Internet Invention: Who Invented the Internet and When?

The Internet was not invented by one person on one day. If the question is who designed the protocol architecture that most directly became the Internet, Vint Cerf and Robert Kahn receive the strongest attribution. If the question is when the Internet began operating on the protocol system that defines it, 1 January 1983, when ARPANET completed its planned switch from NCP to TCP/IP, is the clearest single date.

Those answers do not make 1969 irrelevant. ARPANET became operational that year and provided the environment in which much of the later work developed. Nor do they erase the packet-switching research of Leonard Kleinrock, Paul Baran, Donald Davies and others, the ARPANET leadership of Lawrence Roberts, the work of the BBN teams that built Interface Message Processors, or later protocol and coordination work by researchers such as Jon Postel. The Internet emerged because several distinct technical problems were solved in sequence.

Why the Internet Has More Than One Invention Date

Several dates often appear in accounts of the Internet because they refer to different achievements. A packet-switched network, an internetworking design, a successful test across dissimilar networks, and a mandatory network-wide protocol transition are not the same event.

Packet Switching Came Before the Internet

The Internet depends on packet switching: data is divided into manageable units that can be carried through a network and reassembled at the destination. The idea did not emerge from one laboratory. Research at MIT, RAND, and Britain’s National Physical Laboratory developed related packet-network concepts during overlapping periods, sometimes independently.

Leonard Kleinrock published early work on communication-network theory at MIT beginning in 1961. At RAND, Paul Baran developed distributed communication concepts in the early 1960s in research concerned with survivable communications. In Britain, Donald Davies and his NPL team developed their packet-switching approach in the mid-1960s. NPL dates Davies’s development of packet switching to 1965 and credits his group with the terminology that became standard. [a]

These contributions are related but should not be collapsed into a contest over a single inventor. Baran’s work addressed distributed communication under failure conditions. Davies pursued efficient digital data communication and practical packet networking. Kleinrock contributed mathematical analysis of communication networks and later led the UCLA Network Measurement Center that became the first ARPANET node. The Internet Society’s historical account notes that research at MIT, RAND, and NPL proceeded in parallel before the groups became fully aware of one another’s work. [b]

Why Packet Switching Was Necessary but Not Sufficient

Packet switching solved the problem of moving digital traffic efficiently through a network. It did not by itself solve the later Internet problem: how can one packet network communicate with another packet network when the two were built differently?

A single packet network can have its own internal rules, addresses, packet sizes, transmission media, and control system. An Internet must allow many such networks to remain technically distinct while still participating in end-to-end communication. That second problem is internetworking.

ARPANET Put Packet Networking Into Operation

ARPA’s networking program turned research ideas into a working general-purpose network. In 1968, ARPA contracted Bolt Beranek and Newman, usually known as BBN, to build the Interface Message Processors, or IMPs, that acted as packet switches for ARPANET. The first IMP was installed at UCLA in 1969. By the end of that year the initial network connected UCLA, Stanford Research Institute, the University of California at Santa Barbara, and the University of Utah. DARPA’s historical record identifies these as ARPANET’s first four nodes. [c]

ARPANET mattered because it was not merely a proposal. Researchers could use a real packet network, measure it, build host protocols, create applications, and discover where a single-network design stopped being enough. The Network Working Group completed the initial ARPANET host-to-host protocol, NCP, in 1970, and applications expanded as sites implemented it during the following years.

The First ARPANET Host-to-Host Message Was “LO”

On 29 October 1969, Charley Kline at UCLA attempted to log in to a computer at SRI. The intended word was LOGIN. The letters L and O arrived before the system failed; the full login was completed later. UCLA preserves the event through the IMP log associated with the transmission. [d]

This event is often described loosely as the first Internet message. “First ARPANET host-to-host message” is more precise. The experiment took place years before TCP/IP became ARPANET’s standard and before the open-network Internet had reached its operational form.

ARPANET Was a Network, Not Yet a Network of Networks

Early ARPANET demonstrated that computers at distant research sites could share a packet-switched communications network. Its host protocol, NCP, was designed for communication inside that environment. As other networking projects appeared, this assumption became restrictive.

