How did the internet become a global network?

The answer in one minute

The internet began in 1969 as ARPANET, an American research experiment connecting four university computers. The decisive invention came in the 1970s: TCP/IP, a set of rules that let different networks exchange small packets of data — a network of networks, with no centre and no owner.

Universities built on it in the 1980s; the World Wide Web, invented at CERN in 1989 and given away for free in 1993, made it usable by anyone; commercial carriers took over the plumbing in 1995 and undersea fibre-optic cables stitched the continents together. Because joining required no one’s permission — only speaking the shared protocol — the network grew wherever wires, radios and ambition existed.

Today an estimated 5.5 billion people are online. Nobody runs the internet; a loose web of engineering bodies keeps the agreements working. That openness built the modern world — and its limits, from the offline third of humanity to national firewalls, are now where the story is being written.

Computers that could not speak

In the early 1960s the computer was a building, not an appliance. Each machine was an island with its own language, its own operators, its own queue of waiting scientists. If a researcher in Los Angeles wanted to use a machine in Boston, the practical options involved an airline.

The idea that changed this seemed almost wasteful at first. Telephone networks connected two parties by reserving a circuit between them for the whole conversation — fine for voices, absurd for computers, which speak in short bursts separated by silence. Researchers — Leonard Kleinrock at MIT and UCLA, Paul Baran at RAND, Donald Davies in Britain — worked out a different scheme: chop every message into small, addressed parcels and let them find their own way through a mesh of links, sharing capacity with everyone else’s parcels.

Baran’s version had a Cold War motive — a communications mesh with no centre is hard to decapitate — but the deeper appeal was economic: packet switching let expensive links be shared, and it made the network indifferent to what the packets contained.

Four computers and a crash

The U.S. Defense Department’s research agency, ARPA, funded a network to connect the computers it was paying for at universities. On the evening of 29 October 1969, a student programmer at UCLA, Charley Kline, began to log in to a computer at the Stanford Research Institute, hundreds of kilometres away — the first message ever sent on the ARPANET. He typed “l”, then “o”, and then the receiving system crashed(Kleinrock, UCLA).

The first word the infant internet ever said, its builders like to note, was “lo”. By the end of 1969 four computers were connected; the network then grew by the month. Electronic mail arrived in 1972 and promptly became the largest source of traffic — the first hint that a machine built to share processors would mainly be used by people to reach other people.

The invention that mattered: a network of networks

The ARPANET was one network, owned by one funder, built from one kind of equipment. The 1970s produced many others — packet radio, satellite links, local networks in single buildings — each with its own conventions, each incompatible with the rest. Connecting computers had been solved. Connecting networks had not.

That was the problem Vinton Cerf and Robert Kahn set out to solve in 1974, and their answer — refined over a decade into TCP/IP — is the internet’s real founding invention. The rules are deliberately humble. Every packet carries a destination address in a common format. Special computers called routers pass packets between networks, reading only the address. The network itself promises almost nothing: packets may arrive late, out of order, or not at all, and the sender’s software is responsible for noticing and re-sending. Anything beyond delivery — reliability, secrecy, meaning — belongs to the machines at the ends, not the middle.

The internet's hourglass. Applications above and transmission technologies below can change freely — and have, repeatedly — while the narrow waist, IP, stays the same.

This minimalism is why the internet could swallow every future it met. Fibre optics, wifi, and smartphones did not exist when IP was specified(RFC 791, 1981); all three joined without asking. A network that demands little of its members can accept almost anyone — the design principle that turned a military research project into a planetary commons.

