Spotify

How a Swedish startup rewired the global music industry

By VastBlue Editorial · 2026-03-26 · 18 min read

Series: Made in Europe · Episode 2

The Country That Broke Music

Sweden has a complicated relationship with recorded music. Per capita, no country on earth has exported more commercially successful songs. ABBA, Roxette, Ace of Base, The Cardigans, Robyn — the list of globally dominant Swedish acts stretches across five decades. Behind the artists stands an even more influential layer: the producers and songwriters. Max Martin, born Martin Karl Sandberg in Stenhamra, has written or co-written more number-one Billboard hits than any songwriter in history except Lennon and McCartney. Denniz Pop built the template for modern pop production at Cheiron Studios in Stockholm before his death in 1998. Sweden does not merely participate in the global music industry. It architecturally shapes it.

Sweden also nearly destroyed it.

In 2003, a twenty-year-old Swedish programmer named Gottfrid Svartholm launched The Pirate Bay from a server room in Stockholm. Within two years it had become the largest file-sharing site on the internet, and its most popular category was music. The Pirate Bay did not invent digital piracy — Napster had done that in 1999, and BitTorrent protocol had been published by Bram Cohen in 2001. But The Pirate Bay industrialised it. It provided a clean, searchable, reliably indexed interface for downloading any album ever recorded, at any time, for free. It was, in effect, the music library that every teenager in the world had fantasised about, and it operated from a country whose copyright enforcement mechanisms were, at that moment, spectacularly inadequate to deal with it.

The impact on the recorded music industry was catastrophic and measurable. Global recorded music revenue, which had peaked at $23.8 billion in 1999, fell to $15 billion by 2008. In Sweden specifically, the decline was even more acute — Swedish music revenues dropped by over 50 percent between 2001 and 2008, despite the country producing more internationally successful music than ever before. The industry was watching its revenue evaporate in real time, and the country that had given the world ABBA was also the country hosting the infrastructure that was making recorded music functionally free.

This was the environment in which Daniel Ek grew up. Born in 1983 in Rågsved, a working-class suburb of Stockholm, Ek was programming by age eight, running a web design business from his bedroom by fourteen, and had dropped out of a computer science programme at the Royal Institute of Technology by eighteen — not because he could not do the work, but because he already had clients paying him to do more interesting work outside the university. By his early twenties, he had co-founded and sold a digital marketing company called Advertigo, worked as CEO of the peer-to-peer file-sharing company uTorrent, and found himself in an unusual position: wealthy enough to not need a job, young enough to want one, and deeply embedded in the Swedish tech ecosystem that had both created the music piracy problem and suffered from its consequences.

Martin Lorentzon, thirteen years older, had co-founded Tradedoubler, a performance marketing company that had gone public on the Stockholm Stock Exchange in 2005. Lorentzon was the business architect — the person who understood unit economics, advertiser psychology, and the mechanics of building companies that scaled across European markets with their fragmented regulatory landscapes. Where Ek thought in code and product, Lorentzon thought in margin structure and market entry. They met through the Stockholm tech scene and found in each other the complementary skill set that defines the most consequential co-founder relationships.

In 2006, they founded Spotify. The name was, by some accounts, a mishearing during a brainstorming session — one of them shouted a word, the other misheard it, and the misheard version stuck. They checked that the domain was available and moved on. The name was unimportant. The idea was everything: build a service that was better than piracy.

Engineering the Impossible

The technical problem Spotify set out to solve was, by any measure, extraordinary. The service needed to stream music instantly — with a latency so low that pressing play felt like the song was already loaded on your device. In 2006, average European broadband speeds were between 2 and 8 megabits per second, mobile internet barely existed in usable form, and the infrastructure for content delivery networks was primitive compared to what exists today. Buffering was not a minor inconvenience. It was the default experience of every streaming service that had been tried before.

Spotify's engineering team, working from a small office on Riddargatan in Stockholm, built a hybrid architecture that was unlike anything else in the industry. The system combined direct server streaming with a peer-to-peer distribution layer — essentially borrowing the same BitTorrent-derived technology that powered piracy and repurposing it for a legal service. When you played a song, the first few seconds came from Spotify's servers to guarantee instant playback. But as the track continued, subsequent chunks might come from other Spotify users nearby who had recently played the same song and whose clients had cached the data locally. The system maintained an aggressive local cache on each user's machine, pre-fetching songs the algorithm predicted you were likely to play next. The result was playback that felt instantaneous — faster than loading a locally stored MP3 file in most cases.

