A power fault in one cabinet at a Johannesburg data centre was enough to throw a chunk of the country’s internet into disorder. Afrihost users started reporting failures, then the complaints spread to customers on Axxess, Telkom and Vodacom. By the time the dust settled, the outage had shown how little separation exists between networks most people think of as competitors.

The immediate cause was narrow, almost absurdly so. Afrihost said the problem came from a power failure at its cabinet inside Teraco’s Johannesburg facility, hitting its core network. From there, the disruption rippled outward across fibre and ISP layers that most households never see, but depend on every time a payment terminal connects, a class logs in, or a work laptop wakes up.

What failed

The fault was not a national blackout or a fibre cut somewhere on a highway verge. It was a power problem at a specific cabinet in a Teraco data centre in Johannesburg. That cabinet housed Afrihost network gear, and once the power dropped, the provider’s core traffic handling took the knock.

Data centres such as Teraco are interconnection hubs where ISPs, fibre operators, and content networks keep the equipment that moves traffic around the country. When a central cabinet loses power, the damage can extend far beyond one customer list. A localised failure becomes a network problem.

Afrihost said engineers from both Afrihost and Teraco were sent in to get services back up as quickly as possible. That response was predictable. The bigger point it exposed: if enough plumbing is concentrated in one building, a fault in one rack can turn into a widespread outage before anyone has finished posting screenshots.

Who lost service

Afrihost customers were the first to complain, but not the only ones stuck staring at dead connections and spinning login pages. Similar reports came in from users on Axxess, Telkom, and Vodacom. The outage also spread through major fibre networks, including Vumatel, Openserve, Frogfoot, and MetroFibre.

That list is uncomfortable. These are not fringe players or small regional outfits; they sit near the centre of how most urban and suburban users in the country get online. When the outage touched that group at once, it was clear this was not a brand-specific problem. It was a shared infrastructure problem wearing multiple logos.

For households, the hit was immediate. Work calls dropped. Streaming stopped. Schoolwork stalled. Banking apps became another thing to reload and hope for the best. For small businesses, the damage is more brutal. Card machines fail, cloud systems stop answering, customer service goes dark, and every minute off the network turns into lost sales or wasted staff time.

Many users are buying different products from providers that still depend on the same physical choke points. The bill may come from one company and the line may be advertised by another, but the traffic can still pass through the same building in Johannesburg.

Why one cabinet can knock so much loose

The network industry likes to talk about redundancy, and for good reason. Backup power, alternate routes, and failover systems all exist because engineers know that hardware fails, feeds trip, and things catch fire at the worst possible time. But redundancy only works if it is real, properly separated, and actually tested under stress.

A Teraco facility is designed to be resilient, with the kind of power and connectivity layers that keep most individual failures from becoming disasters. Yet even a well-built data centre can still become a pressure point when too much core infrastructure is concentrated in one place. The more networks that place their heart inside the same building, the more damage one local fault can do.

Johannesburg is especially sensitive here because it is a major exchange point for local traffic. That makes the city efficient, but also loaded. Efficiency and fragility often travel together in telecom. The industry keeps chasing lower latency and lower operating cost, resulting in a system that works beautifully until one central node has a bad day.

This outage also cuts against the comforting idea that outages happen at the edge, where a trench is dug up or a pole is knocked over by a truck. The weak spot here sat deep in the core. That is a different problem, and a more embarrassing one, because the failure was not out on the street. It was inside the part of the network that is supposed to be hardest to break.

What readers should watch next

Providers should give clear explanations once service is restored. People can live with an outage. They do not live well with vagueness. Customers want to know whether the issue was isolated to one cabinet, whether backup paths held, and why so many networks felt the impact at the same time.

This should trigger real hardening of core infrastructure. That means more than a hurried statement about resilience. It means separating critical systems more aggressively, checking whether enough equipment is clustered in a single facility, and making sure failover is not just a slide deck promise.

The broader industry lesson: South Africa’s internet is still too dependent on a few large hubs and a small number of shared physical sites. That arrangement keeps traffic fast and prices workable, but also means one bad power event can disrupt daily life for a lot of people at once. If the outage pushed that fact into the open, good. It should stay there until someone does something about it.