Who Invented the Internet?

By Victor Hoang · Updated Oct 21, 2026

Picture a night in October 1969. A student programmer named Charley Kline sits in a lab at the University of California, Los Angeles (UCLA). Kline types the word "login" to a computer at the Stanford Research Institute.

No single person invented the Internet. The two names most institutions give are Vint Cerf and Bob Kahn. They designed the protocol that joins separate networks and published it in May 1974. But their protocol ran on ideas and machines built by others: packet switching from Paul Baran, Donald Davies and Leonard Kleinrock, and the ARPANET that Larry Roberts planned and Bolt, Beranek and Newman (BBN) built. Tim Berners-Lee, who often gets the credit, built the World Wide Web, which runs on the Internet but is a different thing.

On 29 October 1969 the system crashed after two characters. That fumbled message is the most famous moment in this story, and it shows the problem with the question: nobody in that room was inventing "the Internet." They were testing one network, and the design that would join networks together was still years away. The throughline: the Internet was assembled in layers by people who mostly worked on one layer each, and the answer to "who invented it" changes with the layer you point to.

Timeline at a Glance

This table covers only the attribution milestones. For the full dated chronology, read the Internet timeline.

The Messenger Precursors (antiquity-1840s)

Roads, semaphores and wires carried messages, not packets

Start with a road. In the 5th century BCE, the Greek historian Herodotus described the Persian royal road from Sardis to Susa as a route with 111 stations, and in Rawlinson's translation he reckons the whole journey at ninety days at 150 furlongs a day. It is a precursor, not a network: a state road with rest stops, moving people and messages at the speed of a traveler. But it shows the oldest wish behind the Internet, a ruler wanting words to cross distance faster than a person could carry them.

The first system built to relay a message from station to station came much later. In July 1794, Claude Chappe and his brothers finished a line of 15 stations between Paris and Lille, and the first message on it, sent on 15 August 1794, reported the recapture of Le Quesnoy, according to the IEEE's ETHW history. Each station had two operators who moved the arms of a semaphore, and a code book turned arm positions into signs. Then came wires. On 24 May 1844, Samuel Morse demonstrated an experimental telegraph line from the Supreme Court chamber in the Capitol to a Baltimore depot, and the message was "What hath God wrought?", per a Library of Congress bulletin.

None of this was the Internet. These systems moved one message at a time over a fixed route owned by a government or a company, and a break in the line stopped the message. What they added to the story was the habit of designing a system for relaying information and the expectation that distance could be shortened. The new problem was that none of them could connect to the others, which is exactly the problem the Internet's designers would face a century later.

For teams today: every communication tool starts as a private line. The question that decides its reach is whether it can be joined to the others.

The Packet Papers (1961-1967)

Three groups, two continents, one idea about how to move data

The decisive thinking about the Internet happened on paper, before there was a network to test it on. Four lines of work ran at once, and the people involved did not coordinate at first.

At MIT, J.C.R. Licklider wrote a series of memos in August 1962 describing what he called a "Galactic Network," and in October 1962 he began leading DARPA's computer research program, according to the Internet Society's history. The National Science Foundation renders the name as "Intergalactic Computer Network." Either way, he supplied the vision: computers as a way to reach information, not just to calculate. Leonard Kleinrock, a graduate student at MIT, supplied the mathematics. The Internet Society records that he published the first paper on packet switching theory in July 1961 and the first book on the subject in 1964.

At RAND in California, Paul Baran had been working on a different worry. According to RAND's own history page, he joined RAND in 1959 and designed a network with no centralized switches, in which messages split into blocks that travel separately and are reassembled at the destination. Cold War planners feared that a nuclear attack would destroy the switching centers of the telephone system and military command and control, and Baran's studies from 1960 to 1962 were compiled into On Distributed Communications in 1964. RAND notes that Donald Davies in Britain independently devised a very similar system. Davies worked at the UK's National Physical Laboratory (NPL), where he named the message blocks "packets." The Internet Hall of Fame says he began as a colleague of Alan Turing on the ACE computer, and quotes Vint Cerf: "He invented the term packet." The naming dates to the mid-1960s.

The strands met in 1967. According to the Internet Society, Davies and his NPL colleague Roger Scantlebury presented a packet network concept at the same conference where Larry Roberts presented his ARPANET plan, and the word "packet" was adopted from NPL's work. Roberts had gone to DARPA in late 1966 to develop the computer network concept. Packet switching has no single inventor for exactly this reason: Kleinrock had the theory, Baran the distributed design, and Davies the name and an independent design.