The desktop client was built in C++ for performance, with a custom audio codec pipeline that could dynamically adjust bitrate based on network conditions. The team wrote their own memory allocator. They wrote their own networking stack. They wrote their own audio decoder. In an industry where most consumer software was being built on web technologies and interpreted languages, Spotify was engineering at a level of systems-programming rigour more commonly associated with game engines or operating systems. The reason was simple: if the experience was not instantaneous, people would return to piracy. There was no margin for buffering. Zero tolerance for latency. The product had to be faster than a pirated MP3 file sitting on your hard drive, or it had failed.

The backend was equally ambitious. Spotify needed to serve a catalogue that would eventually encompass tens of millions of tracks, each with complex metadata — artist, album, composer, producer, label, territory-specific licensing restrictions, release date, ISRC codes. The rights management alone was a database problem of staggering complexity. A single song might be owned by one label in Sweden, another in Japan, a third in Brazil, and subject to different royalty rates in each territory. The system needed to know, in real time, whether a given user in a given country had the right to stream a given track — and it needed to make that determination in milliseconds, because any perceptible delay between pressing play and hearing music was a failure.

The engineering team that built this was almost entirely Swedish, drawn from the same universities and the same Stockholm tech ecosystem that had produced the piracy infrastructure they were now competing against. Several early Spotify engineers had previously worked on peer-to-peer systems. They understood the architecture of distributed file sharing at a deep level, and they turned that knowledge toward building something legal, licensed, and fast enough to make piracy feel inconvenient by comparison.

The Licensing Labyrinth

Building the technology was hard. Getting permission to use it was harder.

The music industry in 2006 was traumatised. It had spent five years watching its revenue collapse, suing its own customers, and failing to build digital alternatives that anyone wanted to use. The iTunes Music Store, launched by Apple in 2003, had demonstrated that people would pay for digital music — but only for individual downloads, not for streaming. The record labels viewed streaming with deep suspicion. A download was a sale: you paid $0.99, you owned the file, the transaction was complete. Streaming was access: you paid a monthly fee (or tolerated ads), and you could listen to anything — but you owned nothing. The labels could not see how streaming would generate enough revenue to replace the economics of the album sale, and they had already been burned badly enough by technology companies to distrust anyone arriving from Stockholm with a pitch about making music free.

Daniel Ek spent the better part of two years in licensing negotiations. He flew to New York, Los Angeles, and London repeatedly, meeting with label executives who were, in many cases, openly hostile to the premise. Universal Music Group, Sony BMG, Warner Music Group, and EMI — the four companies that collectively controlled approximately 70 percent of global recorded music — each had to be convinced individually. Each had different concerns, different demands, and different internal politics. Each wanted guarantees that Spotify would not cannibalise download sales. Each wanted advance payments. Each wanted equity in the company. Each wanted minimum per-stream royalty rates that, on paper, would make the service uneconomic.

Ek gave them equity. This was the decisive strategic concession. By offering ownership stakes in Spotify to the major labels, he aligned their incentives with the platform's success. If Spotify grew, the labels' equity would appreciate. If Spotify eventually went public, the labels would hold shares in a technology company valued at multiples far higher than a music company. The labels were no longer just licensors — they were investors. Their hostility did not disappear overnight, but it became tempered by financial self-interest.

The music industry had spent five years suing teenagers for downloading songs. Ek offered them something more powerful than a lawsuit: a business model that made piracy irrelevant by being better than piracy.

Editorial observation

The licensing negotiations also required engaging with a parallel universe of rights holders that most technology companies had never encountered. Recorded music rights are split between the master recording (typically owned by the label) and the underlying composition (typically owned by the publisher, which may or may not be affiliated with the label). A single song therefore requires two separate licences: one for the sound recording, one for the musical composition. The publishers — companies like Sony/ATV, Universal Music Publishing, and Warner Chappell — had their own concerns, their own rate structures, and their own negotiating leverage. In many countries, mechanical reproduction rights and performance rights were administered by separate collecting societies, each with their own rules, each requiring separate agreements.

