In telecommunications, packet switching is a method of grouping data into short messages in fixed format, i.e., packets, that are transmitted over a telecommunications network. Packets consist of a header and a payload. Data in the header is used by networking hardware to direct the packet to its destination, where the payload is extracted and used by an operating system, application software, or higher layer protocols. Packet switching is the primary basis for data communications in computer networks worldwide.

During the early 1960s, American engineer Paul Baran developed a concept he called distributed adaptive message block switching as part of a research program at the RAND Corporation, funded by the United States Department of Defense. His proposal was to provide a fault-tolerant, efficient method for communication of voice messages using low-cost hardware to route the message blocks across a distributed network. His ideas contradicted then-established principles of pre-allocation of network bandwidth, exemplified by the development of telecommunications in the Bell System. The new concept found little resonance among network implementers until the independent work of Welsh computer scientist Donald Davies at the National Physical Laboratory beginning in 1965. Davies developed the concept for data communication using software switches in a high-speed computer network and coined the term packet switching. His work inspired numerous packet switching networks in the decade following, including the incorporation of the concept into the design of the ARPANET in the United States and the CYCLADES network in France. The ARPANET and CYCLADES were the primary precursor networks of the modern Internet.

A simple definition of packet switching is:

The routing and transferring of data by means of addressed packets so that a channel is occupied during the transmission of the packet only, and upon completion of the transmission the channel is made available for the transfer of other traffic.

Packet switching allows delivery of variable bit rate data streams, realized as sequences of short messages in fixed format, i.e. packets, over a computer network which allocates transmission resources as needed using statistical multiplexing or dynamic bandwidth allocation techniques. As they traverse networking hardware, such as switches and routers, packets are received, buffered, queued, and retransmitted (stored and forwarded), resulting in variable latency and throughput depending on the link capacity and the traffic load on the network. Packets are normally forwarded by intermediate network nodes asynchronously using first-in, first-out buffering, but may be forwarded according to some scheduling discipline for fair queuing, traffic shaping, or for differentiated or guaranteed quality of service, such as weighted fair queuing or leaky bucket. Packet-based communication may be implemented with or without intermediate forwarding nodes (switches and routers). In case of a shared physical medium (such as radio or 10BASE5), the packets may be delivered according to a multiple access scheme.

Packet switching contrasts with another principal networking paradigm, circuit switching, a method which pre-allocates dedicated network bandwidth specifically for each communication session, each having a constant bit rate and latency between nodes. In cases of billable services, such as cellular communication services, circuit switching is characterized by a fee per unit of connection time, even when no data is transferred, while packet switching may be characterized by a fee per unit of information transmitted, such as characters, packets, or messages.

A packet switch has four components: input ports, output ports, routing processor, and switching fabric.

The concept of switching small blocks of data was first invented independently by Paul Baran at the RAND Corporation during the early 1960s in the US and Donald Davies at the National Physical Laboratory (NPL) in the UK in 1965.