopts it while mitigating the tax. The average in his organization is maybe 7–8 years, with many people at 14–15. Mitigations include the Rising Stars internal-MBA program and acquisitions such as Echo Nest for deep-learning talent near MIT and Sonantic for cheap voice, bought before LLMs could write the scripts. His explicitly far-fetched hypothesis on Siri is that Apple was “slightly too old to see” AI and should have hired senior talent sooner.
• What keeps him up at night: he thinks we’re in a macro-wave era, not calm extrapolation waters — extrapolating 2015→2025 would have worked, while extrapolating 2005→2015 “would have missed everything.” Spotify grew most during periods of change, so his principle is “always be first.” The AI era’s Ubers and Airbnbs “maybe they are OpenAI and Anthropic, but maybe not. Maybe it hasn’t even happened yet.”

Deep dive

1. An 18-year apprenticeship: how Ek engineered the handoff

• Senra opens with a text from Daniel Ek himself — “ask him how he prepared for the CEO role, we did something quite unusual.” Söderström’s answer: about three years ago Ek made him and Alex Norström co-presidents running the day-to-day, “gradually giving us more and more rope” — first day-to-day, then week-to-week, then month-to-month, with the full P&L and balance sheet. What was genuinely new at the handoff was only Ek’s solo portfolio: PR, government, and being the public face.
• The delegation goes back to the start: in 2008 or 2009, Ek sent a brand-new Söderström alone to set up a deep Facebook integration with Zuckerberg — “that’s a lot of trust for someone who’s new.” It’s why he stayed almost 18 years: “you get a new job almost every year,” and “I don’t think I’ll work for anyone except Daniel again.”
• His portrait of Ek is the episode’s sharpest character study: not “the typical alpha-male founder,” which is precisely why he could get big personalities across music labels, book publishers, and podcast publishers to do something they did not want to do individually or collectively — through logic and non-threat, not force. “Most Americans specifically can mistake humbleness for weakness. It’s the opposite. He’s insanely tenacious, insanely tenacious, just never gives up.”

2. Synchronized swimming: banning “let’s take it offline”

• Ek talks in a “star pattern” — separate one-on-ones with each function, which meant Söderström and Norström would compare notes and find “that’s not what he said to me.” The co-CEOs chose the opposite: no separate leadership meetings, one three-hour Tuesday E-Team with ~14 SVPs spanning marketing, ads, subscriptions, and all product and technology functions, where nobody is ever allowed to say “let’s take it offline” because licensing is “right there across the table.”
• He’s explicit about the cost — 14 or 15 people times three hours weekly of leadership time, with conversations often not immediately relevant — and reframes relevance as timescale: the experience person absorbing ad-stack economics “is going to affect their product decisions in the future… guess what? There are ads in the product. I think of it as investment.”
• The contrarian edge is acknowledged head-on: Elon says leave any meeting the moment it’s not relevant, “and like I said, he’s not wrong. He’s built some of the most successful companies on Earth. It’s just different approaches.” The payoff for Spotify: 14 people with “the full CEO perspective” in a business where dependencies between licensors, contracts, and product make divide-and-conquer hard — “our biggest risk is to ship the org chart.”

3. There is no right org — only choosing which dimension to suck at

• The Sinofsky-derived thesis: “you can’t win. There is no organizational model. There are many, and they all work” — Amazon, Apple, and Elon’s companies have all succeeded with different structures, so pick one that fits your personality. Senra’s supporting pair: Larry Ellison credits Oracle to keeping the core product kernel together for roughly a decade and a half or two decades, while Elon — who takes advice from Ellison — wants maximum fresh blood. Both are highly successful.
• The operational corollary: “figure out what’s important and then optimize for that and just accept that you’re going to be average at the others.” The worst outcome is being great at the unimportant thing. Spotify chose the single experience — a product where one click “triggers all kinds of different business models” (a royalty pool for music, ads for podcasts, and another structure for books), so “on the back end, it’s insanely complicated.”
• His Apple puzzle: on paper, a functional org should dissolve into Yahoo — where he worked through what he thinks was four CEOs in roughly two and a half years while “all the EVPs fought each other to the death.” The only answer he’s found talking to Apple people is tenure: long-serving functional leads with enough trust that sometimes hardware leads and software follows, and sometimes the reverse. “Without the tenure,” functional orgs “just dissolve into politics.” At Spotify, the average is maybe 7–8 years; many of his reports have worked for him 14–15.

