Ray Tomlinson sits in a computer lab in 1971... about to send the first email.
He types a message between two machines sitting three feet apart. Later, he can't remember what the message said. Probably something forgettable, he admits. Something like "QWERTYUIOP."
But that throwaway test? It changed the fundamental economics of communication.
Email had a property Tomlinson didn't fully understand yet. It got more valuable the more people used it. One person with email? Useless. Two people? Barely interesting. A million people? Now you've got something that reorganizes how humans coordinate.
That's a network effect. And it's one of the weirdest value-creation mechanisms we've stumbled onto.
Most things in economics follow simple math. You make a chair... it's worth something. You make ten chairs... they're worth ten times as much. Linear. Predictable.
Networks don't work that way.
Robert Metcalfe figured this out in 1980 when he was thinking about Ethernet — the technology he'd co-invented for connecting computers. He proposed something that sounded almost too good to be true. The value of a network grows proportional to the *square* of its users.
Not just proportional. The *square*.
So if you double your users from ten to twenty? You don't go from ten units of value to twenty. You go from one hundred... to four hundred. Double the users, quadruple the value.
This became Metcalfe's Law. And it explains why some companies become unstoppable... while others with better products vanish.
Think about Theodore Vail in 1908. He's running AT&T, and he's got a problem. Lots of small telephone companies are competing, each with their own network. If you're on the Bell system, you can't call someone on a rival system. The country has fragmented telephone islands. You might have three phones in your house just to reach different networks.
Vail realizes something nobody else is saying out loud. The telephone isn't valuable because of the technology. It's valuable because of who else has one.
So he starts pushing interconnection. One big network instead of fragmented ones. His annual reports from this period are wild. He's basically arguing that AT&T *should* be a monopoly. Not despite competition being good... but because in telephony, competition is waste.
He's not wrong. He understands that in a networked world, the biggest player wins. Not because they're better. Because they're *bigger*.
By 1939, AT&T controls ninety-three percent of American phone lines. The math did exactly what Vail predicted.
And that's where things get uncomfortable.
Here's the pattern that keeps repeating. A network starts small. It's actually kind of bad at first — clunky, limited, missing features. But it reaches some threshold of users... and suddenly it tips. New users flood in because everyone else is there. The network gets more valuable. Which attracts more users. Which makes it more valuable.
It's a feedback loop that doesn't stop until it dominates.
Economists call this "winner-takes-all" dynamics. And the statistics are brutal. In a 2001 study of technology markets, researchers found that ninety percent of the value created by network effects concentrates in the top ten percent of networks. Not evenly distributed. Concentrated.
The second-place network in a winner-takes-all market? Typically captures less than ten percent of the value of the leader. Not forty-nine percent. Single digits.
eBay figured this out in 1995. Pierre Omidyar built a marketplace connecting buyers and sellers. But here's the trick. It's not one network effect... it's two, running simultaneously. More buyers attract more sellers. More sellers attract more buyers.
Jean-Charles Rochet and Jean Tirole studied these "two-sided markets" in the early 2000s and found something surprising. The platform doesn't just facilitate transactions. It becomes the gravitational center. Their math showed that platforms can actually afford to lose money on one side of the market... if it attracts the other side.
Which is why eBay could charge sellers fees but let buyers browse free. You're subsidizing one group to create value for everyone.
Try starting a competing auction site when eBay has millions of buyers and sellers. You're not competing on features. You're competing against the accumulated value of all those connections.
In 2008, several well-funded competitors tried. By 2010, they were gone. eBay's market share barely moved.
So here's the turn. If network effects are so powerful... why doesn't every company that gets big stay big forever?
Clubhouse looked unstoppable in 2020. Social audio app. Invite-only. Everyone from Elon Musk to Oprah showing up in rooms. At its peak in February 2021, it hit ten million weekly active users. The network effects seemed perfect. More interesting people attracted more listeners, which attracted more interesting people. Classic feedback loop.
Then... 2021 happened. Users started leaving. The magic faded. By 2022, it was a cautionary tale. Weekly active users dropped ninety percent.
What went wrong?
This is where the theory gets messy. Metcalfe's Law assumes all connections are equally valuable. They're not.
David Reed proposed something different in 1999 — Reed's Law. He argued the real value comes from subgroups forming within networks. Communities. Tribes. The ability to find your people. His math was even more explosive than Metcalfe's. The value grows exponentially with the number of possible subgroups.
Sounds great. But here's the problem. You can't just throw millions of users together and expect magic. Sometimes more users means more noise. More spam. More content you don't care about.
Network congestion, some researchers call it. The network doesn't just stop growing in value. It starts *losing* value.
Clubhouse got too big too fast. The rooms filled with crypto bros and life coaches. The people who made it magical in the first place? They left. And when they left, everyone else followed.
And there's another thing nobody talks about enough. Most networks run on a tiny fraction of their users. In 2006, researchers at HP Labs analyzed Wikipedia and found that just one percent of users created seventy percent of the content. On Reddit and Twitter, the ratio is even more extreme. One percent of users generate the majority of posts and comments.
One percent. Everyone else is watching.
So the network effect isn't really about having millions of users. It's about having enough of the *right* users. The creators. The super-users. The people who make everyone else want to show up.
