The Internet is "Too Big to Fail": The Dangerous Comfort of Centralization


Transcript of the narrated version (5 min).
Narrated with a synthetic voice (Larry). The writing is Esteban Rey's — kilowatto.com.

---
The Internet is "Too Big to Fail": The Dangerous Comfort of Centralization. 

Yesterday, the corporate world held its breath once again. A massive Microsoft 365 outage left thousands of companies and millions of users staring at error screens instead of working. It wasn’t just about not being able to send an email or co-author an Excel sheet. The impact was far deeper and revealed our systemic fragility: Microsoft Teams, which today functions as the de facto PBX for countless organizations, left contact centers dead silent. Support lines were cut. Even more critically, the failure in email delivery systems prevented millions of users from receiving One-Time Passwords, or OTPs. Suddenly, you couldn’t log into your bank, you couldn’t validate a movie ticket purchase, you couldn’t sign documents, all because a single company, somewhere in the cloud, had a bad day.

The "No-Brainer" Trap. 

 I have spoken many times about the risks of technological concentration, but incidents like yesterday's force us to be more vocal. Most people are unaware that the Internet, a network born with the promise of being decentralized and indestructible, now hangs by four or five very thick threads. And I don't blame them. For Chief Technology Officers and consultants like myself, choosing these giants has been, for years, an obvious choice, a no-brainer. I myself have migrated hundreds of domains to Cloudflare, set up thousands of mailboxes on Google Workspace and Microsoft 365, and deployed critical infrastructure on AWS. Why, because they are cheap, innovative, and, most of the time, they work incredibly well, they are unbeatable allies of efficiency.

The Private Cloud Illusion. 

 The problem arises when we believe we can isolate ourselves from the risk. Many colleagues claim that having their own private cloud eliminates their dependence on others, saying that's why they have their own private cloud, so they don't depend on anyone. However, this is an illusion. Even if you have your servers in your own basement, it is highly likely that your system consumes third-party APIs, uses public DNS, or requires libraries that depend on this concentrated infrastructure. If Cloudflare goes down, half of the applications your private cloud needs to talk to the outside world will stop responding. If AWS has a failure in the us-east-1 region, authentication services, payment gateways, or logistics systems you rely on will stop working. When these giants experience difficulties, your system, no matter how private, will be severely impacted. It is a systemic, invisible, and terrifying dependency.

The Lords of the Cable. 

 To size up the problem, let’s look at who really holds up 99.99% of the Internet. The Cloud, which is the infrastructure, is controlled by just three companies: Amazon AWS, Microsoft Azure, and Google Cloud, which together control about 67% of the entire global cloud infrastructure market. If we add Alibaba, the figure exceeds 75%. The rest of the world fights for the remaining percentage. When it comes to web traffic, including Content Delivery Network and security, Cloudflare has become the doorman of the Internet, with about 20% of all websites in the world and nearly 80% of those using a Content Delivery Network depending on them. If Cloudflare were to unplug a cable, a fifth of the web would instantly vanish. Behind the clouds are the carriers, known as the backbone or Tier 1 ISPs, which are companies like Cogent, Lumen, formerly known as CenturyLink, and Telia, that own the main highways. There are very few of these companies, and when one fails, as happened with Cogent a few years ago, global routing breaks, and entire countries suffer latency or disconnection.

Public Policy or Strategic Diversification?. 

 Yesterday's Microsoft blackout is a reminder that we have put all the eggs of the digital economy into very few baskets. This brings us to an urgent crossroads. We must consider whether to discuss a global antitrust public policy to avoid this systemic concentration and force a return to the Internet's original, distributed architecture, a path that is difficult and politically complex. Alternatively, the solution may lie in the hands of tech decision-makers, who could stop blindly consuming the number one provider by default. Instead, the real resilience strategy for 2026 might be to bet on provider number 3, 4, or 5, looking toward regional data centers and niche cloud providers that offer sovereignty and diversification, even if they do not have a thousand features. Continuing to feed the giants is comfortable and cheap, until one random Tuesday, a failed update in Redmond or Northern Virginia causes a disruption, and it becomes clear that control was never truly in our hands.