# Systems that fail in one place

Explainer · Updated 2026-09-07 · Fourth in the reading path  
Canonical: https://whydecentralize.org/articles/technology/

A technical system is centralized when one operator can take it down, rewrite its records, or lock a user out — even if the interface is used by millions of people. Scale is not the same as distribution.

**Federation is already in production.** Email, the web, and DNS were designed so that many independent operators could interoperate. That they now rest on a smaller set of cloud vendors, resolvers, and certificate authorities is a later concentration — not a reason to treat “web3” as the only alternative.

## Where the single points usually are

- Identity and lockout: standing granted by a platform.
- Cloud and regions: many “distributed” apps share a handful of providers.
- Trust anchors: a small set of TLS certificate authorities and root stores.
- App stores and operating systems: software distribution gated by a vendor.
- Messaging and social graphs: an open protocol with a working graph inside one product.

## What spreading the stack costs

Independent operators must agree on a protocol, absorb their own abuse, and accept slower upgrades. Spam and safety enforcement become harder to do from one desk — a [governance](https://whydecentralize.org/articles/governance/) problem as much as an engineering one. Convenience re-assembles the switch.

## What this page will not do

No token rankings, no unsourced “decentralization scores,” no invented cloud-traffic shares. A distributed ledger is not required for a distributed system. Ledgers often re-centralize at the exchange and the hosted wallet.

Export controls and technology supply chains: [usforeignpolicy.org](https://usforeignpolicy.org). Conflict geography: [armedconflicts.org](https://armedconflicts.org).

Next: [governance](https://whydecentralize.org/articles/governance/).