The History of the Internet: From ARPANET to the World Wide Web

The name says it all: inter-net, between networks. Long before it became the pages we open in a browser, the internet was a university research experiment for years, then a set of shared technical rules, and only at the end the place where the Web lives. Following these three steps in order helps explain why they still get mixed up so often.

A research project, not a single invention

In 1962, J.C.R. Licklider, a researcher at MIT, put in writing, across a series of memos, the idea of a “Galactic Network”: a set of connected computers that would let anyone access data and programs from any point in the network. It was an idea, not yet a working network.

The technical groundwork arrived soon after, from several directions at once. In July 1961, Leonard Kleinrock published the first theoretical basis for packet switching, the method that would let a message be split into many independent pieces instead of requiring a dedicated line. In 1965, Lawrence Roberts connected two distant computers, one in Massachusetts and one in California, over a low-speed dial-up telephone line, showing that different machines could work together at a distance. Around the same time, without either group knowing about the other, similar research on packet switching was also underway at RAND Corporation with Paul Baran and at the UK’s National Physical Laboratory with Donald Davies: the same idea took shape, among several parallel efforts, in more than one place.

Roberts presented the plan for ARPANET in 1967. The project became reality in September 1969, when the company BBN installed the first Interface Message Processor at UCLA: Kleinrock’s lab became the network’s first node, with the Stanford Research Institute as the second node shortly after. By December 1970, a working group led by S. Crocker had completed the first shared host-to-host protocol, the Network Control Protocol (NCP): ARPANET’s computers finally had a common language.

From linking two computers to a network of networks: TCP/IP is born

NCP worked well inside ARPANET, but it wasn’t designed to connect different networks to one another. Robert Kahn, who joined DARPA in 1972, took on the problem: what was needed was an “open” architecture, able to let networks built on different technologies talk to each other without any of them having to adapt to the rest.

Kahn worked on the project together with Vint Cerf, and in September 1973 the two drafted the first version of the protocol, presented at a meeting of the International Network Working Group at the University of Sussex. In May 1974 they published the paper that first described TCP/IP. The protocol became a US defense standard as early as 1980, and January 1, 1983 brought the decisive moment, remembered as “flag day”: every ARPANET host switched, on the same day, from NCP to TCP/IP. By 1990, TCP/IP had displaced competing network protocols nearly everywhere in the world, and ARPANET itself, having served its purpose, was retired.

From this point on, “the internet” stopped being the name of a single network and became what it still is: independent networks, university, corporate, government, that can talk to each other because they share the same set of technical rules.

Practical example: when a European university connected to this network of networks for the first time in the 1980s, it didn’t have to rebuild anything from scratch. Its computers just had to speak TCP/IP: from that moment on, they could exchange data with any other network that had made the same choice, anywhere in the world.

Among the first applications to run on this infrastructure, file transfer and remote login, it was probably email that had the deepest impact, according to the Internet Society document cited as a source: for the first time, people far apart could exchange messages without a phone call or a letter, and that changed the way people collaborated at a distance.

The World Wide Web: a service built on top of the internet

The Web arrived nearly twenty years after ARPANET’s first node, and it was born in one specific place: CERN, in Geneva. In March 1989, researcher Tim Berners-Lee submitted a proposal, “Information Management: A Proposal”, for linking CERN’s information — at the time scattered across incompatible systems — through hypertext. In May 1990 he presented a revised version, and in November 1990, together with colleague Robert Cailliau, he filed a formal project-management proposal titled “WorldWideWeb: Proposal for a HyperText Project”.

By Christmas 1990, Berners-Lee had already defined the Web’s three technical pillars — HTML to write pages, HTTP to transfer them, and URLs to address them — and had written the first browser/editor and the first server software. The first website, hosted on a CERN NeXT computer at info.cern.ch, described the project itself. In March 1991 a simple, text-only browser reached users on CERN’s own computers, and on August 6, 1991, Berners-Lee announced the software on several internet newsgroups: from there, interest spread beyond the community of physicists.

The turning point for its spread came on April 30, 1993, when CERN publicly declared the Web free of royalties, guaranteeing its status as an open standard that anyone could use and build on without paying licensing fees. By the end of that same year, more than 500 known web servers existed worldwide, even though Web traffic still accounted for only about 1% of all internet traffic: other services still carried most of the load in those years.