Sweden, paradoxically, made this easier. STIM, the Swedish Performing Rights Society, was comparatively modern in its outlook and willing to engage with digital licensing in ways that larger collecting societies in the US and UK were not. Spotify launched first in Sweden in October 2008, operating under a limited set of licensing agreements that gave it access to a substantial catalogue in a single, small market. The Swedish launch was a proof of concept — not just technically, but commercially and legally. It demonstrated that a streaming service could operate within the existing rights framework without collapsing the economics of the industry. It demonstrated that users would tolerate a freemium model — free access with advertising, or paid subscription without ads. And it demonstrated that, given a choice between piracy and a legal service that worked as well as piracy, users would choose the legal option.

The Freemium Gamble

The freemium model was not Ek's invention — the term had been coined by venture capitalist Fred Wilson in 2006, and companies like Skype and Flickr had already demonstrated that free tiers could drive adoption for paid products. But applying freemium to music was considered borderline reckless. The labels had fought for years to establish the principle that music had monetary value, that a song was worth at least $0.99, that the industry's existential crisis was rooted in consumers' unwillingness to pay. Now a Swedish startup was proposing to give away unlimited music for free, supported by advertising, and hope that some fraction of users would convert to a $9.99 monthly subscription.

The logic, however, was ruthlessly clear. Spotify's competitor was not iTunes. Spotify's competitor was The Pirate Bay. And The Pirate Bay was free. You could not compete with free by charging $0.99 per track. You could only compete with free by also being free — but legal, faster, better-curated, and integrated into an ecosystem that piracy could never provide: playlists that synced across devices, algorithmic recommendations, social features, artist pages, real-time lyrics, concert listings. The free tier was not generosity. It was a weapon against piracy, funded by advertising revenue that — while modest on a per-user basis — was infinitely more than the zero dollars generated by a pirated download.

The conversion mathematics were brutal. In the early years, fewer than 10 percent of free users converted to premium subscriptions. The advertising revenue from the free tier barely covered the royalty costs of streaming to those users. Spotify was, by conventional analysis, losing money on the vast majority of its user base. The bet was that the free tier would grow the total addressable market — bringing people into a legal music ecosystem who would otherwise have pirated — and that a sufficient percentage would eventually convert to premium as their engagement deepened and their tolerance for advertising diminished.

The bet worked, slowly. By 2012, Spotify had 20 million active users, of whom 5 million were paying subscribers. By 2015, those numbers had grown to 75 million active users and 20 million subscribers. The conversion rate held steady at roughly 25 to 30 percent — far higher than the initial 10 percent, as the product matured and the premium offering expanded to include offline listening, higher audio quality, and family plans. Each percentage point of conversion improvement represented millions of dollars in additional subscription revenue, and the economics of the free tier improved as Spotify's advertising business matured and began attracting major brand advertisers willing to pay premium rates for access to an engaged, demographically desirable audience.

The Machine That Listens

Spotify's most consequential innovation may not be streaming at all. It may be recommendation.

In 2014, Spotify acquired a small Boston-based company called The Echo Nest. Founded by two MIT Media Lab researchers, Tristan Jehan and Brian Whitman, The Echo Nest had spent nearly a decade building something unprecedented: a machine learning system that could analyse music at the acoustic level. Not metadata — not genre tags applied by humans, not "customers who bought this also bought" collaborative filtering. The Echo Nest's algorithms listened to the actual audio signal. They measured tempo, key, loudness, energy, danceability, valence (a measure of musical positivity), acousticness, instrumentalness, liveness, speechiness. They decomposed every song into a multidimensional vector of acoustic features and mapped the entire recorded music catalogue into a continuous mathematical space where similarity could be calculated with precision.

The acquisition was transformative. Before The Echo Nest, Spotify's recommendations relied primarily on collaborative filtering — the same technique Netflix and Amazon used. If users who listened to Artist A also listened to Artist B, the system would recommend Artist B to new listeners of Artist A. This approach worked for popular music but failed catastrophically for the long tail. Niche artists, new releases, and cross-genre work had insufficient listening data to generate meaningful collaborative signals. They existed in a recommendation cold start problem: you could not recommend them because not enough people had listened to them, and not enough people listened to them because they were never recommended.

