The internet experienced a significant disruption on Tuesday, November 18, 2025, after Cloudflare, the US internet infrastructure firm, suffered an internal failure that affected its global network. Because many services rely on Cloudflare to stay online, the outage affected millions of people and disrupted a wide range of services, from social platforms to AI tools. The incident highlighted how heavily the internet relies on a single provider and how quickly things can break when that provider experiences a core failure.
What Cloudflare actually does
To understand why the outage spread so widely, you need to know what Cloudflare does behind the scenes. Cloudflare sits between you and the websites you open. It speeds up those sites, protects them, and maintains stable connections.
When you visit a website that uses Cloudflare, your request passes through Cloudflare’s network. The system checks for threats, optimises the connection, and sends the content back to you without exposing the website’s original server. This makes websites safer and more reliable.
Cloudflare’s core functions include:
- Content delivery for speed: It operates a massive global CDN that stores cached copies of websites in data centres close to you, reducing delays.
- Security and protection: It protects websites from attacks, especially DDoS attacks, and powers Zero Trust systems that decide which users or devices should be allowed through.
- DNS management: It helps run the DNS system that turns website names into IP addresses. If this layer has issues, you may not be able to reach a site, even if the site itself is online.
- Network reliability: Cloudflare utilises an Anycast setup, where the same IP address is present in multiple data centres worldwide. Your request is routed to the closest available centre, and if one location fails, another one takes over.
What happened and who was affected
Around 11:00 UTC (12:00 WAT), websites that use Cloudflare began returning widespread 500 errors. Cloudflare’s own dashboard and API also went offline, which made it harder for the company to respond. Platforms like X, Facebook, OpenAI, and Perplexity were among the many services users could not access.
Cloudflare later confirmed that the problem came from an internal subsystem responsible for monitoring the health of its load balancers. This was not a hardware issue. It was a failure inside the logic layer that decides how traffic is routed across Cloudflare’s global network.
Because Cloudflare handles a vast share of global web traffic, any internal failure spreads instantly. Even services with their own backups had no protection because they still relied on Cloudflare to connect users to their platforms.
How the November 18 outage unfolded
The outage escalated from early warning signs to a global incident within minutes.
When the first signs appeared
Around 6:00 AM ET (12:00 PM WAT), outage trackers showed unusual spikes. Cloudflare confirmed it was investigating a significant network-wide issue. Users encountered 500 errors across websites that relied on Cloudflare, and platforms such as X, Facebook, and several others stopped loading.
By 11:48 UTC (12:48 PM WAT), Cloudflare publicly acknowledged that multiple customers were affected. Their dashboard and API were still offline, limiting the company’s ability to monitor and respond.
Why this was not scheduled maintenance
Cloudflare had routine maintenance scheduled for a handful of regional data centres that day, but these tasks were limited and not expected to affect global traffic. The scale and speed of the outage showed that the disruption was unrelated. Cloudflare later confirmed that the monitoring system responsible for checking the health of its load balancers had failed. The timing overlapped with the maintenance window, but a separate and unexpected internal failure caused the outage itself.
Why the load-balancing system failed
Cloudflare’s findings showed that the outage originated in the subsystem responsible for monitoring the health of its load balancers.
Load balancers distribute incoming traffic across multiple servers to prevent overload and maintain site stability. They depend on real-time checks from the monitoring system to determine which servers are healthy.
When that monitoring system failed, the load balancers lost visibility into server health. Without accurate data, the system defaulted to a conservative safety mode, treating many servers as unsafe. Cloudflare could receive incoming requests, but could not route them correctly. This resulted in the widespread occurrence of 500 Internal Server Errors that users encountered across various platforms.
The situation worsened because Cloudflare’s dashboard and API were also offline. This confirmed that the issue was rooted in the central control layer, which manages routing across the entire network. The Anycast setup could not help because the failure was not in the servers, but in the logic that determines where traffic should be directed. Once that layer collapsed, global routing broke at the same time.
Services and platforms affected
The outage spread across multiple industries almost instantly, demonstrating the deep integration of Cloudflare into the everyday tools you use.
1. Social platforms, news, and finance
X and Facebook experienced significant downtime, leaving users unable to load timelines or pages. Financial and entertainment sites such as Letterboxd and bet365 also went offline, disrupting browsing, betting, and other activities.
2. AI and utility platforms
AI services were heavily affected. OpenAI and Perplexity displayed errors or Cloudflare challenge screens. Many users received a message asking them to unblock a Cloudflare challenge, which occurred because the security layer was unable to verify traffic after the routing logic failed. With Cloudflare’s internal tools offline, those challenges could not be processed, leaving legitimate users locked out.
3. Outage trackers
Even Downdetector, a tool people rely on to check if services are down, loaded inconsistently. This showed how dependent support tools are on the same infrastructure that was breaking.
Why a Cloudflare outage brings down so much of the internet
The outage highlighted the internet’s dependence on Cloudflare’s performance, security, and stability. When Cloudflare’s core control layer failed, the impact spread across social platforms, financial services, AI systems, and essential tools almost instantly.
Cloudflare usually keeps services online by routing traffic to the closest working data centre, but that protection only works when the control layer is functioning. On November 18, the control layer failed, not the servers. Once the monitoring system went down, the network could no longer trust any routing path and stopped handling traffic correctly. This caused widespread 500 errors and locked users out worldwide.
The error messages made the problem clear. The 500 errors showed that Cloudflare’s internal logic was failing. The “unblock challenges” messages on platforms like OpenAI indicated that the security layer had switched into a protective mode due to unreliable routing data. Cloudflare’s API was also down, so the system could not clear those challenges, leaving users stuck despite doing nothing wrong.
