Imagine an email travelling from a laptop in Leeds to a friend abroad. It starts on one company's access network, may cross several other networks and perhaps an ocean, then reaches a server run by another organisation. No central dispatcher approves every hop. The internet works because independently operated networks agree on how to exchange traffic.¹
The owners of the physical paths
Your internet service provider supplies the first link: fibre, mobile radio or another connection to its network. Other companies own long-distance fibre, data centres and facilities where networks meet. Across oceans, submarine cables do most of the heavy lifting. The International Telecommunication Union says they carry more than 99% of international data traffic.² Satellites matter in particular situations, but the global internet's backbone is largely on and under the ground and sea.
A cable owner controls a valuable route, not the entire destination. Networks can buy transit from a larger provider or arrange to exchange traffic directly with another network, a practice called peering. Internet exchange points make such arrangements easier by bringing networks into one facility. Peering can keep local traffic local and reduce cost and delay.³
This creates a mixture of cooperation and commercial bargaining. A large platform may pay for a private connection to an access provider, while a smaller operator may rely on paid transit. The technical ability to connect does not make capacity, price or bargaining power equal. Nor does decentralisation guarantee perfect resilience: a damaged cable or a failed major provider can disrupt many users even when the wider internet continues operating.
The rules that let machines understand one another
Physical connections would be of little use if each network spoke a different technical language. Internet protocols define how data is addressed, transmitted and interpreted. The Internet Engineering Task Force, or IETF, develops many of these specifications through an open process involving engineers from operators, vendors, academia and elsewhere. Its approach emphasises technical discussion and working implementations.⁴
The IETF does not run every network or force every device to adopt a proposal. A standard succeeds when implementers deploy it widely enough that systems interoperate. This is a different kind of influence from owning a cable: the people who shape common rules can make certain designs easier or harder to use, but they still depend on others to build and operate them.
Consider what happens when you type a familiar website name. Your device needs a numerical Internet Protocol address to reach the right service. The Domain Name System, or DNS, helps translate names into addresses. It is a distributed system; the memorable name is not itself the entire route a data packet will take.
What ICANN actually coordinates
ICANN coordinates key unique identifiers used by the internet, especially the top level of the domain-name system and the allocation framework for IP addresses. It helps ensure that a name or address can be resolved consistently across networks. ICANN says plainly that it does not control online content or the whole internet.⁵
A domain has several actors. A registrant wants a name; a registrar sells or manages its registration; a registry maintains the database for a top-level domain such as a particular suffix. ICANN has contracts and accreditation roles for many generic domains, while country-code domains have their own arrangements.⁵,⁶ Buying a domain therefore does not mean buying part of ICANN, and ICANN's role in names does not grant it ownership of your broadband line.
The distinction matters in a dispute. A website can become unreachable because its host removes it, a registrar suspends a name, a network blocks traffic or a government orders restrictions. Those are different levers with different rules and jurisdictions. Saying simply that “the internet” took a page down hides who actually acted.
One ordinary request shows how these layers remain distinct. If a domain name stops resolving, a browser may never learn which server to contact, even though cables and routers are intact. If an undersea cable fails, the name may still resolve while traffic takes a longer route or runs into congestion. If a hosting company removes a page, both naming and routing can work perfectly while the page is gone.³,⁵ The symptom “the site is down” tells you very little about which actor has acted or which component failed.
States, companies and users
Governments shape the internet through telecommunications law, competition policy, privacy rules, criminal law and, in some places, direct restrictions on access. They can regulate operators inside their territory and sometimes order a shutdown. Such power can be severe for users, but it is territorial and implemented through the networks and services within reach, not through a global master switch.⁷
Large technology companies also have substantial practical control over the services people use. A search engine can influence what is found; a cloud provider can become a dependency for many websites; an app store can set conditions for software on its devices. These powers concern particular layers and markets. They should be scrutinised without confusing a platform with the underlying internet.
A useful way to answer “who controls it?” is to ask which part of the journey is at issue. Access providers control a customer's connection. Cable and transit operators control routes. Standards bodies coordinate shared rules. ICANN coordinates names and numbers. Governments set and enforce law in their jurisdictions. The network as a whole persists through their sometimes uneasy cooperation. No one holds all the keys, but some keys open far more doors than others.