The News, Explained

Cloudflare published its third-quarter Internet disruption report on October 27, 2026, drawing on observations from its Radar trends service. The report covers July through September and records 49 disruptions: 19 government-directed shutdowns, 14 power-related incidents and five natural-disaster incidents among them. It says the 14 power-related disruptions occurred across 10 countries. Source

One facility’s power failure affected multiple providers

Honduras offers a concrete example. On September 10, a power failure at an interconnection facility, where networks connect to one another, affected at least 11 major providers. The disruption lasted about three hours, and Cloudflare says country-level HTTP traffic fell by 63%. When different networks exchange traffic through a concentrated facility, a power problem there can affect connectivity across several providers. Source

Cloudflare says cross-border provider Telecomunicaciones de Guatemala and mobile carrier Celtel were unaffected. It suggests their Honduran connectivity may depend more heavily on infrastructure outside the country. The report separates the observation that they were unaffected from a possible explanation of why. Source

The equipment needing power extends beyond a router at home. The report points to routers that connect traffic paths, mobile base stations, data centers and facilities connecting networks. Instability in electricity supply can affect the equipment and paths needed to keep a connection working. Source

Start with what the 63% figure measures

HTTP traffic in this report is a measure of requests sent to websites. Cloudflare’s Radar documentation says its HTTP request data examines end-user requests reaching websites that use Cloudflare’s network to deliver content. This includes both HTTP and encrypted HTTPS connections. The unit to keep in view when reading the Honduras figure is the change in web request traffic within that observation scope. Source Source

Different observed groups can show different patterns. In the report’s Japan earthquake example, the impact appeared mainly in desktop traffic, while mobile traffic remained largely unaffected. Radar’s HTTP documentation explains that device classification comes from the User-Agent header, identifying information sent by the requesting software. Desktop and mobile here describe observations of web requests classified by device type. Source Source

OYOPICK’s Take

Judging continuity of Internet service calls for asking about shared power and interconnection dependencies alongside the number of available connections. In Honduras, a power problem at one interconnection facility affected at least 11 providers. In our view, assessing the room to route around a failure requires looking beyond different provider names into their actual dependencies.

The two unaffected providers offer a clue for examining differences in connectivity paths. Cloudflare presents its explanation for that difference as a possibility. Keeping that distinction clear, simpler explanations from services or providers about power and interconnection dependencies, and the recovery paths available during a failure, could give people stronger grounds for judging resilience. We hope a wider choice of connections comes with a clearer understanding of the failures those options can help address.