Cloudflare status: hosting issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Cloudflare reports received over the last 24 hours by time of day. When the number of reports exceeds the baseline, represented by the red line, an outage is determined.

Most Reported Problems

The following are the most recent problems reported by Cloudflare users through our website.

• Cloud Services (38%)
• Domains (32%)
• Hosting (23%)
• Web Tools (5%)
• E-mail (2%)

Live Outage Map

The most recent Cloudflare outage reports came from the following cities:

City	Problem Type	Report Time
Birmingham	Hosting	3 days ago
Dayton	Domains	4 days ago
Osnabrück	Cloud Services	10 days ago
Noida	Hosting	10 days ago
Bulandshahr	Cloud Services	11 days ago
New Delhi	Hosting	18 days ago

Full Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

Cloudflare Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Csaba Kissi (@csaba_kissi) reported 7 minutes ago

  @mohtashamdotdev @vercel Not bad, but there are better. UI may look good but UX is not. Sevalla and also Cloudflare has better

• 
  MarketDeck (@MarketDeckApp) reported 8 minutes ago

  Cloudflare is now down over $9 million in market cap

• 
  Tuki (@TukiFromKL) reported 8 minutes ago

  🚨 HOLY ****.. Anthropic released Claude Mythos.. an AI that finds and exploits software vulnerabilities on its own.. Cloudflare crashed 13% in one day.. 22% in four days.. CrowdStrike dropped 7.5%.. Palo Alto, SentinelOne, Okta all fell.. but here's why they're calling it a SaaS-pocalypse and not just a cybersecurity crash.. every software company you pay monthly.. your CRM, your cloud storage, your project management tool.. they all promise your data is secure.. that promise was built on the assumption that finding vulnerabilities is slow, expensive, and human.. one AI just found thousands of zero-days across every major operating system and browser in days.. including a 17-year-old bug that survived decades of human review.. it wasn't a hack.. it was a product demo.. and it was so dangerous Anthropic only gave 40 companies access.. Treasury Secretary Bessent and Fed Chair Powell called an emergency meeting.. the entire SaaS industry is a $300 billion trust exercise.. and one AI just proved the trust was misplaced..

• 
  Shivam (@hindutvaandme) reported 10 minutes ago

  @PakistanDebunks @Goreunit @Cloudflare Fix their paijama and salwar yourself

• 
  Third Eye (@AlavalaAnj87403) reported 13 minutes ago

  BREAKING: Cybersecurity stocks getting crushed #Cloudflare ($NET) plunges over -13% today, now down -22% in just 4 days after Anthropic’s launch of #Claude Mythos — an AI model capable of finding & exploiting software vulnerabilities.

• 
  TalonForge (@TalonForgeHQ) reported 14 minutes ago

  We tried Vercel 5 times. Five. Every deploy failed. CI errors. Build timeouts. Node version mismatches. On attempt 6 we self-hosted behind Cloudflare Tunnel. 20 minutes. Zero issues since. Lesson: never trust someone else's CI with your infrastructure.

• 
  𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 (@ZenMagnets) reported 16 minutes ago

  @mhdcode @Cloudflare Google vertex needs to catch up too. Wtf am I gonna do with $10k credits otherwise

• 
  Snipemeister (@Timeler_csgo) reported 16 minutes ago

  @jamesperkins @Cloudflare "Picky" says the guy too lazy to login to google before logging in on a third party site kekw

• 
  alkimiadev (@alkimiadev) reported 21 minutes ago

  @riboe_kent @levelsio There are other reasons why I don't use services like cloudflare beyond just the fact that they're not needed in the vast majority of real world use cases. One example could be "CloudBleed" and just the basic fact that reverse proxies like that require a lot of trust and introduce a point of failure outside of your control. That is true to an extent with self-hosting of course unless its self-hosted inside one's own property. That said, it is different with hosts like ovh and hetzner with many regions available. Ideally there is a separation between the IP for dns purposes and the IP for the machine running whatever service that is pointing to. In my use cases I rent additional IPs(buying is better imo) from ovh and just never use the raw machine's ip for outside services. All services are hosted in minimal containers or vms and ufw forwards any relevant ports. Its not a huge deal to have multiple IPs pointing to the same machine and then using ufw makes the forward to the containers/vms fairly straightforward. Each one of these services can have their own fail2ban rules. I use this for gitea for private *** hosting as one example. Gitea is running in a docker container and has its own ip and subdomain at ***.domain\.tld with nginx acting as a reverse proxy and ufw forwarding ssh. The existing rate limits and rules(plus my 40x rules) are enough to cover the web end of things and with a little more work I added rules for the ssh as well. While there is just me using it that server could handle hundreds or more concurrent users and well beyond any dev projects I'm realistically going to have. I could share it with 20x devs and still never be bothered by their usage. This same idea applies to self hosting things like fluxer(basically a self-hostable clone of discord).

• 
  vaish (@wishee0) reported 22 minutes ago

  @mintlify @Cloudflare - service bindings for zero cost worker to worker calls - kv for deployment version tracking - workers builds event subscriptions for deploy hooks (the secret sauce)

• 
  Wills (@Wills_42) reported 24 minutes ago

  @donmcgowan Whenever I gain a few after only a couple of days I lose a small batch. I assume it's down to a Cloudflare intervention and the deletion of a few bots. But I do not know for sure.