4. Distribution, not product, is the scarce resource

• The strategic lens behind the super app: App Store discovery stopped working and Facebook install ads went away as distribution channels, so “the biggest challenge in the world is going to be distribution.” When Spotify weighed podcasts, there were at least 7 really good podcast apps, including Overcast — and Apple Podcasts still held ~98.5% of usage. “The problem wasn’t that there wasn’t a good enough podcast product. The problem was that they didn’t get any distribution.”
• So Spotify chose the pain of one app — organizationally harder and experientially complex — for the benefit of reaching what was then 300-something million users: “we chose pain and cost somewhere to get benefit somewhere else, and we said, ‘It’s just software.’” Audiobooks followed the same logic, with extra conviction because Sweden’s advanced media habits already showed audiobooks as a big market.

5. “No Regrets”: the survey that became the strategy

• Spotify anonymously surveys users through third parties across the big media platforms, asking both “how do you feel about the time you spent?” and the inverse — how much did you regret? Among the platforms surveyed, Spotify came back as the lowest-regret content on the internet, with Gen Z valuing almost 90% of its time there. The shock was the other side: on many big platforms, young people regretted almost 60% or more of their time. “I kind of mistakenly thought that they were there because they wanted to be there… Turns out when you ask them, they’re like, ‘No, I’m trapped.’”
• The business-model alignment is the load-bearing argument: almost 90% of Spotify revenue is subscription, and “when you vote with your wallet… you’re going to pay for the value you feel,” whereas “in a pure advertising-based model, your incentives are clearly to maximize time spent at any cost.” He hedges the origin honestly — “maybe we lucked into that business model, maybe it was semi-deliberate” — but is now doubling down.
• The strategy’s next expression is fitness — almost 70% of Spotify users work out with it — previewed ahead of Investor Day: tell Spotify, “I want running playlists for an eight-minute mile… on the downbeat,” and it estimates roughly 165 steps per minute for an average-height runner, finds tracks around 160 or 80 BPM, speeds or slows others to match, beat-matches the transitions, and overlays interval commentary. The premise is that people generally do not come out of an hour at the gym or on a run regretting the time.

6. Principles only count when they cost you engagement

• Söderström’s caveat to his own virtue: “everyone convinces themselves that they’re doing good for the world, even the ones who don’t.” Senra’s specimen — a prediction market pitched him a partnership as “a maximum truth-seeking machine” while its homepage led with who wins the Patriots coin flip: “If you’re just saying we want to gamble and make a ton of money, I respect that. But don’t give me this other bullshit.” His frame is the Godfather’s Sollozzo scene: no judgment about how another man makes his living, but not interested.
• Spotify’s proof of costs: as podcasts went video, parents specifically felt their time-well-spent degrading, so Spotify made “the anti-engagement decision” to let anyone turn video off entirely. “If you’re going to have these principles, you kind of have to live with them, and they’re going to have consequences.”

7. From Deep Q-Networks to “computers finally understand English”

• His AI arc as told: neural networks could not do much in late-1990s university because computers were not powerful enough; at Spotify in 2009, Erik Bernhardsson was running collaborative filtering on “the world’s largest Hadoop clusters”; DeepMind’s Deep Q-Network beating Atari “looked like a toy back then, like everything does” but showed the scaling; he thinks he read “Attention Is All You Need” within the first few days after its June 2017 release and “was completely blown away,” after coding his own RNNs, LSTMs, and GRUs on vacations.
• The Sonantic acquisition captures the operating principle: buying cheap voice synthesis before LLMs could even write the scripts, because serving 700 million users a few minutes of expensive voice per day “you’re going to go bankrupt” — “we just bet on intercepting that curve instead of waiting for it.”
• His best product framing: “finally, computers understand English.” Where about 1 million developers on GitHub could talk to computers, now everyone can — so instead of user-research teams studying 10 users and extrapolating, “what if you could do that deep user research with every single user, all 761 million of them, all the time?”
• The personal driver is metaphysical, told at length: he recalls being told that 99% of your atoms swap out every seven years, leading him to ask whether you are your atoms; London cabbies’ brains develop overdeveloped 3D-navigation centers, leading him to conclude you are not even a static structure — “you literally are your thoughts. Literally, not as an analogy.” Spotify, he admits, “is sort of my excuse to get as close as possible to the philosophical side of AI.”

