Skype

Skype was the free internet-calling product that Niklas Zennström and Janus Friis launched in 2003, and it grew faster than almost any software before it because the product itself was the marketing. The through-line across its stories is a single architectural bet, peer-to-peer, that made growth nearly costless and monetization elegant, yet left the company owning a login rather than a social graph, so when calling moved onto the phone Skype's lead evaporated.

Virality: Calling someone is an invitation to install

The problem. A new communications product has no reach on day one, and paying to acquire tens of millions of users through advertising would have destroyed the economics of a free calling service. Skype needed distribution it could not afford to buy.

The approach. Skype built the referral into the core action: a free Skype-to-Skype call only works if the person you are calling is also on Skype, so every user who wanted to talk to someone had a direct incentive to get that person to install the app. The loop was the product, not a bolt-on invite feature.

How it solved it. Acquisition became effectively free and growth was explosive: Skype passed roughly one million users in 2004 and about 50 million registered users by the time eBay agreed to buy it in September 2005, roughly two years after the August 2003 public beta. Marketing spend never had to scale with that curve because each user recruited the next.

Network Effect: More people on Skype made Skype worth more

The problem. Voice communication is worthless in isolation. Early on, the odds that any given contact was reachable on Skype were low, which is the classic cold-start weakness of any network product.

The approach. Skype leaned entirely on the direct network effect: every additional user increased the chance that any other user's contacts were reachable for free, compounding the product's value with each install rather than through any new feature.

How it solved it. The compounding was real and sustained, carrying Skype from about 50 million registered users in 2005 to roughly 170 million monthly users by Microsoft's 2011 acquisition and a peak above 300 million monthly active users around 2016. The caveat, which later proved decisive, was that this effect only defended the business while switching stayed costly.

Founder-Market Fit: The Kazaa team already ran a global P2P network

The problem. Carrying voice over the internet for free normally required telco-scale infrastructure, an enormous capital cost that would have made a free product impossible to operate. Solving that demanded rare expertise in running large decentralized networks cheaply.

The approach. Zennström and Friis had just built Kazaa, the peer-to-peer file-sharing network, so pushing work onto users' own machines was exactly the technical foundation they knew how to operate. They reused that lineage directly, routing calls and presence through "supernodes," ordinary user machines with spare bandwidth and a public IP.

How it solved it. Because media relay, routing, and NAT traversal ran on users' computers rather than Skype's servers, marginal cost per user was near zero and the network scaled to tens of millions without telco infrastructure. The founders were not learning P2P on the job; they were applying the specific architecture they had already proven at scale with Kazaa.

Business Model: Free calls build the network, paid calls off it make the money

The problem. If Skype-to-Skype calls were free forever, and they had to be free to drive the viral loop, the company still needed revenue to survive and justify its valuation. The growth engine and the money engine appeared to be in tension.

The approach. Skype ran a freemium model that kept the two engines separate: free on-network calls remained the network-building asset, while SkypeOut let users pay to dial regular landlines and mobiles and SkypeIn let them receive calls from them. The paid tiers monetized the long tail of off-network calling without ever taxing the free behavior that fueled growth.

How it solved it. Revenue climbed with the model, from roughly $7 million in 2004 toward a projected figure in the hundreds of millions by 2006, even as the user base ran far ahead of it because the free product was deliberately the point. eBay paid approximately $2.6 billion upfront in 2005 on the strength of that base, and Microsoft paid $8.5 billion in 2011.

Moats: A login is not a social graph

The problem. Skype's network looked like an unassailable moat, but it was architected for an always-on desktop and broadband world. The supernode design did not map cleanly onto battery- and bandwidth-constrained phones, and, more fundamentally, users' identities lived inside Skype rather than in a graph Skype controlled.

The approach. Microsoft tried to defend it by re-centralizing the architecture, moving supernodes into Azure, and later reframed Skype as a feature to bundle while investment shifted to Teams. But the defense addressed the plumbing, not the graph: switching away carried almost no cost because a user's real contacts lived in the phone's address book, iMessage, and WhatsApp.

How it solved it. It did not. Mobile-native rivals delivered "free calls to your contacts" with better UX on top of graphs people already used, and Skype's standalone network lost its pull. Monthly active users collapsed to roughly 36 million by 2023, and Microsoft retired Skype on May 5, 2025, folding remaining users into Teams, about 21 years after launch.