What Is the Network Effect, and Why Does Bitcoin Have One?

Some things are worth exactly as much as the number of people who also have one. A fax machine. A language. A currency. Hand someone the only fax machine on Earth and you have handed them a heavy object with a cable.

This is not a quirk of technology. It is the defining property of anything whose purpose is to connect people, and money is the purest example there is. Bitcoin began in exactly this position, with one participant and nothing to connect to.

The First User Owns Nothing

On 3 January 2009, the Bitcoin software produced its first block. At that moment the network had one participant, one computer and a set of rules nobody else had agreed to. Whatever else it was, it was not money. A currency that one person accepts is a private note to self.

Nine days later, on 12 January 2009, ten bitcoin moved from Satoshi Nakamoto to the developer Hal Finney in block 170. That single transaction changed the category of the thing. It was now a network with two participants, and for the first time the rules described a relationship between people rather than a program running alone.

The economist Carl Menger had described this logic in 1892, more than a century before anyone wrote a line of Bitcoin code. Money, he argued, does not emerge because a government declares it. It emerges because a good is more saleable than the alternatives, meaning its holder can reliably dispose of it at a sensible price. Saleability is not a property of the object. It is a property of everyone else, and it sits underneath every one of the characteristics that make money work.

You Already Know This Pattern

You have lived through this several times without naming it.

A messenger app with one user is useless, because there is nobody to message. With two users it becomes marginally useful. With everyone you know on it, leaving becomes the expensive option, because leaving means leaving them.

A telephone in 1878 was close to worthless. A telephone in 1970, when 87 percent of American households had one, was infrastructure.

The internet in 1993 reached roughly 10 percent of American households and was mostly a tool for researchers. By 2016 it reached 88 percent, and being offline had become the unusual state.

In each case the thing itself barely changed. What changed was how many other people were reachable through it.

Why the Value Does Not Grow in a Straight Line

Here is the part that makes networks behave differently from ordinary products.

Add one user to a network and you do not add one unit of usefulness. You add a connection to every user who was already there. The number of possible two-way connections in a network of n participants is n(n−1)/2, and that number climbs much faster than n itself.

Participants	Possible connections
2	1
5	10
10	45
100	4,950
1,000	499,500

Doubling the participants from 100 to 200 does not double the connections. It roughly quadruples them, from 4,950 to 19,900.

This observation is usually called Metcalfe's law, after the engineer Robert Metcalfe, who sketched it in 1980 while selling Ethernet hardware. Two details are worth knowing, because they are almost always left out. Metcalfe's original version counted compatible communicating devices, not users. And the name was not his: the writer George Gilder attached it to him in a 1993 magazine column.

Whether real networks actually gain value at that rate is disputed among researchers, and we take that argument apart with the underlying data in our analysis of where network effects come from and how you prove them. What is not disputed is the direction. Connections grow faster than members, and that is what makes networks behave the way they do.

How Fast This Can Happen

The growth is not steady. It is slow for a long time, then sudden, then it flattens out as the network runs out of people to add. Sociologists call the resulting shape an S-curve, and it shows up in technology after technology.

The measured numbers make the point better than the description does.

• Internet use worldwide went from 15.6 percent of the population in 2005 to 73.6 percent in 2025. The ITU counts roughly 6 billion people online in 2025, with 2.2 billion still offline.
• Mobile phone subscriptions worldwide went from 0.21 per 100 people in 1990 to 8.12 in 1999, then to 76.6 in 2010 and 111.5 in 2025. There are now more subscriptions than people.
• WhatsApp reported over 450 million monthly users when Facebook acquired it in February 2014, 1 billion in February 2016, 2 billion in February 2020 and more than 3 billion in April 2025. That last figure is about 36.5 percent of everyone alive, and roughly half of everyone who is online at all.

Look at the middle of any of those curves and the growth looks unstoppable. Look at the first fifteen years and it looks like nothing is happening. Both views are of the same network.

