Independent Publication · Established 2026 · Open Access
SYSTEM 06
Foundational Study
← All Systems

Internet Backbone

The aggregation, routing, and interconnection layer beneath nearly every digital service.

Overview

The internet backbone is the layer of infrastructure that most users never encounter directly and most analysts find difficult to characterise. It is neither the cables alone, nor the routers alone, nor the businesses alone, but the operating practice through which all of them produce a global routing system out of thousands of independently managed networks.

Its defining feature is that no single authority operates it. The backbone exists because tens of thousands of network operators have agreed, contractually and technically, to exchange traffic on terms that have evolved over decades. That arrangement is more durable than it appears, but its durability is a property of the community rather than of any institution.

§01

A network of networks, not a network

The internet is built from autonomous systems — independently administered networks identified by globally unique numbers — that exchange routing information using a small set of protocols, primarily the Border Gateway Protocol. There is no central directory of who connects to whom; the topology emerges from the bilateral and multilateral agreements between operators.

This federated structure is the source of both the internet's resilience and its quieter fragilities. Routing problems in one network can propagate globally within minutes when announcements are misconfigured. The same openness that permits permissionless connectivity also permits route leaks, hijacks, and the silent rerouting of traffic across paths its owners did not intend.

§02

Interconnection economics and concentration

Interconnection between networks happens at internet exchange points and through bilateral peering or transit arrangements. The economics of these arrangements have shifted substantially over the past two decades. A small number of large content networks now exchange traffic directly with access networks at the edge, bypassing much of the traditional transit market.

The result is a backbone whose centre of gravity has moved from carrier-operated long-haul networks toward content-driven private interconnection. This shift has improved performance and reduced cost for users, but has also concentrated traffic flows through a comparatively small number of operators whose internal architectures are not visible to outsiders.

§03

Routing security as a structural problem

The protocols that hold the internet together were designed for a smaller, more trusting community than the one that now uses them. Mechanisms to authenticate routing announcements have been deployed slowly, unevenly, and without strong incentives for individual operators to adopt them. The result is a routing system whose integrity depends substantially on operator vigilance rather than on cryptographic enforcement.

Incidents in which traffic is rerouted through unintended networks — whether by error or by intent — are common enough to be routine. Their consequences range from brief and local to globally consequential, depending on which prefixes are affected and how long the misroute persists. Strengthening routing security is widely understood to be necessary and structurally difficult to coordinate.

§04

Dependence on physical and naming infrastructure

The backbone runs on physical infrastructure — fibre, data centres, power, cooling — that it does not itself control. It depends on naming infrastructure — the domain name system, public key directories, certificate authorities — that is operationally distinct but no less essential. Any analysis of the backbone that ignores these adjoining systems misses where its real vulnerabilities sit.

Outages attributed to the internet itself are often outages in one of these adjoining layers. A misconfiguration in a name service, a failure in a certificate authority, or a fibre cut in a single corridor can render large portions of the visible internet unreachable, even when the routing fabric itself is functioning normally.

§05

Governance through community institutions

Coordination across the internet's many operators happens through a constellation of community institutions: regional address registries, standards bodies, operator forums, and informal working groups. These bodies have no enforcement powers in the conventional sense. Their authority rests on technical credibility and the willingness of operators to participate.

This governance model has been remarkably effective over decades, but it is increasingly tested by state actors seeking formal control, by commercial actors whose internal decisions affect global routing, and by the sheer scale and heterogeneity of the modern operator community. The institutions that once produced consensus among a few hundred operators now struggle to do so among many thousands.

§06

Operational opacity at the centre

Much of the modern backbone's traffic now traverses private networks operated by a small number of large platforms and cloud providers. These networks are engineered to a standard that often exceeds that of the public internet, but their internal topology, capacity, and failure modes are largely opaque to the broader operator community.

When these private networks fail, the public internet around them experiences the consequences without visibility into the cause. The asymmetry between the importance of these systems and the public information available about them is one of the defining governance problems of the contemporary backbone.

§07

Why the backbone matters here

Almost every other critical system now depends on the internet backbone for at least some of its operation. Payments, telecommunications, logistics, and increasingly even electrical grid coordination ride on networks that the backbone makes possible.

Critical Dependencies studies the internet backbone as a case in which dependency concentration, operational complexity, governance lag, and the limits of community-led coordination are all simultaneously visible — and in which the boundary between technical and institutional problems has long since dissolved.

We study systems, not actors.

Related Systems

Other systems studied alongside this one for governance, resilience, and dependency context.