When One Company Takes Down Half the Internet
The Cloudflare Outage of November 18, 2025
Published November 18, 2025 • Last updated ~14:00 UTC
Today, millions of us woke up to a bizarre reality: X (Twitter) wouldn’t load, ChatGPT threw errors, even the site that tracks outages — Downdetector — was down. The culprit? A single company most people have never heard of: Cloudflare.
By mid-morning UTC on November 18, 2025, Cloudflare reported an “internal service degradation” that cascaded across the globe, triggering widespread HTTP 500 errors and connection failures. It wasn’t a cyberattack (at least none confirmed), just a painful reminder of how interconnected — and fragile — our digital world has become.
What Actually Happened?
The trouble started around 11:48 UTC (early morning US East Coast time). Cloudflare’s status page lit up with alerts:
- Widespread 500 Internal Server Errors
- Dashboard and API access failing
- Intermittent impacts on CDN, DNS, security services, WARP, and Access
High-profile casualties included:
- X (Twitter) – timelines frozen, posts failing to load
- OpenAI / ChatGPT – login and query failures
- Letterboxd, bet365, parts of League of Legends multiplayer
- Ironically, Downdetector itself (it came back quickly)
By 13:00 UTC, Cloudflare announced they had identified the issue and begun rolling out fixes. Error rates dropped steadily, with services like WARP and Access fully recovered. As of this writing, most of the web is breathing again — but lingering intermittent errors persist for some users.
Why Does One Outage Break So Much of the Internet?
Cloudflare isn’t just a CDN — it sits in front of an estimated 20–30% of all web traffic worldwide, operating from over 330 cities.
When their edge fails, it’s like pulling a main power line in a city grid: lights go out everywhere downstream, even if the individual houses are fine. This is the classic single point of failure problem in modern cloud architecture.
The Human Side: Frustration, Memes, and a Forced Digital Detox
Social media (what little worked) exploded with jokes:
“Cloudflare down = the most productive Tuesday of 2025”
“Even the outage tracker is outaged. We’ve reached peak 2025.”
But for businesses relying on real-time services — e-commerce, SaaS tools, gaming platforms — this was real lost revenue and angry customers.
What Can You Do When Cloudflare (or Any Major Provider) Goes Down?
As an everyday user, options are limited — but not zero:
- Just wait — Most issues are transient. Refresh in 5–10 minutes.
- Switch networks — Try mobile data vs Wi-Fi or vice versa.
- Use the app instead of the website — Many services bypass Cloudflare routing in their native apps.
- Change DNS temporarily — Switch to Google (8.8.8.8) or Quad9 (9.9.9.9).
- Incognito mode or clear cache — Sometimes bypasses stuck Cloudflare challenges.
If you’re a site owner or developer:
- Enable “Pause Cloudflare” (Development Mode) in an emergency.
- Consider multi-CDN strategies (Cloudflare + Fastly + Akamai) for critical sites.
- Monitor status pages and set up redundant alerting.
The Bigger Question: Are We Too Dependent?
Today’s outage was bad, but it wasn’t the worst we’ve seen (remember the 2021 Fastly incident or 2022 AWS chaos?). Yet each one chips away at the illusion of an unbreakable cloud.
We love the speed, security, and cost savings these giants provide — but events like this force the conversation: How much centralization is too much?
Decentralized alternatives exist (IPFS, ENS, etc.), but they’re not ready for prime time. Until then, the best we can do is build with failure in mind.
Cloudflare will almost certainly publish a detailed postmortem in the coming days. When they do, expect brutal honesty — they have a strong track record there.
In the meantime, the internet is mostly back. Go touch grass… or refresh X one more time.
Stay safe out there on the fragile tubes we call the web.