
What is the history of Ethernet?
Ethernet history is not just a story of link speed rising from a few megabits to hundreds of gigabits. It is also the story of shared coaxial cable giving way to twisted-pair star wiring, hubs giving way to switches, half-duplex collision domains becoming full-duplex point-to-point links, and a LAN technology expanding into data centers, vehicles and industrial communication. One reason Ethernet has lasted so long is that the basic frame and MAC model could survive while the physical media and data rates changed radically.
The usual starting point is 1973. At Xerox PARC, Robert Metcalfe was working on a local network for Alto computers and networked laser printing, drawing on ideas from ALOHAnet. In a memo dated 22 May 1973 he proposed dropping the name “Alto Aloha Network” and called the design the “Ether Network.” With David Boggs, he developed the early experimental Ethernet system, which operated at roughly 3 Mb/s.[1][2]
Image note: the timeline above represents the move from Xerox PARC and early cable systems through NICs, hubs/switches, fiber, industrial Ethernet and modern connected systems. A title attribute provides a mouse-hover explanation, while alt text is retained for accessibility.
The 1970s: why Ethernet was invented
Xerox PARC needed a practical way to connect many computers and shared resources inside the same building. Unlike wide-area research networks such as ARPANET, the problem was local: workstations, printers and file services had to communicate economically at office scale. Ethernet adapted the shared-medium idea from ALOHAnet to a cable-based local network.[3]
The first Ethernet was not the RJ45 star network familiar today. Stations shared coaxial cable, listened before transmitting and detected collisions if more than one device transmitted at once. That behavior became the foundation of CSMA/CD. In modern switched full-duplex Ethernet, collisions are normally absent, so the old shared-medium behavior is historical context rather than the everyday mechanism of a current LAN.
1980: DIX turned Ethernet into a multi-vendor industry specification
A decisive step came when Digital Equipment Corporation, Intel and Xerox published the 10 Mb/s DIX Ethernet specification in 1980. That moved Ethernet beyond a Xerox research system and gave multiple vendors a common target for commercial hardware. IEEE's 50-year review identifies the 1980 DIX specification as the bridge from PARC development to the IEEE 802.3 standardization process.[4]
1983: IEEE 802.3 standardization
The IEEE standards process approved the original 802.3 work on 23 June 1983. This provided a common standards framework for Ethernet MAC and physical-layer development. Later revisions and amendments kept adding new media and new speeds to the same family.[5]
So “Ethernet was invented in 1983” is inaccurate. The invention/prototype story starts in 1973, DIX industrialization came in 1980, and IEEE standardization followed in 1983.
1990: 10BASE-T moved Ethernet toward twisted-pair star wiring
Early Ethernet relied heavily on shared coaxial bus cabling. IEEE 802.3i-1990 standardized 10BASE-T over unshielded twisted pair, accelerating the move toward star-wired installations where each endpoint had its own cable to a central hub or, later, a switch.[6]
This change reshaped practical networking. Cabling, fault isolation and endpoint replacement became much easier, and topics such as Ethernet cable types, maximum cable length and the network interface card became part of everyday installation work.
The 1990s: switching and full duplex changed the network model
A hub still created a shared collision domain even if the cables formed a physical star. As switching spread, each link could become a separate point-to-point segment, enabling full-duplex operation. Ethernet increasingly moved away from shared-medium collision behavior toward switched links where a device could transmit and receive at the same time. IEEE's 50th-anniversary material highlights this as one of the major transformations of the 1990s.[7]
This is why saying “Ethernet uses CSMA/CD” without historical context is incomplete. CSMA/CD belongs to shared half-duplex Ethernet; it is not the normal operational mechanism of modern full-duplex switched links.
1995: Fast Ethernet reached 100 Mb/s
IEEE 802.3u-1995 introduced Fast Ethernet at 100 Mb/s. 100BASE-TX became one of the most widely deployed copper LAN technologies and provided a major step between classic 10 Mb/s Ethernet and the coming Gigabit generation.[8]
1998–1999: the Gigabit Ethernet generation
IEEE 802.3z-1998 standardized 1000 Mb/s Ethernet, especially for fiber and short-reach media. IEEE 802.3ab-1999 then defined 1000BASE-T over four-pair Category 5 balanced copper cabling. That distinction is useful: Gigabit Ethernet arrived in 1998, while the familiar copper 1000BASE-T specification followed in 1999.[9][10]
For practical design details, the separate Gigabit Ethernet guide covers negotiation, cabling and the difference between 1 Gb/s link rate and real application throughput.
2002: 10 Gigabit Ethernet pushed beyond the desktop LAN
10 Gigabit Ethernet, standardized in 2002, strengthened Ethernet's role in backbone, data-center and service-provider networks. The technology was no longer simply a way to connect desktop PCs. Fiber PHY development made it possible to preserve the Ethernet MAC/frame family while extending speed and reach dramatically.[11]
2003 and later: Ethernet cabling started delivering power
Power over Ethernet added another dimension. IEEE 802.3af standardized power delivery to compatible endpoints over Ethernet cabling, followed later by PoE+, four-pair PoE and higher power classes. Ethernet cabling could now deliver both communication and power to phones, cameras, access points and other devices. The separate Power over Ethernet page covers the power classes and PSE/PD roles.
2010: 40 and 100 Gigabit Ethernet
IEEE 802.3ba-2010 introduced 40 Gb/s and 100 Gb/s Ethernet. The market was now moving beyond the old pattern of a single 10× jump and into a broader family of data-center and transport speeds.
2017: 200 and 400 Gigabit Ethernet
IEEE 802.3bs-2017 added MAC, physical-layer and management parameters for 200 and 400 Gb/s operation. IEEE records show board approval on 6 December 2017 and publication on 12 December 2017.[12]
These rates are far removed from the needs of a typical PLC link, but they show how far the same Ethernet family expanded into data-center and carrier infrastructure.
2024: 800 Gigabit Ethernet became a published standard
IEEE 802.3df-2024 added 400 and 800 Gb/s MAC/PHY capabilities. IEEE lists approval on 15 February 2024 and publication on 15 March 2024. That means 800 GbE is now a published part of Ethernet history, not merely a future target.[13]
By contrast, 1.6 Tb/s Ethernet remains part of the P802.3dj development work in 2026. A history page should not present an active standards project as if it were already a completed historical milestone.[14]
How did Ethernet enter industrial automation?
Industrial adoption was about more than attaching factory computers to the office LAN. PLCs, HMIs, SCADA, drives and data-collection systems increasingly used Ethernet-based upper networks. Industrial protocol families such as EtherNet/IP, PROFINET and EtherCAT then built different device and timing models on top of Ethernet-related infrastructure.
The names can be misleading. EtherCAT and traditional Ethernet may share Ethernet framing and physical-layer technology, but they do not use the same process-data scheduling model. “It has Ethernet” does not identify the industrial protocol running on the port.
The 2010s and 2020s: Ethernet is no longer only a LAN technology
Ethernet now appears in data centers, factories, building automation, vehicles, cameras, access points and edge devices. Single-Pair Ethernet brings Ethernet PHYs into automotive and industrial wiring, while IIoT, edge computing and data logging use Ethernet as one of the common transport foundations between field equipment and higher-level systems.
That expansion also changed the design questions. Engineers now have to think about segmentation, authentication, remote access and Ethernet network security, not only cable continuity and link speed.