Two Kinds of Network Effect

Economists separate two mechanisms here, and Bitcoin has both. The distinction comes from a 1985 paper by Michael Katz and Carl Shapiro.

A direct network effect means more users make the network itself more useful. For Bitcoin this is liquidity and acceptance: with more participants there is more depth to trade into, more places that will take the asset, and a smaller gap between what a buyer pays and a seller receives.

An indirect network effect means more users attract more of the things built around the network. For Bitcoin this is the surrounding industry: wallets, exchanges, custodians, hardware manufacturers, payment processors, auditors, developers. None of these existed in 2009. Each one that appears makes the network easier to use, which brings more users, which attracts more builders.

The two feed each other, which is why network effects are described as self-reinforcing rather than merely additive.

Why One or Two Winners Take Almost Everything

If joining the largest network is the rational move for each individual, the outcome for the market is concentration. This is not a theory. It is what the data from the last two decades shows.

In the third quarter of 2010, seven platforms were still selling smartphones in measurable numbers. Gartner counted Symbian in front with 36.6 percent of units sold, Android at 25.5 percent, Apple's iOS at 16.7 percent, BlackBerry at 14.8 percent, Windows Mobile at 2.8 percent, Linux at 2.1 percent and everything else at 1.5 percent. Symbian was the incumbent. Android had come from 3.5 percent a year earlier.

In August 2026, StatCounter measured Android at 67.61 percent of worldwide mobile operating system use and iOS at 32.36 percent. Everything else on Earth shares the remaining 0.03 percent. Symbian does not appear at all.

Those two measurements are not the same kind of number, and the difference is worth keeping in view: Gartner counted devices sold in one quarter, while StatCounter counts devices actually browsing today. They answer different questions. What they agree on is the shape of the outcome.

Search behaves the same way, and for a related reason. A search engine gets better as more people use it, because every query and every click is training data for the next answer. In August 2026, StatCounter measured Google at 91.1 percent of worldwide search, Bing at 4.5 percent, and every other search engine on Earth sharing the remaining 4.4 percent.

Neither outcome is caused by network effects alone. Capital, default placement, patents and distribution deals all did work here, and it would be dishonest to pretend otherwise. What the network effect explains is why the position is so hard to attack once it exists: a challenger has to be better than the incumbent and better than the incumbent's user base, and the second part is not something engineering can fix.

Bitcoin's market shows the same shape, with an important qualification. CoinGecko lists 19,445 coins, and Bitcoin accounts for roughly 58 percent of the combined value of all of them, a figure usually called Bitcoin dominance. That is the number to use for the market as a whole, because it counts everything, including Ethereum at 295 billion dollars and every other large platform token.

The much-quoted figure of 99 percent refers to something narrower: Bitcoin's share of the assets that are held specifically as a store of value. Measured against Bitcoin Cash, Litecoin and Dash it is 99.40 percent. Add Dogecoin and it is 98.60 percent. Add Monero and Zcash and it is 97.17 percent. That comparison deliberately leaves out Ethereum and Solana on the grounds that they are platforms rather than savings vehicles, which is a defensible distinction and also a convenient one. Anyone quoting the 99 percent owes you the basket it was calculated from.

How Big Bitcoin's Network Is Today

These are the measurable sizes of the network as of 1 and 2 September 2026.

Measure	Value
Addresses holding a balance above zero	56,789,267
Addresses active on a given day	677,321
Reachable public nodes	25,689
Bitcoin in circulation	20,078,253
Age of the network	17.66 years
Share of all cryptocurrency value	roughly 58 percent

One row deserves a second look. In January 2011, 65,813 addresses held a balance. Today it is 56,789,267. That is a factor of 863 in under sixteen years, and the count has fallen in exactly one of those years. Between January 2018 and January 2019 it dropped by 4.4 million, or 16.6 percent, as the addresses created during the 2017 mania were emptied and consolidated. Every other year it rose, including through the 2022 collapse: the price fell 65 percent that year and the address count still gained 9.2 percent.

