The Internet Has Invisible Switchboards. The Internet Society Just Warned They Are Centralizing.

Almost everything you do online passes through a piece of infrastructure you will never see. When you load a website, stream a video, or send a message, your data almost certainly transits an internet exchange point, an IXP, a physical location where dozens or hundreds of networks plug into shared switches and trade traffic directly with one another instead of paying an upstream provider to carry it. IXPs are the internet's switchboards, and for a decade the global policy agenda was simple: build more of them, and keep local traffic local.

That agenda just changed. On October 13, the Internet Society published a full revision of its 2015 IXP policy brief, and the message is striking: in many countries, the question is no longer whether an exchange exists. It is whether the local interconnection ecosystem around it can survive as traffic concentrates into fewer facilities, platforms, and locations.

What an internet exchange actually does

An IXP is conceptually simple. Networks, ISPs, mobile operators, content delivery networks, cloud providers, connect to a shared switching fabric. They then peer with one another, agreeing to exchange traffic directly. If a broadband ISP's customer in Manchester wants video from a CDN that is also present at the Manchester exchange, the data hops across the room rather than traveling to London, or worse, across an ocean and back.

This is the "keep local traffic local" principle, with three concrete benefits: lower latency (fewer hops), lower cost (less paid transit), and resilience (more paths between networks). A related concept is the route server, described in RFC 7947, which distributes routing information among participants so they do not need hundreds of bilateral sessions. The route server normally does not forward packets; the switching fabric, each network's routing policy, and the transport to the exchange remain separate pieces of the arrangement. Being "connected to an IXP" therefore says surprisingly little on its own about where a network actually is or which local peers it reaches.

The exchange can remain on the map while the ecosystem around it thins. The new policy question is how to tell the difference.

What the new brief actually says

The Internet Society's 2026 revision changes the unit of analysis. An exchange, the brief argues, is best understood as part of a local interconnection ecosystem: active peering communities, viable IXP operations, local hosting and content, data centres, technical capacity, and varied ways to interconnect. The proposed policy response is correspondingly broader than building a switch fabric: preserve neutrality and voluntary peering, remove barriers to participation, and avoid regulation that treats an exchange like a telecom carrier simply because networks meet there.

The sharpest passage concerns remote peering, the practice of reaching an exchange without equipment at the site. The brief acknowledges that remote access and virtual services expand reach and are not inherently harmful. Its concern is substitution: if remote access or a centralized platform displaces local participation, the exchange may continue to report healthy traffic while losing the people, operators, and skills that made it a local institution.

The brief also puts limits on its own scorecard. Published IXP traffic statistics, it notes, do not capture every form of interconnection. Private peering can bypass a public exchange entirely, and measurement studies have found that visible routing data can miss links. Traffic growth shows more activity on one fabric; it cannot by itself establish diversity, affordability, resilience, or community health. Membership counts have a similar limitation if they do not distinguish active local participants from remote access or selective peering. The Internet Society points to its own analysis of 50 countries receiving IXP development and peering-ecosystem support, reporting that broadband affordability improved in 40 of them, while acknowledging that this is an association, not a controlled estimate of what an exchange caused.

The counterpoint: exchanges are still growing fast

Here is the interesting tension. The same week the warning was published, real exchanges posted records that suggest local interconnection is thriving, not thinning. On October 13, the London Internet Exchange (LINX), the member-owned, not-for-profit exchange with roughly 900 members spanning broadband ISPs, mobile operators, and CDNs, launched its Cloud Connect service at its Manchester exchange. Cloud Connect gives members direct access to Amazon Web Services and Google Cloud Platform through their existing LINX ports, cutting latency and improving control over cloud connectivity.

The launch follows shortly after LINX Manchester tipped the 1 Tbps mark in traffic, placing it among the top 10 exchanges in Europe. Manchester now hosts more than 130 connected networks across five data centre partner points of access. LINX CEO Jennifer Holmes described a three-year strategy centered on growing the north of the UK through strategic partners and product innovation, while keeping the member experience and technical support constant. Cloud Connect is now available at both LINX Manchester and LON1 in London.

Meanwhile, the global interconnection provider NetIX has launched NetIX 2.0, folding terrestrial and satellite connectivity into a single network architecture rather than treating satellite as a separate access technology. Traffic is classified by its requirements and routed accordingly, which gives ISPs and carriers finer control over how connectivity is engineered. NetIX's footprint now spans more than 220 data centres in over 100 cities across 65 countries, connecting more than 200 members to over 10,000 visible networks and 60-plus internet exchanges, with daily traffic peaks above 7 Tbps.

Why this matters to anyone who uses the internet

These two stories, the warning and the growth, are the same story told from opposite ends. The internet's founding architecture is decentralized: no single network controls the whole, and resilience comes from many paths between many networks. That decentralization is not a law of nature. It is a property that has to be maintained, facility by facility and peering agreement by peering agreement.

Centralization creeps in quietly. When a small ISP finds it cheaper to reach one big exchange remotely than to invest in local peering, local capacity atrophies. When traffic concentrates in a handful of mega-facilities, outages and policy decisions at those chokepoints have outsize consequences. When exchanges get reclassified as carriers by regulators, the volunteer-run, community-funded model that built much of the world's peering fabric becomes harder to sustain. None of this breaks the internet tomorrow. It makes the internet more fragile and more expensive in the ways that compound over years.

The Internet Society's brief is best read as a policy agenda, not an audit. It says the task has shifted from deployment to persistence, and names the institutional conditions that support it. Whether those conditions are holding in any particular country is a separate question, one answered by local records: the share of local versus remote participants, hosting and content presence, operating finances, governance, and how the ecosystem responds to outages.

For most of us, the takeaway is simpler. Every time a regional exchange like Manchester adds networks, capacity, and services like Cloud Connect, the internet gets a little more distributed, and a little more resilient. The switchboards are invisible, but what happens to them is anything but abstract. It determines how fast your traffic moves, how much your ISP pays to move it, and how many independent paths exist when something goes wrong.