Untitled

DANIEL EK AND ANDREAS EHN spent much of the summer of 2006 recruiting the best engineers around. Daniel attracted some of the consultants who had helped him at Stardoll. Andreas proved invaluable in drawing engineers from KTH, where he’d earned a reputation as one of the brightest students in his graduating class.

In August, a small group of engineers flew to Barcelona for Spotify’s kick-off. Over tapas and red wine, Daniel and Martin Lorentzon explained that they wanted to create a legal, torrent-based platform for the distribution of music, and possibly video. The service would be ad-funded but free to use, they explained, because that was the only way to fight piracy.

They also made it clear that developing the product would be their top priority. Commercial licenses and agreements could wait. Fortunately, the gang in Barcelona didn’t grasp how difficult those final challenges would prove to be. If they had, they may never have attempted to build the platform. In the era of Kazaa and The Pirate Bay, the word “free” caused record label executives’ eyes to roll, their heads to shake, and their doors to close. Spotify’s founders had never negotiated licenses before. They had no idea how strongly the industry would resist their attempt to build an ad-based streaming service by employing the same technology that was being used for illegal file sharing.

Yet the gang munching tapas on the coast of Catalonia had at least three things going for it: Martin’s experience and considerable wealth; Daniel’s clear and unwavering vision of a product; and Andreas’s ability to attract and inspire Sweden’s best programmers.

In the fall of 2006, the Spotify engineers moved into the company’s first office. It was located on the second floor of an apartment building on Riddargatan, where Stockholm’s central business district meets the posh residential area of Östermalm. Here, the Spotify crew spent their first week lugging IKEA flat packs up the stairs and assembling office furniture. They unpacked whiteboards and installed computers and servers. It was a humbling task, literally building a company from scratch. No one could know that the work they’d started would, over time, turn hundreds of Spotify employees into millionaires.

The king of torrent technology at this time was a Swede named Ludvig Strigeus. He was a self-taught, twenty-five-year-old hacker who had single-handedly built μTorrent (pronounced “microtorrent”), one of the world’s most popular file-sharing clients. The ultralight program was used to download files rapidly from file websites such as The Pirate Bay. Early on, Daniel Ek and Martin Lorentzon realized that “Ludde,” as he was called, could be key to Spotify’s success.

Martin had set up a meeting through an acquaintance named Niklas Ivarsson, a charismatic engineer from Borås and head of the European division of ATI, a company that sold software to the automotive industry. He had worked with Ludde and was blown away by the young programmer’s talent for analyzing code and quickly building his own programs.

Ludvig Strigeus was the kind of genius who, as a child, would spend time disassembling household appliances to see how they worked. His mother told a story of how, in kindergarten, he was able to fix a broken dishwasher before the repairman showed up. But the young Ludde also suffered from a tragic medical condition. As a toddler, his parents had noticed that something wasn’t right with his hips. The doctors determined that he had spinal muscular atrophy, an incurable disease that breaks down the body over time. By the time Ludde was eight years old, he was confined to a wheelchair.

Ludde’s first computer was a Commodore VIC-20. He later inherited a PC, learned to code in Basic, and began creating simple programs and games. In his teens, he became a hacker and, quite famously, cloned Scumm, a popular gaming engine used to build computer games in the late 1980s. Scumm ran on protected source code, which stopped most young coders from looking under the hood. But Ludde was able to analyze the underlying assembler code, which had passed through a compiler and been translated to a language that only computer processors could understand. Line by line, he deciphered the code and translated the information back into a human programming language. It took him several years, but in the end Ludde had built something he called “ScummVM.” Its users could now convert Scumm games into any desired operating system. Ludde then released his clone as open source code so that anyone could contribute. It was a remarkable feat of programming, especially for a teenager.

Ludde’s youthful antics made him a phenomenal and unconventional programmer. He preferred to code in C++, an older language that was considerably more difficult than popular alternatives like Java and C-sharp. But when used properly, C++ resulted in fast and lightweight programs. In 2005, when Ludde released μTorrent, the entire file weighed only fifty kilobytes, approximately half as much as a low-resolution photograph. It was the kind of stuff that turns hackers into legends.

Daniel Ek and Martin Lorentzon began to woo Ludvig Strigeus in the summer of 2006. They traveled to ATI’s offices in the Gårda business district of Gothenburg, where Niklas Ivarsson made the introduction. The two founders explained that Spotify’s system would be based on bittorrent technology. They needed his help to build it, and they were willing to buy μTorrent in order to recruit him. They were proposing an “acquihire” long before the term became widely used in the tech world.