Year	Milestone
1962	Licklider describes the concept of a global network
September 1969	ARPANET’s first node, at UCLA
May 1974	Cerf and Kahn publish the TCP/IP design
January 1, 1983	ARPANET switches from NCP to TCP/IP: the internet as we know it begins
March 1989	Berners-Lee proposes the World Wide Web at CERN
April 30, 1993	CERN places the Web in the public domain
April 1995	The NSF ends public funding for the US internet backbone

How it works: packets, addresses, names

Underneath everything is still the idea of packet switching: every piece of information you send, a photo, a message, a page request, gets split into smaller packets, and each one carries the address of its destination. That address is the IP address, a sequence of numbers like 142.250.190.78; routers, as MDN Web Docs explains, read the address on each packet and forward it toward the next network, one hop at a time, following a routing algorithm that works out the best path each time.

Inside a single local network, another layer of addresses, MAC addresses, normally assigned by the hardware manufacturer, lets a switch deliver packets to the right device, learning which connection each message arrived from. An ISP (Internet Service Provider) is the company that runs the routers able to connect your local network to other operators’ networks, all the way to any other network in the world.

Nobody, though, memorizes number sequences like 142.250.190.78. That’s what the DNS is for: the system that translates human-readable names, like “google.com”, into the matching IP address.

Practical example: when you type an address into the browser’s address bar and hit Enter, your device first asks the DNS which number matches that name; it then opens a connection to the machine answering at that number and requests the page; the reply comes back split into packets that travel through different routers; the browser puts them back in order and renders the page. All of this, typically, in under a second.

One distinction is worth stressing, since it’s the one most often confused: the internet is not the Web. The internet is the network of networks just described, made of cables, routers and protocols; the Web is one of the services running on top of it, made of pages linked together and viewed with a browser over the HTTP protocol. Email, messaging and video streaming are other services that travel over the same infrastructure without going through the Web at all. Even pages built with server-side languages such as PHP remain, technically, part of the Web: they are programs that generate HTML pages to send to the browser over that same packet network.

How many people use the internet worldwide

Estimates vary depending on who calculates them and which period they cover, and the two main ones should always be cited separately. According to the ITU, the United Nations agency for telecommunications, about 6 billion people — 74% of the world’s population — used the internet in 2025, up from 5.8 billion in 2024; growth in 2025 ran at 3.3%, faster than the 2.9% recorded the year before. The gap remains wide: penetration reaches nearly 94% in high-income countries against 23% in low-income countries, and by region it ranges from 88-93% in the Commonwealth of Independent States, Europe and the Americas down to 36% in Africa.

DataReportal, which publishes estimates using a different method, calculated that at the start of 2026 there were 6.04 billion internet users worldwide, or 73.2% of the population, an increase of 294 million people (+5.1%) over the previous twelve months; about 2.21 billion people remained offline. The two figures shouldn’t be added together or blended into one: they measure different periods with different methodologies, but they point to the same underlying trend — growth that slows down as it approaches the size of the world’s population.

What the internet is used for

Email remains, according to the Internet Society document cited as a source, probably the service that changed remote collaboration the most, going back to ARPANET’s earliest years. From there, its uses have multiplied considerably. At the start of 2026, according to DataReportal, 5.66 billion user identities were active on social media worldwide (68.7% of the population), with the average person using close to 7 different platforms a month; 91.7% of online adults watch streaming television regularly, and streaming services now account for more than half of all TV viewing time worldwide. On the advertising side, global advertising spending in 2025 came to about $1.16 trillion (up 6.5% from the year before); digital channels accounted for 74.4% of that total.

A more recent field, running on that very same infrastructure, is artificial intelligence platforms: more than 1 billion people use them every month according to DataReportal, with generative AI tools processing requests on remote servers reached through exactly the same packets, IP addresses and routers as any other service: what changes is the service running on top, not the transport underneath, the same transport that has connected every computer in the history of computers since TCP/IP existed.

Still, there are the nearly two billion people the ITU estimates were offline in 2025: internet access is far from a given everywhere, and it remains closely tied to a country’s economic development.