The Echo Nest's acoustic analysis broke the cold start. A new song, uploaded to Spotify with zero listeners, could be immediately analysed at the signal level, positioned in the acoustic feature space, and recommended to users whose listening history suggested affinity for songs with similar acoustic properties. The system did not need humans to listen first. It did not need genre tags. It did not need editorial curation. It needed only the audio file itself. This was the technical foundation for Discover Weekly — the personalised playlist, launched in 2015, that presented each user with thirty songs every Monday, algorithmically selected from music they had never heard but were statistically likely to enjoy.

The scale of this system is difficult to comprehend. Spotify's recommendation engine processes over 100 million tracks, analyses the listening behaviour of over 600 million users, and generates billions of personalised recommendations daily. It operates a constellation of machine learning models — collaborative filtering, natural language processing of music reviews and blog posts, audio feature extraction via convolutional neural networks, reinforcement learning systems that optimise for long-term engagement rather than immediate clicks. The models run across distributed computing infrastructure that processes petabytes of listening data in near-real time. Each time you press play, skip a song, add a track to a playlist, or let an album finish without interruption, you are generating training data that refines the model's understanding of your preferences with increasing granularity.

The consequence for the music industry has been structural. Before Spotify, music discovery was mediated by gatekeepers: radio programmers, music journalists, label A&R representatives, MTV producers. These gatekeepers were human, fallible, biased toward established artists, and concentrated in three cities — New York, Los Angeles, and London. After Spotify, music discovery became algorithmic, global, and decentralised. An artist recording in a bedroom in Lagos could reach listeners in Helsinki without ever being played on radio, reviewed in a magazine, or signed to a label. The gatekeepers did not disappear — editorial playlists like RapCaviar and Today's Top Hits still wield enormous influence — but they were supplemented by a recommendation infrastructure that operated at a scale no human curation system could match.

What Spotify Changed — and What It Did Not

The simplest measure of Spotify's impact is revenue. In 2014, the global recorded music industry generated $14.2 billion in revenue. By 2023, that figure had grown to $28.6 billion — surpassing the pre-piracy peak for the first time, adjusted for inflation. Streaming accounted for 67 percent of total global music revenue, and Spotify was the single largest contributor, responsible for roughly a third of all streaming revenue worldwide. An industry that had spent fifteen years in contraction was growing again, and the growth was being driven primarily by a service built in Stockholm by a team that had started with two founders and a conviction that piracy was a service problem.

But Spotify's impact extends beyond revenue recovery. The service fundamentally altered how music is consumed, and that alteration has consequences that the industry is still processing. The album, as a creative and commercial unit, has been diminished. In the streaming era, individual tracks compete for attention independent of their album context. Listening is increasingly playlist-driven — curated by algorithms, by Spotify's editorial team, or by users themselves — and playlists are organised by mood, activity, and moment rather than by artist or genre. A workout playlist might contain hip-hop, electronic dance music, and pop-punk in sequence. A "chill evening" playlist might blend jazz, ambient, and acoustic singer-songwriter. The organising principle is no longer "who made this" but "how does this feel."

This shift has reshaped how artists create. Songs have gotten shorter — the average length of a top-ten hit has decreased from four minutes and twenty seconds in 2000 to three minutes and seven seconds in 2023. Intros have been compressed or eliminated — in a skip-driven environment, the first five seconds determine whether a listener stays. Genres have blurred as artists optimise for playlist inclusion rather than genre-specific radio formats. The economic incentive is clear: Spotify counts a stream after thirty seconds of playback, so a three-minute song generates the same per-stream revenue as a seven-minute song but can be played twice as often in the same listening session.

The economics of streaming have also created a new distribution of value that is sharply contested. Spotify operates a pro-rata royalty model: all subscription and advertising revenue for a given period is pooled, and each rights holder receives a share proportional to their total streams as a percentage of all streams on the platform. This means that an artist's royalty is determined not by how much their own listeners pay, but by how their stream count compares to every other artist on the platform. A subscriber who listens exclusively to a single independent artist still contributes their $9.99 to the global pool, where it is distributed according to total platform-wide listening patterns — which are dominated by the most-streamed artists. The model has been criticised for systematically disadvantaging independent and niche artists, and Spotify has begun experimenting with alternative models, but the fundamental tension between scale economics and equitable compensation remains unresolved.

• Physical media (CDs, vinyl) accounted for 96% of music revenue in 1999. By 2023, physical formats represented just 17% of global revenue.
• Spotify pays between $0.003 and $0.005 per stream on average. An artist needs approximately 250 streams to earn one dollar.
• The top 1% of artists on Spotify account for approximately 90% of all streams — a power-law distribution that mirrors, and in some ways intensifies, the pre-digital star system.
• Spotify's marketplace for artists — tools for playlist pitching, audience analytics, and promotional campaigns — has created a parallel economy where artists invest time and money to optimise their algorithmic visibility.
• In 2023, over 100,000 tracks were uploaded to Spotify every single day, creating a catalogue management and discovery challenge unprecedented in the history of recorded music.

The per-stream royalty debate has become one of the defining cultural arguments of the streaming era. Musicians — particularly mid-tier and independent artists — argue that the rates are unsustainably low, that streaming has created a system where only superstars and catalogue owners earn meaningful income, and that the platform extracts disproportionate value from the creative labour it distributes. Spotify argues that it has returned over $40 billion to rights holders since its inception, that it has grown the total revenue pie rather than merely redistributing it, and that the comparison should not be between streaming rates and CD-era album sales, but between streaming revenue and the zero revenue generated by piracy. Both arguments contain truth. Neither contains the whole truth.

Stockholm to the World

Spotify went public on the New York Stock Exchange in April 2018, via a direct listing rather than a traditional IPO — another unconventional choice by a company that had built its identity on doing things differently. There were no underwriters, no roadshow, no artificial price support. The company simply listed its existing shares and let the market determine the price. The opening trade valued Spotify at approximately $29.5 billion, making it one of the most valuable European technology companies in history and validating the Stockholm ecosystem that had produced it.

The direct listing was characteristically Swedish in its aversion to unnecessary ceremony. Ek did not ring the opening bell at the NYSE. He was in Stockholm. The company posted a short note on its blog. There was no champagne-popping photograph, no CNBC interview from the trading floor, no breathless founder narrative about the journey from garage to global dominance. The company had always operated with a Scandinavian restraint that contrasted sharply with the performative exuberance of Silicon Valley culture. Internal meetings were efficient. Hierarchy was flat. Decisions were made by consensus when possible and by data when consensus failed. The engineering culture prized craftsmanship and reliability over speed and disruption — a reflection of the Swedish engineering tradition that also produced Ericsson, Volvo, and Saab.

Today, Spotify employs over 9,000 people across 42 offices in 20 countries. Its headquarters remain in Stockholm, at a campus in the Järla Sjö district that houses engineering, product, and design teams. The company has expanded far beyond music into podcasts, audiobooks, and what it calls "audio-first" experiences — a strategic bet that the same infrastructure built to stream music can be adapted to stream any audio content, with the same recommendation engine connecting listeners to content they did not know they wanted.

The podcast expansion, in particular, represents a wager worth hundreds of millions of dollars. Between 2019 and 2023, Spotify spent over $1 billion acquiring podcast companies and exclusive content — including Gimlet Media, Anchor, and a $200 million exclusive licensing deal with Joe Rogan. The strategy was to replicate in spoken audio what Spotify had achieved in music: become the default distribution platform, use exclusive content to drive subscriber growth, and build a recommendation layer that makes the platform indispensable. The results have been mixed — the exclusive content strategy was partially reversed in 2023 as the company shifted toward profitability — but the underlying ambition remains: to own the infrastructure through which the world accesses audio.

Spotify did not just build a product. It built the distribution layer for an entire medium. In doing so, it demonstrated something that Europe's technology ecosystem has struggled to prove: that a European company can define a global consumer category and sustain dominance within it.

Editorial observation

In the first quarter of 2024, Spotify reported its first-ever operating profit as a public company — €168 million, after years of losses that had tested investor patience and fuelled scepticism about the viability of the streaming model. The path to profitability required difficult decisions: workforce reductions of approximately 1,500 employees in late 2023, price increases across most markets, renegotiated licensing terms with major labels, and a strategic retreat from some of the most expensive podcast content deals. The profitability milestone did not prove that streaming was easy. It proved that streaming was possible — that a European company could build a global consumer platform, navigate the most complex licensing environment in entertainment, and eventually generate more revenue than it spent.

