A wide area network (WAN) is a telecommunications network that interconnects sites across large geographic distances — cities, countries, or continents — typically by carrying traffic over links leased from or operated by carriers rather than infrastructure the user owns end to end. WANs join local area networks into a single reachable whole using technologies ranging from leased lines, MPLS, and carrier Ethernet to broadband internet, cellular, and satellite, with the internet itself the largest example, and they trade the high bandwidth and low latency of the LAN for reach, at recurring circuit cost.
Semantic Classification
Content
Definition
A wide area network spans the distances a Local Area Network cannot: between buildings in different cities, national branch estates, continental data-centre pairs, and ultimately the whole planet. The defining boundary is administrative and economic as much as geographic — within the LAN an organisation owns its cabling and switches outright, but to cross public rights of way it must ride infrastructure operated by carriers, so WAN capacity is leased by the month and priced by bandwidth and distance. The internet is the WAN of WANs: tens of thousands of autonomous systems stitched together by BGP peering and transit into a single global reachability fabric, and every enterprise WAN today either supplements or rides upon it.
The technology stack has turned over repeatedly. Point-to-point leased lines (T1/E1) and circuit technologies (X.25, Frame Relay, ATM) gave way in the 2000s to MPLS VPN services, in which a carrier’s label-switched core presents each customer with a private any-to-any routed network with contractual latency, loss, and availability SLAs — still the benchmark for predictable enterprise connectivity. Alongside sit carrier Ethernet (metro E-Line/E-LAN services), dark fibre and DWDM waves for those who want to run their own optics, and, at the access edge, broadband, DOCSIS, 4G/5G fixed wireless, and LEO satellite (Starlink) — the last two having made usable WAN access nearly location-independent. Site-to-site Virtual Private Network tunnels (IPsec, WireGuard) overlay private address space and encryption on any of these substrates.
Physics disciplines every WAN design: Telecommunications links are constrained by the speed of light in fibre (~5 µs per kilometre, so ~60–70 ms round-trip London–New York before queueing), making latency — not bandwidth — the binding constraint at distance. This drives content delivery networks, regional application replicas, TCP congestion-control evolution (BBR), and the general principle of moving computation towards users rather than data across oceans.
Current Landscape
The enterprise WAN has been remade twice in a decade. SD-WAN — Software-Defined Networking applied to the branch edge — replaces static routing over a single MPLS circuit with centrally orchestrated overlays that measure loss, latency, and jitter across multiple underlays (MPLS, broadband, 5G) per application and steer traffic accordingly, cutting cost and lifting cloud-application performance; it is now the default branch architecture. The second remake is SASE/SSE, which fuses SD-WAN with cloud-delivered security (secure web gateway, CASB, zero-trust network access) so that the security perimeter follows users and workloads rather than sites. Meanwhile hyperscalers operate private global backbones that increasingly carry enterprise inter-region traffic (cloud WAN offerings), and 400G/800G coherent optics keep the underlying carrier economics on their long deflationary curve. The WAN’s centre of gravity has shifted accordingly: from connecting branches to a head office, to connecting everyone to the cloud.
Recent market data underlines the pace of this shift. Dell’Oro Group reported the SASE market up 22% year-on-year in Q2 2025 to 16–25bn by 2027–2029.
Sources: