Few inventions are as hard to date as the network most of us use every hour, which is why the question of how the internet evolved has no single answer. There was no launch day, no inventor in a lab coat, no moment when a switch flipped. A research experiment among a handful of American universities slowly stopped being an experiment, picked up new rules, and then picked up several billion users. This article follows that path from the 1960s to the present, separating what the historical record supports from the neat stories that tend to get told about it.

What was ARPANET, and why was it built?

ARPANET was a research network funded by the US Defense Department’s Advanced Research Projects Agency. Larry Roberts presented the proposal in October 1967, and the contract to build it was awarded in 1969. The first link, between UCLA and the Stanford Research Institute, went live on October 29, 1969, at about 10:30 in the evening. According to the usual account, the system crashed while the team was typing the word LOG, the start of “login”. By December of that year the network had four nodes, the fourth at the University of Utah.

The motive is where popular history drifts. A common story says ARPANET was designed to survive a nuclear attack. Historians generally describe something plainer: expensive computers sat at a few universities, and researchers wanted to share them without travelling. The survivability idea belongs to separate work on distributed communications, and the two threads got tangled together later. Treat the nuclear-war version as folklore, not established fact.

Who invented packet switching?

Nobody owns it alone. Donald Davies at Britain’s National Physical Laboratory conceived packet switching in 1965, and the name stuck because he used it. The core idea is simple enough to state in a sentence: chop a message into small chunks, label each one, and let the chunks find their own way through the network to be reassembled at the other end.

That differs sharply from the telephone system, which reserved a dedicated circuit for the length of a call. Packets share lines, so a busy link can carry pieces of many conversations at once. It was a more efficient use of expensive wires, and it is still how the network works today.

Paul Baran, working in the United States around the same time, developed related ideas about distributed networks. Which of these men deserves the most credit is a question historians still argue about, and it depends on whether you weight the concept, the name, or the first working implementation.

When did TCP/IP and the word “internet” arrive?

By the early 1970s there were several separate networks, and none of them could talk to each other. Vint Cerf and Robert Kahn published their design for connecting them in May 1974. The first attested use of the term “internet” appears in a document from December 1974, RFC 675, as shorthand for networks joined together.

The protocol was split into two layers in 1978: TCP to handle reliable delivery, IP to handle addressing and routing. RFC 791 described IPv4 in 1981, when the whole network counted about 213 hosts. IPv4 went onto the ARPANET itself in January 1983. Many retellings give a precise day for that switchover, but the sources I checked do not, so I will leave it at the month.

This is the real birth of the internet, if you insist on one. Not a machine, but an agreement: a common set of rules that any network could adopt without asking permission.

How did NSFNET turn a research tool into public infrastructure?

The ARPANET itself stayed small and tightly controlled. The larger leap came from the US National Science Foundation, which created NSFNET in 1986 as a backbone running at 56 kilobits per second. That is roughly a thousand times slower than a modest home connection today. It was upgraded to 1.5 megabits per second in 1988 and to about 45 megabits per second in 1991.

The ARPANET was decommissioned in 1990. By then it had been handed to the Defense Communications Agency (in 1975) and outgrown by the networks that sprang up around it. Universities, labs, and eventually companies connected to regional networks that fed into the NSFNET backbone.

There was one catch. The backbone came with rules that restricted commercial traffic. Those rules would not last.

What is the difference between the internet and the Web?

People use the words interchangeably, and they should not. The internet is the global system of interconnected networks, the plumbing. The World Wide Web is a collection of documents and other resources linked by hyperlinks and addresses, reached over the internet using HTTP or HTTPS. Email, online games, and video calls use the internet without using the Web in the traditional sense.

Tim Berners-Lee submitted his proposal to CERN in March 1989 without giving the system a name. By the end of 1990 he had a working browser and server. The technology was released outside CERN in January 1991 and announced to the wider internet that August. A decisive step came on April 30, 1993, when CERN made the protocol and code available royalty-free and placed the Web in the public domain.

Later that year the NCSA released Mosaic, a graphical browser that could show images inline with text. Popularity grew quickly after that. A network built for researchers now had a face that anyone could use.

How did the internet go commercial?

Commercial providers appeared before the rules officially allowed them. The first commercial dial-up provider in the US, a company called The World, opened in 1989. In 1992 a law, the Scientific and Advanced-Technology Act, allowed NSFNET to interconnect with commercial networks. On April 30, 1995, the NSF ended its sponsorship of the backbone, which removed the last restrictions on commercial traffic. Private carriers including MCI, PSINet, and Sprint took over the job; by 1996 Sprint was the largest carrier of internet traffic.