Ethernet history at a glance
| Year | Milestone | What changed? |
|---|---|---|
| 1973 | Xerox PARC Ethernet concept and experimental network | Shared packet LAN for computers and printers |
| 1980 | DIX 10 Mb/s Ethernet | Open multi-vendor industry specification |
| 1983 | IEEE 802.3 approval | Common MAC/PHY standardization framework |
| 1990 | 10BASE-T | Twisted-pair star wiring replaced much coaxial bus cabling |
| 1995 | Fast Ethernet, 100 Mb/s | 10× LAN speed increase |
| 1998–1999 | Gigabit Ethernet / 1000BASE-T | 1 Gb/s over fiber and then four-pair copper |
| 2002 | 10 Gigabit Ethernet | Backbone and data-center use expanded |
| 2003 | Standardized PoE | Power could travel with Ethernet data |
| 2010 | 40/100 GbE | High-capacity data-center and backbone links |
| 2017 | 200/400 GbE | Higher-density infrastructure |
| 2024 | 800 GbE | Published 800 Gb/s Ethernet standard |
| 2026 | 1.6 Tb/s development continues | Active IEEE project, not yet a completed historical milestone |
Why has Ethernet survived for more than fifty years?
Because Ethernet is not one cable, one speed or one application. It preserves a common MAC/frame model while changing the physical layer. Coaxial cable, twisted-pair copper, fiber and single-pair Ethernet can serve very different environments. Higher-layer IP/TCP services and many protocols that operate over Ethernet can then use that base.
This adaptability explains why the same broad Ethernet family can sit behind an office workstation, a camera, a PLC programming port, PROFINET, EtherNet/IP or a data-center link without all of those applications being the same protocol.
Four historical details that are often mixed up
First: Ethernet was not invented in 1983; that was the IEEE standardization milestone, while the experimental work dates to 1973. Second: today's RJ45 star network was not the original Ethernet architecture; early systems used shared coaxial bus media. Third: CSMA/CD is historical shared-medium behavior, not the everyday mechanism of a full-duplex switched link. Fourth: “Gigabit Ethernet in 1998” and “1000BASE-T copper in 1999” refer to different milestones.
Ethernet topic guide
• What is Ethernet?
• How does Ethernet work?
• What types of Ethernet cables are there?
• What are the different Ethernet speeds?
• What is the Ethernet protocol?
• What is an Ethernet card?
• What is the difference between a switch and a hub?
• What protocols can operate over Ethernet?
• Advantages and disadvantages of Ethernet
• What is Power over Ethernet?
• What is the structure of an Ethernet frame?
• What is a MAC address?
• How has Ethernet evolved and what are the future trends?
• What is Gigabit Ethernet?
• Ethernet network security
• How is data transmitted in Ethernet?
• What is the maximum Ethernet cable length?
Technical references
[1] Computer History Museum — 22 May 1973, Metcalfe's Ethernet memo. Original Ethernet memo
[2] Computer History Museum — Metcalfe/Boggs and Ethernet's 1973 origin. 1973 Timeline
[3] Robert Metcalfe — “Ether Acquisition,” Xerox PARC archive copy, 22 May 1973. Memo PDF
[4] IEEE SA — Ethernet's 50-year development, DIX 1980 and standardization history. Ethernet Through the Years
[5] IEEE SA — IEEE 802.3 history, board approval 23 June 1983. IEEE 802.3 history
[6] IEEE SA — IEEE 802.3i-1990, 10BASE-T. 10BASE-T
[7] IEEE SA — transition toward switched/full-duplex Ethernet. 50th Anniversary
[8] IEEE SA — IEEE 802.3u-1995 Fast Ethernet. 802.3u
[9] IEEE SA — IEEE 802.3z-1998, 1000 Mb/s Ethernet. 802.3z
[10] IEEE SA — IEEE 802.3ab-1999, 1000BASE-T over four-pair Category 5 copper. 802.3ab
[11] IEEE SA — Ethernet speed history including the 10 GbE generation. IEEE Ethernet history
[12] IEEE SA — IEEE 802.3bs-2017, 200/400 Gb/s. 802.3bs
[13] IEEE SA — IEEE 802.3df-2024, 800 Gb/s. 802.3df
[14] IEEE SA — P802.3dj and the continuing 1.6 Tb/s Ethernet work. Ethernet's Next Bar