The Invisible Infrastructure of Control Online
Almost the entire internet runs, at some physical point, through infrastructure owned by a remarkably small number of companies. A practical look at the cables, servers, and choke points most people never think about.
Roughly 95% of intercontinental internet traffic — video calls, financial transactions, ordinary web browsing — travels through undersea fibre-optic cables, of which there are a few hundred in total, owned and operated by a surprisingly concentrated group of telecommunications companies and, increasingly, the large cloud and platform companies themselves. Most people who use the internet daily have never seen a diagram of these cables and would be surprised at how few of them there actually are, physically, holding up something that feels as diffuse and everywhere as "the internet."
This is the least discussed layer of digital control, precisely because it's the most physically invisible: not the apps and platforms people consciously choose, but the infrastructure underneath all of them that almost nobody chooses at all.
The physical layer: cables, servers, and a handful of owners
Beyond the undersea cables, the cloud infrastructure that hosts an enormous share of the internet's actual computing — the servers running the databases, applications, and services people use daily — is concentrated among three companies controlling a substantial majority of global cloud infrastructure market share. A meaningful share of the modern internet, including websites and services that appear to be independent competitors to each other, quite literally runs on servers owned by one of these three companies.
This matters beyond abstract concern about concentration for its own sake. When a major cloud provider suffers an outage, the effect isn't contained to that provider's own services — it can simultaneously take down banking apps, food delivery services, government portals, and unrelated startups, all of whom made an independent, reasonable business decision to host on the same infrastructure, without any coordination between them that would have revealed how correlated their supposed independence actually was.
The naming layer: DNS and its choke points
Beneath even the cloud layer sits the Domain Name System — the infrastructure that translates a human-readable address like a website name into the numerical address computers actually use to find it. This system depends on a small number of root servers and a concentrated set of registrars and resolvers. Disruption or deliberate restriction at this layer doesn't take down a single platform — it can make entire categories of websites unreachable for whoever depends on the affected piece of infrastructure, which is precisely why DNS-level blocking has become a preferred tool for both legitimate content moderation and, in some jurisdictions, more troubling forms of state-level censorship.
Why this concentration happened, and why it's hard to reverse
None of this consolidation happened through any single dramatic event — it followed the same logic explored elsewhere in how platforms centralised generally: significant infrastructure requires significant capital, and significant capital naturally concentrates among whoever can deploy it at the largest scale, since bigger infrastructure investments generally produce better per-unit economics. A smaller, more distributed set of cloud providers or cable operators would likely be technically less efficient, at least in the narrow sense of cost per unit of capacity, even though it might be considerably more resilient against the correlated-failure risk that concentration creates.
This is a genuine trade-off, not a simple failure of judgment by anyone involved: concentrated infrastructure is more efficient in the ways markets typically reward, and more fragile in ways that only become visible when something goes wrong across an entire region or provider simultaneously.
What visibility into this layer actually offers
Most individuals have limited direct leverage over infrastructure-layer concentration — you can't meaningfully choose which undersea cable your data travels through, in the way you can choose which app to use. But visibility into this layer still matters, for a specific practical reason: it reveals that "using a different platform" or "switching to a competitor" often provides less actual independence than it appears to, if the supposedly different services are all running on infrastructure owned by the same small handful of underlying companies.
Genuine resilience against infrastructure-layer concentration is mostly a policy and engineering question — antitrust attention to cloud market concentration, investment in more genuinely diverse infrastructure, and design choices by services themselves to avoid single-provider dependency where the stakes justify the extra cost. For an individual, the more modest, achievable benefit of understanding this layer is simply recalibrating how much independence "choosing a different app" actually buys you, and treating claims of platform diversity with an appropriately informed degree of scepticism.
Frequently asked questions
How much of the internet actually depends on undersea cables?
The large majority of intercontinental internet traffic travels through a relatively small number of undersea fibre-optic cables, owned by a concentrated group of telecommunications and large technology companies.
Why does it matter that so much cloud infrastructure is controlled by so few companies?
Because outages or policy decisions at that layer can simultaneously affect many seemingly unrelated, seemingly competing services that all happen to run on the same underlying infrastructure, revealing a concentration risk that isn't visible from the outside.
What is DNS, and why is it considered a choke point?
DNS translates human-readable website addresses into the numerical addresses computers use. Because it depends on a relatively small number of root servers and registrars, restricting access at this layer can make broad categories of websites unreachable at once.
Does using a different app actually reduce my dependency on this infrastructure?
Often less than it appears — many apparently competing services run on cloud infrastructure owned by the same small handful of underlying providers, so switching between them may not meaningfully diversify your actual infrastructure exposure.
Can individuals do anything about infrastructure-layer concentration?
Directly, not much — this is primarily a policy, antitrust, and engineering question. The main individual benefit is understanding the layer well enough to calibrate realistic expectations about how much independence switching platforms actually provides.