Daily active addresses did not behave that way at all. They peaked above one million in January 2021, fell, recovered, and sit at 677,321 today. Activity follows the price cycle. Participation does not.

These figures describe the base layer only. Bitcoin's payment layer, the Lightning Network, is a separate network with its own participants and its own growth story, and that story currently looks different.

Network Effects Are Not Forever

This is the part that gets left out of most explanations, and leaving it out makes the concept sound like a law of physics. It is not one.

The American telephone network reached 41 percent of households by 1929 and then lost a quarter of them during the Great Depression, falling to 31 percent by 1934. Economic collapse turned out to be stronger than the network effect.

The same network later reached 95 percent of American households in 2002. By 2018 it was at 41.7 percent. It lost more than half its users in sixteen years, not because it got worse but because something with better properties arrived and people moved.

Desktop operating systems tell a milder version of the same story. Windows was once quoted at around 90 percent of the market, and that figure is still repeated. In August 2026, StatCounter measured Windows at 62.67 percent, Apple's two macOS entries at 26.91 percent combined and Linux at 8.88 percent. The lock-in is real. It is also eroding.

It is worth naming what did the replacing. The landline did not lose to a better landline. It lost to mobile phones, a network that went from 0.21 subscriptions per 100 people worldwide in 1990 to 111.5 in 2025. The two curves cross because one network could do something the other could not, and households moved one at a time until the older network was the unusual choice.

That symmetry is the part most explanations skip. Every network in this article is both a possible incumbent and a possible challenger, depending on which decade you look at. The telephone displaced the telegraph and was then displaced by the mobile phone. Symbian led a market it went on to lose entirely.

What It Would Take to Replace a Money

The obvious next question is what this means for money, and it is worth asking mechanically rather than as a prediction.

Money is a network, which makes today's national currencies and gold the incumbents. A challenger would need two things. It would need properties the incumbent cannot match, which is the same condition mobile phones met against the landline. And it would need to overcome a switching cost that is far higher than anything else in this article.

That second condition is where money is genuinely different. Leaving a messenger costs you the effort of persuading your contacts. Leaving a money means persuading everyone you buy from, everyone who pays you, your bank, and eventually your tax authority. The thing being carried by a money network is the agreement itself, not a message about something else.

This is also why the change, when it comes, tends to be generational rather than gradual. The United States ended the dollar's convertibility into gold in 1971, and the monetary system everyone now treats as normal is younger than most of the people using it. We follow what that did to saving in Impatience Is a Monetary Policy. Whether Bitcoin becomes the next such change is not a question this article can answer, and anybody who tells you they know is describing their confidence rather than the evidence.

None of this makes the network effect weaker as an idea. It makes it a force rather than a guarantee, which is the more useful way to hold it, and it cuts in both directions at once.

What This Means for Bitcoin

Bitcoin is a network whose product is agreement. Its usefulness comes from how many people accept it, how deep the market is, and how much has been built around it. That places it in the same category as the telephone, the internet and the messenger on your phone, and it explains why the first years looked so unpromising and why the later years did not.

It also sets the limits of what the idea can tell you. The network effect explains why participation and usefulness reinforce each other. It does not tell you what any of it should cost, and anyone who converts it into a price target is adding assumptions that the concept itself does not contain.

Two things are worth carrying out of this article. Networks gain value from the connections between their members, not from the members alone, which is why their growth bends rather than climbs in a line. And no network is safe from a better one, which is why the interesting question is never whether Bitcoin has a network effect. It plainly does. The interesting question is what would have to be true for something else to break it.

That question is the subject of a separate essay, Nobody Was Watching. That Is Why It Worked., which argues that Bitcoin's starting position cannot be copied and asks who actually holds the network today.