Ludde thought the idea sounded cool enough, but he was hesitant. Spotify’s offer was low on cash, and he already had two American suitors. One was BitTorrent Inc, founded by the programming legend Bram Cohen, who had authored the original protocol. The other was Azureus, who had a competing bittorrent client that was popular, but not as fast as the one Ludde had built.

By the early fall of 2006, Ludvig Strigeus felt overwhelmed by his options. Should he sell μTorrent to Spotify and start working there, or accept one of his American offers? The people from Azureus had already flown into Gothenburg to wine and dine him at the Elite Hotel. Representatives from BitTorrent Inc had gifted him a trip to the film festival in Cannes, all expenses paid. All three of his bidders seemed impatient. The problem was too much for Ludde’s analytical mind. There were too many unknown variables.

Azureus sent a contract full of American legalese that Ludde struggled to comprehend. The ramifications of selling his company to a US operation were making him nervous. After all, his program was being used for illegal downloading. Just a few months earlier, The Pirate Bay had been the target of a police raid in Sweden. What if he, too, was sued? All things considered, Ludde decided to accept the offer from Spotify. It felt reassuring to work for a Swedish company, close to his colleagues.

In mid-October 2006, Spotify finalized the deal with μTorrent. At the same time, Niklas Ivarsson left ATI to join Spotify. He would soon prove an asset in the difficult negotiations with the record labels.

Spotify paid Ludvig Strigeus a small amount of cash, and enough shares to make him Spotify’s fourth-largest shareholder.

A few weeks after the deal, Spotify sold μTorrent to BitTorrent Inc. Over the years, Ludde’s shares would be diluted as new investors came on board, but his stake was large enough to be valued in the hundreds of millions of dollars by the time Spotify went public.

From the start, Daniel Ek knew what he expected of his CTO, Andreas Ehn. Spotify’s client needed to be quick and nimble. There was no room for the types of glitches often found in other media players on the market. The music needed to flow like water from a faucet. Delays due to buffering would not be tolerated at Spotify. It’s “not cool to have to wait,” as an early version of the Spotify.com website put it.

The plan was to base the system on bittorrent technology. The users would download the Spotify client and offload its own servers by storing parts of the songs on their own hard drives, sharing them with other users in the network. The arrangement would speed up the system and outsource some of Spotify’s broadband expenses. Unlike The Pirate Bay, Spotify intended to share a part of its advertising revenue with artists and record companies. But the founders clearly felt their users shouldn’t have to pay for music.

“Our service is ad-funded, so it costs you nothing to use,” the website stated, long before the service had been launched.

Programming chops were paramount to Andreas as he expanded his team. He made one key recruitment in October 2006, when he called Fredrik Niemelä, a 27-year-old doctoral student at KTH, and offered him a job. Fredrik was a soft-spoken computer scientist from Norrland, in the north of Sweden. He sported a ponytail and goatee, and had been captain of a coding team at KTH that would soon win a world championship in programming.

Fredrik Niemelä would recall how Andreas, over dinner in Stockholm’s bourgeois district of Vasastan, painted a grand vision of Spotify’s future. The goal was to build an “agnostic” streaming platform, Andreas explained. Music was just the first step. After that, Spotify would expand into streaming television, film, and more. They were building a company, sure, but the product would be technically sophisticated, and ideologically akin to the file-sharing movement.

The Spotify CTO made a profound first impression on Fredrik, who was immediately drawn in by the prospect of applying his knowledge to solve a contemporary problem. After all, the battle between young file-sharers and record companies was on everyone’s lips. Fredrik also had a great deal of respect for many of the other engineers who had left KTH to join Spotify.

To start, Fredrik agreed to take on a part-time role as “technical advisor.” A few months later, he started working full-time. Soon, he was staying at work late into the night. Eventually, he would stay behind until the early morning, taking cabs back to his apartment in the suburb of Rinkeby. A few hours later he would wake up, take the subway back into the city, and start over. Daniel Ek tracked his progress, and figured Fredrik Niemelä would soon be due for a promotion.

The office on Riddargatan quickly became a second home for the young engineers building Spotify’s desktop client. Many of them were in their mid-twenties working their first real job. They would show up late in the morning but work long hours. The staffers had squeezed a foosball table into the office, and some would stay behind for some late-night poker, too.