The problem this period left behind was that it was all theory. There was a plan on paper and no machine that proved it.

For teams today: when several teams solve the same problem independently, the credit often goes to whoever names it well. Name your own work early, because the vocabulary that sticks outlives the draft that first used it. The same pattern shows up in who invented the computer.

Roberts' Network (1966-1972)

Four nodes, one crashed message, and the first written rules

This era overlaps the last one: planning ran from 1966 while the first rules were being written. The turning point is the moment the plan became hardware. In August 1968, DARPA put out a request for quotations to build the Interface Message Processor, the small computer that would sit beside each host. BBN, led by Frank Heart, won the contract in December 1968, and, as the Internet Society records, it installed the first IMP at UCLA in September 1969. SRI became the second node a month later. UC Santa Barbara and the University of Utah followed, and by the end of 1969 four host computers were connected.

While the hardware was built, the people who would use it were writing the rules. On 7 April 1969, Steve Crocker, then at UCLA, circulated RFC 1, "Host Software," on the division of the network's software between the IMPs and the host computers, and it named UCLA, SRI, Utah and UC Santa Barbara among the participants. The Internet Society says Crocker established the Request for Comments (RFC) series that year, and that the Network Working Group he led finished the Network Control Protocol, the first host-to-host rules, in December 1970.

Then came the message. Kleinrock's UCLA page records that at 22:30 on 29 October 1969, his student Charley Kline tried to log in from the UCLA host to the SRI host, where Bill Duvall was working. The system crashed after the letters "l" and "o," and the full login worked about an hour later. Kleinrock supervised, Kline typed, and Duvall received.

Roberts' contribution deserves a precise statement. The Internet Society credits him with the plan and with designing and optimizing the network's topology, and it says the plan's line speed was raised from 2.4 to 50 kilobits per second. The protocols that joined networks came later, from Kahn and Cerf.

The ARPANET solved the problem of sharing computers among research sites. It created a new one: it was a single network on a single protocol for a single community, and it could not talk to other networks.

For teams today: the first version of any platform proves the model works. The harder question, whether it can connect to systems you don't control, is the one that decides its life. API architecture is how modern software answers it.

The Cerf-Kahn Protocol (1972-1983)

One design joins many networks, then everyone switches on a single day

The turning point was a question, not a machine. Bob Kahn arrived at DARPA in 1972 and, according to the Internet Society, introduced the idea of open-architecture networking there. In spring 1973 he asked Vint Cerf to help design the protocol. A first version was distributed as INWG#39 in September 1973, and a refined version was published in 1974.

That 1974 publication is the paper institutions point to. Cerf and Kahn's A Protocol for Packet Network Intercommunication appeared in IEEE Transactions on Communications, volume COM-22, number 5, in May 1974. Its abstract presents a protocol for sharing resources that exist in different packet switching networks. It handles packet sizes that vary from network to network, transmission failures, sequencing and flow control, and it exposes problems like internetwork routing and timeouts. In plain terms, it describes how to join networks that were built differently without rebuilding any of them, which is what makes it an internet and not just a network.

Others made the design real. Jon Postel served as RFC Editor and managed protocol number assignments until his death on 16 October 1998, per the Internet Society. TCP/IP was adopted as a defense standard in 1980.

Be careful with dates here. The protocol was designed in 1973 and 1974, but it was deployed in 1983. On 1 January 1983, the Internet Society records, the ARPANET switched from NCP to TCP/IP in a "flag-day" style transition that required every host to convert at once. Anyone claiming the Internet "started" in a given year is choosing between those two events.

Kahn and Cerf's prize case is clear. The ACM's Turing Award citation for 2004 credits them with "pioneering work on internetworking" and the design and implementation of the Internet's basic protocols, TCP/IP, and "inspired leadership in networking." The Presidential Medal of Freedom followed in November 2005, honoring them for helping create the modern Internet, according to a White House archive page.

The problem this era created was access. The network now worked across many networks, but it still required command lines and expert knowledge.

For teams today: a shared standard is worth more than a better private product. Teams that adopt open interfaces are the ones that can later swap tools without rebuilding their whole stack. See digital transformation strategy for how to plan around that.

The Web Arrives (1989-1993)

CERN gives the network a face, and a new name gets the credit

In March 1989, Tim Berners-Lee at CERN circulated "Information Management: A Proposal," according to the W3C's history page. In 1990 he named the browser and editor "WorldWideWeb," and the first web server and first web page went online in November of that year. On 30 April 1993, CERN declared the technology free for anyone to use.