The European Question

Spotify's story matters beyond the music industry because it is one of the very few examples of a European consumer technology company achieving genuine global scale. The list is short and sobering. SAP, Spotify, Skype, Wise, Klarna — a handful of companies from a continent of 450 million people with some of the best universities, the deepest engineering traditions, and the most sophisticated consumer markets on earth. The question of why Europe does not produce more Spotifys is one of the central puzzles of the global technology economy, and Spotify's own history offers some clues.

First, Spotify benefited from a problem that was uniquely acute in Sweden. The combination of a world-leading music industry and world-leading piracy infrastructure created a specific pain point that demanded a specific solution. Ek did not set out to build a streaming platform in the abstract. He set out to solve a Swedish problem — and the solution happened to scale globally. This pattern — local specificity generating global relevance — recurs in European technology success stories more often than the Silicon Valley model of building for a global market from day one.

Second, Spotify was able to launch in a single, small, culturally coherent market and prove its model before expanding. Sweden's population of 10 million was large enough to generate meaningful data but small enough to negotiate manageable licensing agreements. The company spent two years operating only in Sweden, then expanded to the UK, France, and Spain before eventually launching in the United States in 2011 — five years after its founding. This patient, market-by-market expansion is antithetical to the Silicon Valley playbook of launching globally on day one, but it allowed Spotify to refine its product, prove its economics, and build relationships with rights holders gradually rather than attempting to negotiate with the entire global music industry simultaneously.

Third, and most painfully, Spotify eventually had to go to Wall Street. The company listed on the NYSE, not on the Stockholm Stock Exchange, not on Euronext, not on any European exchange. The European venture capital ecosystem in 2018 could not support a company of Spotify's scale and ambition. The institutional investor base was too small, the analyst coverage too thin, the liquidity too constrained. Spotify is headquartered in Stockholm and built by Swedish engineers, but its shareholders are overwhelmingly American institutional investors. This is the European technology paradox in miniature: the continent produces the engineering talent and the companies, but the capital markets that fund their growth and determine their valuations are located elsewhere.

Despite this, Spotify remains a fundamentally European company — in its engineering culture, its corporate governance, its approach to employee relations, and its relationship with the regulatory environment. It operates under European data protection law. It negotiates with European works councils. It pays European tax rates. And it has demonstrated, at global scale, that European engineering and European business models can compete with and, in some domains, defeat the best that Silicon Valley produces.

The music industry that Spotify rewired was, by 2006, a system in collapse — haemorrhaging revenue, suing its own customers, and unable to offer a digital experience that competed with piracy. Two decades later, that industry generates more revenue than at any point in its history, and the infrastructure enabling that recovery was engineered in Stockholm by a team that understood, from the inside, both the technology that had broken music and the engineering required to fix it. The pirate and the engineer came from the same country. They attended the same universities. They worked in the same programming languages. The difference was that one built a tool for taking, and the other built a tool for paying. Both tools worked. Only one of them built an industry.

Sources

1. IFPI Global Music Report 2024 — https://www.ifpi.org/wp-content/uploads/2024/04/IFPI-Global-Music-Report-2024.pdf
2. Spotify Technology S.A. — Q1 2024 Earnings Release — https://newsroom.spotify.com/2024-04-23/spotify-reports-first-quarter-2024-earnings/
3. Spotify Technology S.A. — SEC Form F-1 Registration Statement (2018) — https://www.sec.gov/Archives/edgar/data/1639920/000119312518063434/d494294df1.htm
4. Aguiar, L. & Waldfogel, J. "Streaming Reaches Flood Stage: Does Spotify Stimulate or Depress Music Sales?" NBER Working Paper No. 24939, 2018 — https://www.nber.org/papers/w24939
5. Nordgård, D. "The Music Business and Digital Impacts: Innovations and Disruptions in the Music Industries." Springer, 2018 — https://link.springer.com/book/10.1007/978-3-319-91887-7
6. Jehan, T. & Whitman, B. "The Echo Nest Musical Brain." Proceedings of ISMIR, 2005 — https://ismir2005.ismir.net/proceedings/1035.pdf
7. Ingham, T. "Spotify Has Now Paid More Than $40 Billion to the Music Industry." Music Business Worldwide, 2024 — https://www.musicbusinessworldwide.com/spotify-has-now-paid-more-than-40-billion-to-the-music-industry/