Lose them, and you lose everything... even if your total user count looks fine.
Facebook understood this when they expanded beyond Harvard in 2004. They didn't just open the floodgates. They moved university by university. Created density at each school before moving to the next. Mark Zuckerberg was obsessive about this. A school needed at least sixty percent adoption before he'd consider it successful.
Because a social network where none of your friends are? That's just a website with your photos on it. But a social network where everyone you know is? That's infrastructure. That's where you coordinate your life.
By 2010, they'd built something that was genuinely hard to leave. Not because the product was so great. Because everyone you needed to reach was there.
When researchers asked people why they stayed on Facebook despite privacy concerns, the most common answer wasn't "I like it." It was "I have to."
Uber did the same thing city by city starting in 2014. More drivers meant shorter wait times for riders. More riders meant more income for drivers. The cross-side effects reinforced each other.
But Uber's network effects are local. Being the biggest ride-sharing app in San Francisco doesn't help you in Tokyo. The network effects don't travel. Which is why Uber had to fight market by market, burning billions of dollars to reach critical mass in each city.
Between 2014 and 2016, they lost four and a half billion dollars. Not million. *Billion*.
They were buying network effects. Subsidizing rides to get enough drivers and riders that the loop could sustain itself. Network effects are powerful... but they're not magic. They're expensive to build.
And this is where it gets strange. We're starting to see network effects show up in places that don't look like traditional networks.
Machine learning models get better as more data flows through them. The algorithm improves. Which attracts more users. Which generates more data. Which improves the algorithm. Same feedback loop... just in code instead of connections.
Google's search got better not because their engineers were smarter, but because billions of searches taught the algorithm what people actually wanted.
Bitcoin and Ethereum run on network effects too. The more people mining and transacting, the more secure and valuable the network becomes.
Even biology works this way. In the 1970s, Robert May studied ecosystem stability and found that highly connected food webs are more resilient to shocks than sparse ones. Add or remove one species, and the ripples move through the whole system.
The Amazon rainforest creates its own weather because of network effects between trees, fungi, insects, and moisture. Remove enough trees, and the whole network collapses into savanna. It's not gradual. It tips.
Michael Katz and Carl Shapiro studied this in 1994 and found something that should worry us. Network effects naturally tend toward monopoly. Not because the dominant company is evil. Because the math pushes that direction.
They called it "excess inertia." Once you're locked into a network, switching costs are so high that you stay... even when something better comes along.
Remember how long we all stuck with QWERTY keyboards even though Dvorak was proven faster? That's excess inertia.
And once a network achieves dominance, it's almost impossible to displace. Not with a better product. Not with better prices. You'd need to convince millions of users to coordinate a simultaneous switch. Which almost never happens.
So policymakers are asking harder questions now. Are network effects always good? Facebook, Google, Amazon — they all leveraged network effects to become essentially unchallengeable in their markets. Is that innovation? Or is that just math playing out in a way that kills competition?
Jean Tirole won a Nobel Prize in 2014 for his work on two-sided markets. He's been arguing that we need new regulatory frameworks. Because traditional antitrust law assumes competition is about prices and quality. But in networked markets, the competition happens early. Once someone wins, the game is over.
By the time regulators notice a problem, the network effects have already created a moat so wide that competition is theoretical.
Here's what keeps me up at night. We're building more of our infrastructure on network effects. Social connection, transportation, finance, information. And network effects concentrate power in ways that are really hard to reverse.
Reid Hoffman, who founded LinkedIn, understands this better than almost anyone. He built a professional network that became essentially mandatory. Not because LinkedIn is so wonderful. Because everyone you need to reach professionally is there. Seven hundred fifty million users as of 2023.
That's power. And it's power that doesn't come from making a better product. It comes from getting there first and reaching critical mass.
Hoffman's written about this openly. He calls it "blitzscaling." Move so fast that you achieve network effects before competitors can respond. Ethics and sustainability come later. First, win.
But maybe there's something useful in understanding this.
Next time you feel stuck with a platform you don't even like that much? You're not crazy. You're experiencing the math of network effects. The value isn't in the product. It's in the people.
And that means the power isn't really in the company's hands. It's in ours. Collectively.
Which is why when users coordinate and leave together — like they did with MySpace, with Digg, with Clubhouse — the network effect runs in reverse. The same feedback loop that built the empire tears it down.
MySpace lost ten million users per month in 2011. Once the exodus started, it accelerated. The math that made them unstoppable made them collapse faster.
So look at the networks you're in. Ask which ones actually serve you... and which ones you're serving.
Because network effects work both ways. You're not just receiving value. You're creating it. Every time you show up, you make that network more valuable for everyone else.
Facebook's market cap is over eight hundred billion dollars. That value didn't come from their servers or their code. It came from you posting, you commenting, you showing up.
Which means you have more leverage than you think.
The next time a platform takes you for granted, remember this. Without you and people like you, there is no network effect. There's just expensive servers and empty rooms.
And here's the thing that should actually scare companies: we're getting better at coordinating. Better at leaving together. The same tools they built to create network effects? We're learning to use them to dissolve network effects.
One collective decision at a time.