One statistic captures how fast the shift happened. The internet carried about 1 percent of two-way telecommunications information in 1993, about 51 percent in 2000, and more than 97 percent by 2007. The phone network did not disappear. It was absorbed, and today much of it is itself carried as data.

The dot-com bubble ran from about 1997 to 2001. Many of the companies failed, yet the infrastructure they paid for stayed in the ground, and the habits they introduced (shopping, searching, messaging online) did not go away either.

How did broadband, mobile, and social platforms change it?

Dial-up gave way to broadband, which made always-on connections ordinary. The term “Web 2.0” gained prominence in 2004 to describe sites built around user contributions, such as blogs, wikis, and social networks, rather than pages published by a few.

Phones did the next round of work. NTT DoCoMo launched the first commercial 3G network on October 1, 2001, after a full mobile internet service in Japan in 1999. Apple announced the first iPhone on January 9, 2007 and released it in the US on June 29 of that year. That first model lacked 3G and could not install third-party apps; the App Store followed in July 2008. Even so, a pocket-sized screen with a real browser moved the internet from desks to everywhere.

The numbers now are large. The International Telecommunication Union estimated that about 6 billion people used the internet in 2025, roughly three quarters of the world, up from a revised 5.8 billion in 2024. Roughly 2.2 billion remained offline. The gap is wide by income: use was about 94 percent in high-income countries and 23 percent in low-income ones, and 85 percent in cities against 58 percent in rural areas.

Is the internet running out of addresses?

In a sense, it already did. IPv4 addresses are 32 bits long, which allows 4,294,967,296 of them. The last block of IPv4 addresses was assigned at the regional registry level in January 2011. IPv6, with 128-bit addresses, offers 2 to the power of 128, a number so large that exhaustion is not a practical worry.

The transition has been slow, and the source I read expects it to keep taking a long time. Networks cope with workarounds, notably sharing one public address among many devices. This is a good example of how the internet changes: nobody orders an upgrade. Each operator moves when it suits them, and the old and new run side by side for years.

What is still unsettled?

Several questions are open, and they are mostly about people rather than protocols. Who pays for the backbone as video and cloud traffic grow? How much should governments regulate platforms built on top of a network designed to be neutral? Can the 2.2 billion people still offline get affordable access? The ITU notes that mobile data prices remain out of reach in about 60 percent of low- and middle-income countries.

There is also the security legacy. The early protocols assumed that everyone on the network was a colleague. Retrofitting trust onto a system built without it is a project that has run for decades and has no finish date. Forecasts about where all this goes (satellite constellations, AI-generated content, new interfaces) are opinion, not history, and they should be read as such.

Common questions

What was the very first message sent over ARPANET?

It was sent on October 29, 1969, from UCLA to the Stanford Research Institute. The system crashed while the word LOG was being typed, so the first message was a fragment of “login”.

Did Al Gore or anyone else “invent” the internet?

No single person did. Packet switching came from Davies and Baran, the protocols from Cerf and Kahn and many collaborators, and the Web from Berners-Lee. Governments, universities, and companies all paid for pieces of it.

Is the internet the same thing as Wi-Fi?

No. Wi-Fi is a way of connecting a device to a local network by radio. That local network may or may not be connected to the internet.

How many people use the internet?

The ITU estimated about 6 billion users in 2025, around three quarters of the world’s population, with about 2.2 billion people offline.

Who runs the internet?

No one entity does. Thousands of independent networks agree to exchange traffic using shared protocols, and standards bodies publish the specifications. Governments regulate what happens within their borders.

What to remember

The internet grew from an agreement, not an invention. ARPANET proved that packet switching worked in 1969; TCP/IP, adopted on the ARPANET in January 1983, let different networks join; NSFNET carried it through the late 1980s; and the end of NSF backbone funding in April 1995 opened the door to commerce. The Web, a separate layer on top, made it usable by people who had never heard of a protocol.

Be careful with the tidy origin stories. The nuclear-survival motive is more legend than documented purpose, and the exact date of the TCP/IP switchover varies from one retelling to another.

Further reading

A well-known narrative history of ARPANET and the people who built it. See it on Amazon (affiliate link)

As an Amazon Associate I earn from qualifying purchases. A commission may be earned if you buy through this link, at no extra cost to you.

Sources