The guys working on the “back end” technology had camped out in the apartment’s largest room, past the closet with the humming servers. Floral-patterned curtains hung over the windows, right next to a whiteboard that Andreas Ehn presided over. Fredrik Niemelä had quickly become his right-hand man, primarily dedicated to building the streaming technology. Others were occupied with the database or the advertising platform. A programmer named Magnus Hult was in charge of all the metadata that threaded together songs, artists, and album art. The team would organize Spotify’s music database like books in a library. Users would be able to search by artist or song, and discover other music by the same artist. They could also find cover songs, guest appearances, and songs by the same producer. Later, Spotify would pay for short biographical texts and become a sort of encyclopedia for music. Eventually, the company would develop algorithms that made recommendations, pushing music by Rihanna on Beyoncé fans.

There was a smaller room for those working on the “front end” of the client. This was the domain of Ludvig Strigeus, who built the user interface for Windows, and then for Mac, all by himself. Ludde worked out of Gothenburg but made regular visits to the team in Stockholm. He worked closely with Rasmus Andersson, a designer who would touch up his work and create Spotify’s graphic design.

At Spotify’s first offices on Riddargatan 20.
Daniel Ek and Rasmus Andersson (seated). (Rasmus Andersson)

A third room housed the rest of the Spotify team. That included Petra Hansson, Spotify’s general counsel, and Daniel Ek, when he wasn’t out negotiating licenses with the record labels. In February 2007, they were joined by Jonathan Forster, Spotify’s first director of sales. He soon realized that his job wouldn’t be easy, complaining that clients didn’t see the value of targeted ads.

“You can’t sell steak to people who want ground beef,” the Brit would tell his colleagues.

It fell on Martin Lorentzon to deal with investors and venture capital firms. On a regular day, he would hit the gym and go out for lunch with a business acquaintance. Now and then, he would drop by the office and crack jokes with the engineers.

Co-founder Martin Lorentzon interrupting a meeting in Spotify’s second offices with a little dance. (Rasmus Andersson)

The programmers chuckled. They knew he was up to something important, but not exactly what. Then again, he didn’t really have a clear grasp of what they were up to, either.

Martin would set the tone for a culture of pranks and practical jokes around the office. He would later describe how employees slipped inappropriate magazines into one another’s bags, or entered fake status updates on one another’s Facebook pages.

At one point during an early pitch at a venture capital firm in the US, Lorentzon recalled opening his laptop and unleashing a pre-installed program that performed a Google search for the term “How to enlarge a small penis.”

The prank, initiated by an employee, caught Daniel off guard and the pitch ended up failing.

Spotify’s engineers were now forging ahead to build a first-class product. Daniel Ek had insisted that users should be able to find a song instantly and play it without delay. Soon, all the technical discussions revolved around latency, which had a number: 200 milliseconds. Fredrik Niemelä had read that if a song started playing within 0.2 seconds of a button being pushed, people would experience it as immediate. The bandwidth in Sweden’s broadband networks was already good enough for songs to load faster than they were being played. So, in theory, it should be possible to start playing a song while it was still loading.

The challenge with bittorrent was that the files were broken down into pieces that arrived in a “torrent,” so in no particular order. Forcing the protocol to change this would slow the process down. As Andreas Ehn’s team worked on solving this challenge, his “back end” team turned their office into a war room. For weeks, the engineers scribbled on the white board, vigorously debating how the bittorrent technology could be repurposed to suit their needs. Finally, one of the sketches seemed to make sense. The solution even looked elegant.

One illustration showed how the existing torrent system cut the music files vertically, like a loaf of bread. These pieces were downloaded, assembled, and then played from left to right, or beginning to end. Since the pieces arrived in random order, they all needed to be in place for the song to begin to play without interruption. Another illustration showed the same music file cut lengthwise, like the layers of a submarine sandwich. It was sliced vertically, too, but only four times. That meant that whatever piece of the song arrived first, it was far more likely to contain a part of the first few seconds—hence the song could start streaming more quickly. The team tested the theory, which worked in practice. The files were streamed from the servers in the nearby closet, and the music played instantly.

With this major challenge solved, the engineers started developing a patchwork of solutions that would ensure that the music kept playing, even when the connection was shaky. The main trick was to load, or buffer, the songs that the user was likely to want to hear next. The very first selection would always be hard to anticipate. But if the user was listening to a playlist, the next song on the list would likely follow. The system, the engineers reasoned, could thus start loading the next few anticipated songs into the computer’s cache memory, and simply clear it if the user made another selection.

After a bit of trial and error, the coders figured that any unexpected selection would have to stream from Spotify’s own, high-end servers. The peer-to-peer technology was used only for song choices that were easy to predict, such as the next few songs on an album or a playlist. This division of labor also cut costs, as the other users’ computers helped store and transfer slices of songs. Another trick was to download a user’s favorite songs and store them in their own computer’s memory. These hacks would ensure a smooth user experience and guard against glitches and delays. Spotify’s first protocol became a sort of hybrid. It was a torrent network supported by a central server, bolstered by a bunch of homespun features. It wasn’t flawless, but it was close enough.

In January 2007, the first demo of Spotify’s music player was ready. Everyone who tried it found the experience magical. The engineers could hardly believe what they had accomplished in just a few months. Spotify’s client was better than anything else on the current market.

“It was rocket science, for real,” an early admirer would express. Daniel was delighted with his engineering team and quickly promoted Fredrik Niemelä to the role of Chief Product Officer, CPO.

At this point, practically all the songs in Spotify’s database were pirated music files from The Pirate Bay and similar websites. The irony escaped no one at the company. Spotify, after all, aimed to be a legal alternative to piracy. But what were they supposed to do—buy millions of songs on iTunes to fill up the database? Even if they did, the service still wouldn’t be legal when the beta version came out.

Spotify needed to finish building the product, patent its newly developed technology, and seal licensing deals with the record companies. Then, everything would probably work out. With a little luck, Daniel Ek hoped to launch the service in the fall of 2007.

As Spotify was building its music player, music industry executives were growing increasingly frustrated with all the file-sharing in Sweden. Per Sundin, the Nordic head of Sony BMG, would recall how he lost it at a dinner in Stockholm with some newfound acquaintances.

“So what you’re saying is that you’re a family of thieves?” the 43-year-old blurted out over the table.

He and his wife Jenny were guests at the home of a couple they’d recently gotten to know, and the subject of piracy was about to ruin yet another pleasant evening. Jenny rolled her eyes. She often found herself wishing her husband would shut up about what he did for a living.

“Yes, but everyone downloads stuff. We download. We’ve stopped buying CDs altogether,” the husband on the other side of the table said.

Moments like these made Per Sundin’s blood boil. He knew that his wife would chastise him when they got home, but he couldn’t contain himself. A new flood of accusations came pouring out.

“How would you feel if your kids shoplifted at the local convenience store? Would that be okay, just because it’s easy? What do you think artists are supposed to live on?”

By early 2007, piracy had caused the record industry’s revenue to decline for six years straight. Nowhere in Europe was the problem worse than in Sweden, where 1.2 of the nation’s 9 million people were said to share pirated music files. Per Sundin had already been forced to lay off staff a number of times during the past few years.

When his international colleagues within Sony BMG asked him why file-sharing was so common in his country, he’d point to several factors. One was the late arrival of iTunes. Apple’s music store didn’t launch in Sweden until 2005, and by then, the nation had already given the world Kazaa, The Pirate Bay, and μTorrent. Another reason was Sweden’s “home PC reform” from 1998, a government program that subsidized home computers and spawned all kinds of computer literacy, including hacking and file-sharing. Then there were all the incentives that Swedish politicians had provided for the expansion of broadband networks. Most American households could only dream of the high speeds and low prices readily available in Sweden.

But it wasn’t just the prevalence of piracy that bothered Per Sundin. It was the lax attitude toward the problem, shown by lawmakers and the public alike. For him, the low-water mark arrived with the national elections of 2006. A newly formed political party called The Pirate Party was gaining support among young voters. As a result, neither leader of the country’s two major political parties was willing to slam file-sharing.

“We need to make sure that young people who do this downloading don’t see themselves as criminals,” Prime Minister Göran Persson said in a live television debate a few weeks ahead of the election.

Instead of criticizing his opponent, Fredrik Reinfeldt, the leader of the conservative opposition party, wavered on the issue.

“Are you watching the debate?” Per Sundin’s mother said when she called him on the phone. “You need to quit the record industry!”

Per Sundin would soon leave Sony BMG, but he would remain in the industry. He was about to become the head of Universal Music in Sweden. Soon, a new streaming service called Spotify would pop up on his radar.

