Why is the internet so fragile?

Why is the internet so fragile: Infrastructure risks

Why is the internet so fragile? While designed as a decentralized network, modern efficiency creates dangerous bottlenecks. Massive consolidation of infrastructure and reliance on thin, vulnerable physical cables leave the global grid susceptible to widespread outages. Understanding these hidden structural weaknesses helps explain why digital connectivity breaks during major incidents.

Why is the internet so fragile?

Why is the internet so fragile because it is built on a robust yet fragile model that was originally designed for decentralization but has become highly centralized in practice. While the network can handle random local failures, it is extremely vulnerable to systemic collapses within its concentrated hubs - specifically a small handful of critical companies like Cloudflare, AWS, and Microsoft. When a single software update or routing error occurs at these choke points, it triggers a massive cascading failure that can effectively vanish large portions of the digital world.

A common misconception is that internet services are fully independent from one another. In reality, many seemingly separate applications and websites rely on the same cloud regions, data centers, or infrastructure providers. When one of these shared dependencies fails, multiple services can become unavailable at the same time.

The Paradox of Centralization in a Decentralized System

Technically, the internet is a network of networks with no central authority, but modern efficiency has forced everything into a few massive silos. Today, nearly 33% of the worlds cloud infrastructure is controlled by a single provider. ^[1] This creates a dangerous efficiency. We moved to the cloud because it was more reliable than local servers, but in doing so, we created a single point of failure for millions of businesses simultaneously. If that one provider has a bad day, the global economy feels the ripple effect within seconds.

Its an uncomfortable truth. Most of the independent websites you visit daily rely on the same two or three Content Delivery Networks (CDNs). When a CDN fails, it doesnt just take down one site - it breaks the glue of the internet. Wait for it - even the most redundant systems are often just layers of the same foundation. I’ve seen teams spend months building high availability architectures only to find out they were all connected to the same DNS provider that didnt have its own backup.

Physical Choke Points and Undersea Vulnerabilities

Beyond software, the internets physical reality is surprisingly thin. Approximately 99% of international data traffic travels through subsea communications cables, not satellites. ^[2] These cables are often no thicker than a garden hose. In certain regions, like the Luzon Strait or the Red Sea, dozens of these cables are bunched together in narrow corridors. A single dragging ship anchor or a localized earthquake can sever multiple lines at once, cutting off entire nations from the digital grid. This highlights the severe internet infrastructure vulnerabilities inherent in our global connectivity.

Global cable maps reveal how much international connectivity depends on a relatively small number of major subsea routes. Although networks include redundancy, simultaneous damage to multiple cables can significantly reduce capacity and increase latency, affecting financial systems, cloud services, and international communications.

The Invisible Protocols: BGP and the Trust Problem

The internets routing mechanism, the Border Gateway Protocol (BGP), was built on a foundation of trust that no longer exists in the modern world. BGP is essentially the GPS of the internet, telling data which path to take. However, it was designed when the internet was a small club of researchers. If a network accidentally (or intentionally) announces that it is the best path for traffic it doesnt actually own, the rest of the internet blindly believes it. This is known as a BGP hijack. These BGP routing security issues remain one of the top causes of internet fragility today.

BGP route leaks and hijacks can spread rapidly because route announcements are propagated between networks worldwide. Security mechanisms such as Resource Public Key Infrastructure (RPKI) help validate routing information, but adoption remains incomplete, leaving parts of the internet vulnerable to misconfigurations and malicious announcements. Many experts argue that the current risks of internet centralization outweigh the benefits of hyper-efficient routing.

Comparison: Centralized vs. Decentralized Risks

Understanding the trade-offs between how the internet was designed and how we use it today is vital for any resiliency strategy.

Modern Cloud vs. Traditional Decentralization

Centralized Cloud (Current Model)

• Hardened by world-class experts, but a single breach affects millions

• Cascading; one service failure breaks thousands of dependent apps

• Extremely high for individual users, but 100% impact during provider failure

• Low for developers; infrastructure is managed by a third party

Decentralized (Original Vision)

• Fragmented; security depends entirely on local administration

• Isolated; the rest of the network continues unaffected by local loss

• Lower on average, but failures are contained to specific nodes

• High; each entity must manage its own hardware and routing

The Day the 'Everything' Cloud Broke

In 2021, an engineer at a major CDN made a standard configuration change to improve performance. He expected a routine update, but a hidden bug in the code triggered a massive loop that crashed their entire edge network.

Within minutes, many popular websites, online services, news outlets, and applications became inaccessible because they depended on the affected CDN's infrastructure. The incident demonstrated how a failure at a widely used provider can disrupt a large portion of the web simultaneously.

The team realized that their automated 'health checks' were actually making the problem worse by flooding the remaining servers. They had to manually override the system, a breakthrough that took nearly an hour of high-stress coordination.

The outage lasted only 49 minutes, yet it resulted in an estimated loss of 150 million USD in e-commerce revenue and proved that the internet's 'robustness' is often an illusion built on a single line of code.

A local lesson on undersea cable vulnerability

Hung, a developer in Da Nang, managed a warehouse system for international clients. He was confident in his redundancy strategy, having deployed servers in both Singapore and the United States.

However, a rare incident occurred when three major subsea fiber-optic cables (AAG, APG, and IA) failed simultaneously. Vietnam's international bandwidth dropped by 75%, effectively paralyzing Hung's system.

Hung realized that even though the servers themselves remained operational, the international network paths connecting users to those servers had become congested due to bottlenecks at key cable landing stations and transit routes.

After two weeks of struggling with connectivity issues, Hung had to deploy an additional local server cluster to ensure operations remained stable during future subsea cable outages—a stark lesson on the physical dependencies of the internet.