ARPA was supporting packet radio and satellite networking alongside ARPANET. These systems did not have the same physical characteristics or internal operating rules. Radio links could be intermittent. Satellite links differed in delay and transmission behavior. Local networks were also developing. Requiring every network to be rebuilt as a copy of ARPANET would defeat much of the reason for having separate networks in the first place.

The Internet problem therefore became a problem of network independence: a local network should be able to keep its own technology while exchanging traffic with hosts on other networks through a common higher-level protocol system.

Cerf and Kahn Designed an Open-Network Internet

Robert Kahn began working on the problem of connecting independent packet networks and asked Vint Cerf to join the protocol design work in the spring of 1973. Cerf brought direct experience with ARPANET’s NCP. Kahn brought an architectural approach in which distinct networks would remain autonomous rather than being redesigned into one uniform network.

The Internet Society’s historical account describes several rules behind this approach: each network would stand on its own, gateways would connect networks, communication would use best-effort delivery at the network level, and there would be no single global operational controller. A first written version of the developing design circulated within the International Network Working Group in September 1973. Cerf and Kahn then published a refined version in May 1974 under the title A Protocol for Packet Network Intercommunication. [b]

This is why Cerf and Kahn are often called the fathers of the Internet. Their strongest claim is not that they invented packet switching, ARPANET, every Internet protocol, or every network that later joined the Internet. Their direct contribution was the protocol design and open-network approach that allowed unlike networks to communicate as parts of a larger internetwork.

What “Open-Network” Meant in Practice

The design separated local network technology from end-to-end communication. A host could send data through its local network, pass it to a gateway, cross another network with different internal rules, and continue toward the destination. The networks did not need to share identical hardware.

Gateways—later generally called routers—handled the boundaries between networks. They moved Internet datagrams from one network environment to the next. This separation allowed packet radio, satellite links, wired networks, local-area networks, and later many other transmission technologies to carry Internet traffic without forcing all of them to become the same kind of network.

The 1977 Three-Network Test Showed Internetworking Working

A protocol proposal becomes much more convincing when it works across the incompatible systems it was designed to connect. In 1977, a set of tests linked ARPANET, the Packet Radio Network (PRNET), and the Atlantic satellite network (SATNET) using the developing Internet protocols. The Computer History Museum describes the November test as a three-network demonstration in which packets traveled through different network technologies, including a mobile packet-radio link. [e]

The importance of the test lies in the boundaries the data crossed. ARPANET used fixed network links, PRNET used packet radio, and SATNET used satellite communication. Successful end-to-end communication across those systems demonstrated the central Internet idea: the underlying networks could remain different while a common internetwork protocol carried data across them.

For invention history, 1977 deserves more attention than it often receives. The 1973–1974 work described the architecture; the three-network tests showed that it could operate across the types of networks that had created the problem in the first place.

TCP Was Split Into the TCP/IP Architecture

The name “TCP/IP” can hide an important change in the design. The original Cerf–Kahn proposal used a broader Transmission Control Program that combined functions later separated between network-layer delivery and reliable transport.

As experiments expanded, researchers found that not every application needed the same reliability behavior. The design was reorganized. Internet Protocol (IP) handled addressing and delivery of datagrams across interconnected networks. Transmission Control Protocol (TCP) provided reliable, ordered communication for applications that needed it. User Datagram Protocol (UDP) provided a lighter transport option for applications that did not require TCP’s full service model.

By September 1981, RFC 791 specified the Internet Protocol and RFC 793 specified TCP for the DARPA Internet program. The layered arrangement made IP the common delivery service beneath transport protocols, rather than requiring one transport behavior for every application.

Why 1 January 1983 Is the Strongest Single Birth Date

ARPANET could not remain indefinitely divided between NCP hosts and Internet-protocol hosts. Jon Postel’s November 1981 NCP/TCP Transition Plan, RFC 801, set a concrete goal: complete the switch from NCP to IP/TCP by 1 January 1983. Its milestone schedule described January 1983 as “Full Internet Service,” with all hosts TCP-capable, NCP removed from service, relay services ended, and services operating on the TCP base. [f]

That transition explains why 1983 is different from 1969. In 1969 ARPANET was a new packet-switched network. By 1983 its hosts were operating through the protocol architecture intended to connect multiple networks. The shift was not the beginning of networking research and not the first use of TCP, but it established TCP/IP as the operational basis of the ARPANET environment and the wider Internet that was forming around it.