In March 2007, a few members of Spotify’s engineering team had just pulled another all-nighter at the office. By the early morning, Andreas Ehn, Fredrik Niemelä, and a developer named Mattias de Zalenski had put together the music player’s back and front ends, thus giving birth to the service. When it was time to register the first account, Andreas and Fredrik exchanged glances. After a tense moment, Fredrik conceded to the man who had recruited him.

Andreas would recall that, indeed, he registered Spotify’s very first account. Soon, the team would create more, selectively distributing beta invites among their friends. These soon sought-after invites were even said to be sold for money on the KTH campus. Sophia Bendz, the now-head of information and marketing at Spotify who had been recruited from the PR firm Prime, was in charge of generating hype in the right circles.

Late night at Spotify’s Stockholm headquarters in 2007.
Seated: Sophia Bendz. To her right: Martin Lorentzon. (Rasmus Andersson)

Ludvig Strigeus and Rasmus Andersson had built a sleek client. Spotify’s beta users could search for music and create playlists. There were separate pages for artists and albums, and links between the two. And, most importantly, the music began streaming immediately. It felt like pressing play on a locally stored file in iTunes. In the Spotify client, encrypted files were temporarily saved in the cache memory on the user’s computer. The decryption key wasn’t saved on the hard disk, but in the cloud. Users gained access to the music, but they didn’t own it. The age of music streaming was born.

Daniel Ek was amazed by the progress, but he was also plagued by a guilty conscience. He was one of few people at Spotify who knew how poorly the negotiations with the record labels and music publishers were going. Still, after paying a lump sum to Sweden’s performance-rights organization, STIM, he was able to secure test licenses for a limited beta-release.

Spotify’s beta version hit the market in April 2007. To mark the occasion, the team gathered for a group photo. In it, fourteen nerdy-looking young men are seen smiling into the camera. Developer Gunnar Kreitz is wearing a black t-shirt with KTH’s logo on the chest. Next to him stands Andreas Ehn, in one of his neatly pressed shirts. Further back stands Jonathan Forster, the only one to wear a suit jacket. Daniel Ek’s friend from high school, Martin Birkeldh, tasked with administrative duties around the office, looks over the shoulder of the lanky Fredrik Niemelä, who is sporting his goatee and wearing his long hair pulled back. Crouching in the front row is the designer Rasmus Andersson, dressed in a blue and white striped shirt.

The Spotify team in April 2007. Standing from left: Mattias Arrelid, Gunnar Kreitz, Andreas Ehn, Per Malm, Jonathan Forster, Magnus Hult, Daniel Ekberger, Tommie Gannert, Martin Birkeldh, Fredrik Niemelä. Crouching from left: Rasmus Andersson, Jon Åslund, Andreas Mattsson.
Front row: Daniel Ek. (Rasmus Andersson)

Spotify’s twenty-four-year-old co-founder, Daniel Ek, is crouched in the very front, dressed in a blue polo shirt, with tousled hair and a bald spot. The young chief executive is grinning. His co-founder Martin Lorentzon is not in the picture.

Pär-Jörgen Pärson stared into the glow of the laptop computer. To him, the music player looked like a dark gray version of iTunes. But when he typed in his favorite band, Killswitch Engage, and pressed play, the song started immediately, without the slightest delay. It didn’t seem to download or buffer. It was just there.

On the other side of the table sat Martin Lorentzon, smiling wryly. Pär-Jörgen Pärson had once been his boss at the venture capital firm Cell Ventures. That firm no longer existed, but Pär-Jörgen was now a partner at the venture capital firm Northzone, with offices in one of the five high rises near the Hötorget square in central Stockholm. Martin had come there to dazzle him, and had clearly succeeded.

Martin showed him that the computer was connected to the internet through his cell phone. Pär-Jörgen was taken aback. Over the years, he’d seen hundreds of investors burn their fingers on tech companies promising to reshape the music industry, but what he had in front of him could turn into something truly great, he thought. Eight years ago, Pär-Jörgen had missed his chance to invest in Martin’s ad-tech company Tradedoubler. He didn’t want to make the same mistake again. Spotify, he thought, could be the biggest thing to come out of Sweden since Skype.

The seasoned investor did his best to maintain his poker face. He didn’t want to seem too interested. He knew that negotiating with Martin wasn’t going to be easy, since he was already so goddamn rich.

Finally, the Northzone partner hinted that he would consider investing in Spotify. But Martin didn’t want to state the terms quite yet. He just smiled and said he’d get back in touch.