The web is why Berners-Lee's name turns up in "who invented the Internet" searches. It also explains a quieter detail: when Cerf and Kahn wrote about their November 2005 medals, they distinguished their work from Tim Berners-Lee's World Wide Web and Bob Metcalfe's Ethernet. The web is a system of linked pages that rides on the Internet's protocols. Our sibling article, the Internet vs the World Wide Web, covers that line in full, and the Internet timeline sets the dates in order.

The problem this era created was a credit problem. Millions of people met the Internet through the web, so the web's inventor became the face of the whole thing.

For teams today: the part of a product that users see usually gets the credit for the parts they don't. Make sure the people who built the layer underneath are recognized too.

The Credit Years (1997-present)

The builders write their own account, and the public keeps asking

By the mid-1990s the people involved were already writing the history. The Internet Society's Brief History of the Internet, published in 1997, lists nine authors: Barry Leiner, Vinton Cerf, David Clark, Robert Kahn, Leonard Kleinrock, Daniel Lynch, Jon Postel, Larry Roberts and Stephen Wolff. It is an account by participants. The National Science Foundation tells a similar group story: DARPA ran the ARPANET and funded the design of TCP/IP, NSF built NSFNET and funded the Mosaic browser, and commercial firms later built their own networks.

The public version of the question picked up politics. Kahn and Cerf wrote a statement in September 2000, archived from a public mailing list, in which they quote Al Gore's words as "During my service in the United States Congress I took the initiative in creating the Internet." They say they don't think Gore meant to claim he "invented" it, and they wrote that "No one person or even small group of persons exclusively "invented" the Internet." They credit Gore with promoting the network's importance early, urging agencies to join their networks into an "Interagency Network," and helping pass the High Performance Computing and Communications Act of 1991.

That statement is a useful model for reading any inventor claim: it names what a person actually did and when.

For teams today: the version of history that gets repeated is the one that is easiest to say in a sentence. If your team's work is shared, write the longer version down.

Which Definition Gives Which Inventor

Use this table to settle the argument. Each row is a different meaning of "the Internet," each with a different name on it.

Three Myths, Stated Precisely

Three popular statements about the Internet's origin are worth stating precisely.

"It was built to survive a nuclear war." The Internet Society's history says the rumor started from the RAND study, and that "This was never true of the ARPANET, only the unrelated RAND study on secure voice considered nuclear war." RAND's own history of Baran confirms the nuclear-attack concern was his. The same source says later work on joining networks did emphasize robustness and survivability, including the ability to withstand losses of large portions of the underlying networks, so the truth sits between the rumor and its denial.

"Al Gore said he invented the Internet." The words Kahn and Cerf quote are about taking the initiative in "creating" it, and they say they don't think he meant to claim he invented it. Their credit for his work is specific: early promotion and legislation, not engineering.

"The first message was 'Lo.'" The attempt was to type "login." According to UCLA's record, the system crashed after "l" and "o," and the full login worked an hour later. "Lo" is what arrived, not what was meant.

Key People

Key Facts

• The ARPANET's first IMP was installed at UCLA in September 1969 and four hosts were connected by the end of that year.
• The first ARPANET message was sent at 22:30 on 29 October 1969, and the system crashed after two letters.
• Cerf and Kahn's internetworking paper appeared in IEEE Transactions on Communications in May 1974.
• The ARPANET switched to TCP/IP on 1 January 1983, almost nine years after the paper.
• Cerf and Kahn received the 2004 Turing Award for internetworking.
• CERN declared the web free for anyone to use on 30 April 1993.

The Pattern Behind the History

Our synthesis, not published research: the credit for the Internet followed four habits.

• Layers have different inventors. Packets, the first network, the protocol and the web were each made by different people.
• Independent invention is normal. Baran, Davies and Kleinrock reached packet switching without a shared plan.
• The standard-setters get the textbook credit. Cerf and Kahn are cited because their paper defined the joining of networks.
• The visible layer takes the popular credit. The web gives users the Internet, so its inventor is the name most people know.

What It Teaches Teams Today

• Ask which layer a claim is about. Most credit disputes at work, from who built the process to who won the account, are really disputes about which layer is being credited.
• Write the open version. Crocker's RFCs made network design a public record. Written, dated decisions settle later arguments.
• Credit the independent parallel. Two teams reaching the same idea is evidence the idea was ready, not a duel.
• Check the pre-history. The Internet's designers built on a century of work in relaying messages; the same is true of the tools your team uses. For the parallel story in hardware, see the history of computers, and for where it all fits, return to the history of the Internet.