Calling 1983 the Internet’s “birthday” is therefore a useful shorthand if its meaning remains precise. It marks an operational protocol transition, not the instant at which packet switching, internetworking, routers, or every later Internet service was created.

Who Deserves Credit for Inventing the Internet?

The most accurate attribution depends on the technical layer being discussed. The table below separates contributions that are often compressed into the phrase “invented the Internet.”

For a short answer, Cerf and Kahn are the best-supported names when “invented the Internet” means designed the internetworking protocol architecture that became TCP/IP. A historically complete answer must still include the packet-switching researchers, ARPANET teams, gateway developers, protocol implementers, and other network projects that contributed ideas or working systems.

The Nuclear-War Origin Story Mixes Two Different Histories

A common story says ARPANET was created specifically so U.S. communications could survive a nuclear attack. That merges related but separate research histories.

Paul Baran’s RAND work did examine distributed communications in a military setting concerned with survivability. ARPANET, however, was developed as an experimental computer network for research communication and resource sharing. The Internet Society’s historical account explicitly traces the nuclear-war claim to confusion between the RAND secure-communications study and ARPANET. It also notes an important distinction: later Internet research did place weight on survivability and continued communication when portions of underlying networks failed. [b]

The Internet and the World Wide Web Are Different Inventions

Tim Berners-Lee did not invent the Internet. By the time he proposed the World Wide Web at CERN in March 1989, Internet networking already existed. Berners-Lee’s work created a hypertext information system that used network infrastructure already available to connect documents, clients, and servers.

CERN records that the 1989 proposal described a client-server hypertext system and that Berners-Lee completed an early version of the WorldWideWeb application in 1990, together with the first web-server software, httpd. [g]

Email also predates the Web, which is another useful way to see the distinction. Internet services do not have to be web pages, and the Internet did not begin when browsers appeared.

What Survived From the Original Internet Design

The protocols and standards have changed substantially since the early experiments, but one of the central design choices remains visible: the Internet does not require every participating network to use the same local transmission technology. IP creates a common layer across very different networks.

TCP also did not freeze in its 1981 form. RFC 9293, published in 2022 as the current consolidated TCP Internet Standard, states that TCP evolved over decades and incorporates changes made since RFC 793. [h] IPv6 is likewise a later form of the Internet Protocol; RFC 8200 specifies IPv6 as an Internet Standard and expands addressing far beyond IPv4’s 32-bit address space. [i]

That continuity helps explain what was actually invented in the 1970s. The enduring idea was not a particular cable, computer model, or fixed list of networks. It was a method for allowing independent networks to participate in end-to-end communication through shared Internet protocols. The network technologies beneath that common layer could change without requiring the Internet to be reinvented each time.

Why Cerf and Kahn Are Central Without Being the Only Inventors

Internet history becomes misleading when every contribution is forced into one “inventor” slot. Packet switching had multiple roots. ARPANET required planners, contractors, programmers, hardware teams, host-protocol developers, and university researchers. Internetworking drew lessons from projects in the United States and Europe. TCP/IP itself changed through testing, implementation, protocol separation, specifications, and deployment.

Cerf and Kahn occupy the center of the invention story because they addressed the problem that distinguishes an Internet from an individual network: how to connect independent packet networks without requiring them to become one uniform network. Their 1973–1974 work gave that problem a technical design. The 1977 experiments demonstrated the design across different network types. The 1983 ARPANET transition made TCP/IP the operational basis of a growing Internet.

The most defensible historical answer is therefore layered rather than singular: packet switching was developed through several research lines; ARPANET became operational in 1969; Cerf and Kahn designed the open-network protocol approach in 1973–1974; multi-network TCP testing succeeded in 1977; and 1 January 1983 is the strongest single operational date for the TCP/IP Internet.