8. The Apple war: three counter-positions and no exit

• The setup as he lived it: a small Swedish startup watches “the biggest, most respected product company in the world, and the person you look up to the most, Steve Jobs,” buy Beats and bring in Dre and Iovine. “‘We’re going to kill you.’ That’s what they literally said in their internal meetings. ‘They gave us six months.’ Those were hard times.” Senra adds color from his friendship with Iovine — Jimmy still swears “Apple held me back.”
• The strategy was three deliberate counter-positions to Apple: freemium, because Apple’s control-heavy iAd effort did not work; personalization, because Apple was, “I would almost say, against data”; and ubiquity on Samsung TVs and Android. His scorecard today: Apple pulled back its free radio tier, “still struggle[s] to be good on non-Apple hardware,” and, with personalization, “I want to say we’re better.”
• The non-structural reason, per Senra’s Josh Kushner quote — bet on who wants it most: “If Apple Music didn’t work out, they would’ve been fine.” Söderström agrees: “for us, this was not optional. We either died or we did this.” Senra’s Game of Thrones maxim as the bow: “those on the margins often come to control the center.”

9. Premeditated media: the agent that filters rage bait

• The recently shipped Save to Spotify lets agents in Claude Code, Codex, or similar tools research topics, write and generate a podcast, and upload it privately. Söderström’s own version merges personal documents and interests with what Spotify knows from his listening into a deeply personal update about what is happening in the world — which kept resurfacing Senra’s Evan Spiegel episode with commentary that there was “product wisdom in there.” Both reject judging Spiegel by market capitalization alone; Senra cites four roughly 75% declines for Steve Jobs as an example of what can happen over a career.
• The instruction set is the point: “I literally ask my agent to filter for rage bait, filter for clickbait, filter for politics.” He confesses his own capture — hours-long road-rage rabbit holes — and now pre-commits: “Don’t show me road rage.” His coinage: “premeditated media” — deciding your own future consumption ahead of the moment when “someone puts candy in front of you.”
• Senra’s pushback — don’t most of the 700 million just want it done for them? Söderström concedes via the 1/9/90 rule and Spotify’s own history: roughly 10 billion user playlists, built by a heavily skewed minority, powered recommendations for everyone. The engaged few can improve Spotify for themselves, while less-engaged users with similar tastes can benefit without doing all the work.

10. Truth-tellers, tenure taxes, and the Rising Stars fix

• Senra recounts what he says was Ek’s story about Söderström telling him, “you’re running product and you’re not good at it” — Ek’s initial reaction was “I have to fire this guy,” followed by the realization that Söderström had done him a service. The episode gets its mirror: Söderström keeps long-tenured people who have learned that shouting “No, I think that’s a stupid idea” carries no consequences — “that comfort takes a long while to build,” especially for arrivals from corporate America. Tenure’s second dividend, via Senra’s MrBeast anecdote: 30,000 hours of shared context — “you don’t get that in 100 hours.”
• The taxes get mitigated deliberately: a Rising Stars program of roughly 20–30 people yearly, an internal MBA that exposes them to markets such as India, the US, and Germany and has them sit with both co-CEOs, plus senior hiring by acquisition when a skill is genuinely new — The Echo Nest in Boston “just to get very close to MIT” when Spotify needed deeper-learning talent. His explicitly far-fetched analogy: Apple’s AI stumble (Senra: “is that why Siri is so bad?”) — its organization was “slightly too old to see” the shift; “they should have gotten someone at a more senior level sooner.”

11. Which era are we in? “Maybe it hasn’t even happened yet”

• His closing framework: the world moves in macro waves — Spotify itself rode cheap broadband, then had to adapt when smartphones broke its computer-only free tier because streaming over cellular in 2008 “wouldn’t even start.” It responded with offline sync and a mobile model. An analyst extrapolating 2015→2025 — more mobile, more subscriptions, more ads — “would have done really well.” The same method applied 2005→2015 “would have missed the smartphone… missed Uber… missed the entire world.”
• His read on now is strongly hedged but clear: “I think we’re in one of these where everything changes.” The comfort inside the fear: “guess when Spotify grew the most? Periods of change” — when things are stable, market share stays stable; “you don’t eat market share.” So the principle is “always be first,” even while the destination stays open: the era’s Ubers and Airbnbs “maybe they are OpenAI and Anthropic, but maybe not. Maybe it hasn’t even happened yet.”