The switch

From experiment to infrastructure

1969
29 October 1969
The first ARPANET message
UCLA attempts to log in to SRI. Two letters — "lo" — cross the network before the receiving computer crashes.
1972
1972
Email takes over
Message software written for ARPANET engineers becomes the network's dominant use almost immediately.
1974
May 1974
Cerf and Kahn publish the internetworking design
Their protocol for connecting unlike networks matures into TCP/IP — the internet's founding agreement.
1983
1 January 1983
The ARPANET switches to TCP/IP
The "flag day" cutover: from now on, joining the internet means speaking IP. The domain name system follows within the year.
1986
1986
NSFNET opens the network to all of academia
The U.S. National Science Foundation builds a national backbone linking its supercomputer centres and, soon, hundreds of campuses.
1989
March 1989
The web is proposed at CERN
Tim Berners-Lee drafts a system of linked documents for physicists; by Christmas 1990 the first website runs on his desk.
1993
30 April 1993
CERN gives the web away
The web's software is released royalty-free, for everyone, forever. The Mosaic browser brings it to ordinary screens the same year.
1995
30 April 1995
The government backbone retires
NSFNET is decommissioned; commercial providers now carry the internet's traffic. The dot-com era begins.
1998
1998
Naming gets a steward
ICANN, a non-profit, takes over coordination of domain names and addresses — governance by committee rather than by government.
2000
Late 2000s
The internet goes mobile
Smartphones put the network in billions of pockets; for most people ever connected, the internet has never involved a desk.
2024
2024
5.5 billion people online
The ITU estimates 68% of humanity uses the internet — while a third, mostly in poorer and rural regions, still does not.

Two dates deserve their fame. On 1 January 1983 every machine on the ARPANET had to switch to TCP/IP on the same day — the “flag day” that turned a network into the seed of an internetwork. And on 30 April 1993, CERN put the World Wide Web’s software in the public domain, renouncing royalties permanently. One decision made the network joinable; the other made it usable — and free.

The web is not the internet

The web deserves its own telling — it is what made the internet worth joining for people who would never type a command. A physicist’s tool for linking laboratory documents became, within five years of being given away, the front door to news, shopping, government and each other. The giveaway was the masterstroke: because no one owed CERN a licence fee, every business, newspaper and teenager could build without permission, exactly as the network beneath had intended.

What “global” is physically made of

A word like “cyberspace” hides how physical the internet is. Between continents, it is glass: more than 600 active and planned submarine cables — over 1.5 million kilometres of fibre, enough to wrap the planet dozens of times — laid along the seabed by ships, following routes surveyed to avoid anchors, trawlers and fault lines(TeleGeography). Satellites matter at the edges, but they carry a fraction of one per cent of international capacity; when you video-call another continent, you are almost certainly speaking through the ocean floor.

The cables answer a common puzzle — how can a network with no owner span oceans? Because each cable has ordinary owners: telecoms, consortia, and increasingly the big content companies, each with commercial reasons to connect. The internet’s genius is that it needs no owner in common. Every wire is someone’s; the agreement to exchange packets across them is everyone’s.

Nobody in charge — on purpose

The internet’s governance is as unusual as its engineering. Its technical standards come from the Internet Engineering Task Force, a body with no membership fees and no voting, where anyone may propose a protocol and arguments are settled, in its famous credo, by “rough consensus and running code”(IETF, Tao of the IETF). Names and numbers — who gets which address, what “.org” means — are coordinated by ICANN, a California non-profit answerable to a global community of stakeholders rather than to any state.

It is worth pausing on how strange this is. The infrastructure carrying the world’s commerce, diplomacy and family photographs is held together by volunteer committees, memoranda of understanding, and the fact that defecting from the shared protocols would cut the defector off from everyone else. The arrangement has been attacked, litigated and envied for thirty years — and it has, so far, held.

Why this mattered

The internet did to distance what the railways did to geography — but for information, and in one generation. Whole industries (mail, media, retail, finance) were rebuilt on top of it; new ones (search, social media, cloud computing) exist only because of it. Its openness set the period’s economic tone: a Stanford student or a Hangzhou teacher could reach a billion users without asking a gatekeeper, an experiment in permissionless invention with no precedent in industrial history.

Why this still matters

The internet’s founding bet — that openness beats control — is now being re-tested. About a third of humanity remains offline, overwhelmingly in poorer and rural regions, making connectivity a new axis of inequality(ITU, 2024). States have learned to fence the supposedly unfenceable: national firewalls, deliberate shutdowns and data- localisation laws have pushed watchdogs like Freedom House to record more than a decade of declining internet freedom. And the traffic itself increasingly flows through a handful of private platforms and clouds — a re-centralisation of a network built to have no centre.

Whether the next billion users join the open internet or a set of national ones is a live political question. The history is the useful compass: every time the network faced a choice between a closed, controlled design and an open, humble one, the open one won — not because it was noble, but because it grew faster than anything closed could match.