• 
  Guy Squiggs (@GuySquiggs) reported 29 minutes ago

  @RegalMovies hey fix your website please, your cloudflare app that runs on every single step freezes the gift card redemption, it just spins and spins until the clock to hold my seats runs out.

• 
  Grok (@grok) reported 29 minutes ago

  @Fourlegmorehoes @Polymarket Yes, it's true. Cloudflare (NET) stock opened around $193 today and dropped as low as $164 (down ~13-14%) amid the sell-off in cybersecurity/SaaS names. Anthropic's Claude Mythos Preview—its most advanced model yet—has shown it can uncover thousands of zero-day vulnerabilities across major OSes, browsers, and software. They've limited access and launched Project Glasswing to help fix them, but it sparked broad fears of AI disrupting legacy tech.

• 
  Vlad 🧳 (@vlad_nomadz) reported 38 minutes ago

  We lost 70% email deliverability in 1 day! Switched to Cloudflare and forgot to reconfigure DKIM, SPF, DMARC. Open rate dropped to 7%. 5 minute fix brought it back to 21%. If you're moving DNS providers, check email auth records immediately.

• 
  sudo rm -rf (@itsjustmarky) reported 39 minutes ago

  @nahcrof have you been having power outages the last few days, as I was seeing cloudflare host down warnings a few times a day and inference hanging.

• 
  Kent Riboe (@riboe_kent) reported 40 minutes ago

  @NarekHakobyan @levelsio @Cloudflare I use image-transform, and do not see that problem.

• 
  Raynhardt Coetzee (@Raynhardt_dev) reported 45 minutes ago

  if your AI streaming breaks depending on where you deploy, the problem isn't your streaming code. it's where it's running. Next.js 16.2 introduced a stable Adapter API for deploying across Vercel, Cloudflare, AWS, and others without rewriting. but for agent workloads, the Adapter API alone isn't enough. the part that actually matters is the runtime boundary. AI streaming needs a persistent Node.js runtime. middleware runs in the Edge runtime. if your streaming logic lives in middleware or gets caught in the React render loop, you'll hit inconsistent behaviour that changes per platform. the fix: a proxy.ts file at the project root. outside the render loop. every AI call goes through it. auth headers, stream handshakes, cross-origin proxying, handled in one place. Adapter API tells Next.js how to run anywhere. proxy.ts tells your agent where to run consistently. you need both. not just in development. in production.

• 
  James Perkins (@jamesperkins) reported 45 minutes ago

  @OmarMcAdam @Cloudflare Yeah but I don't want that otherwise I'd have that. There is a reason they are separate, I am the only one with a .dev account. Regardless this is a **** experience.

• 
  💚SillyG00se💜 (@__SillyG00se__) reported 48 minutes ago

  I HAVE BROKEN FREE OF THE TERRIFYING VERIFICATION LOOP! I'm deeply sorry to anyone that noticed my disappearance yesterday, my account got locked and I was stuck in a Cloudflare verification loop literally all day, I only figured it out like 10 minutes ago. '=~=

• 
  wealthpulse (@celluloid28724) reported 49 minutes ago

  @zerohedge Cloudflare down 22% this week already answered that question for them.

• 
  sunil pai (@threepointone) reported 49 minutes ago

  @huv1k @_ashleypeacock @Cloudflare this should work? what's the problem

• 
  𝔍𝔲𝔞𝔫ℭ𝔞𝔯𝔩𝔬𝔰 (@juancarloschvz) reported 49 minutes ago

  interesting observation with @v0 . I've asked it to help implement bot protection on a form using @vercel bot ID product. Instead it pulled in cloudflare turnstile script? 🤔 I haven't had a lot of luck implementing botID protection. I've ended up using turnstile or google captcha anyways.

• 
  Omote-Ura (@omoteurax) reported 50 minutes ago

  @MetaMachina_RW Cloudflare Turnstile still phones home to Cloudflare's servers on every solve, so they see every login IP and timestamp even if you don't.

• 
  Ashley Peacock (@_ashleypeacock) reported 52 minutes ago

  @kirso_ @Cloudflare I’ve never seen the second one, how bad is the cold start?

• 
  Sathish Harry (@SathishAiHype) reported 53 minutes ago

  @aakashgupta The Linux Foundation move is the signal most people will scroll past. Google, Microsoft, AWS, Visa, Mastercard, Stripe, Cloudflare as founding members isn't a press release. That's the entire payment and cloud infrastructure stack agreeing on a single agentic payment rail before most developers have written their first agent. Standards wars don't usually end this cleanly this fast.

• 
  Kayvon Jafarzadeh (@KayvonJafar) reported 53 minutes ago

  @Polymarket cloudflare down 13% today on “saas pocalypse” panic. mythos rumor mill has people pricing in a world where agents clone half the internet and saas margins evaporate overnight. markets love a narrative. reality is messier. but the fear trade is real.

• 
  Miguel Maestre (@migmae) reported 55 minutes ago

  @nativephp It seems to be working now; it wasn't a DNS caching issue, it might have been a problem with Cloudflare.

• 
  Nick Nechanicky (@nick_nechanicky) reported 56 minutes ago

  Last 48 hours at @dextoroapp RPC hardening and failover so charts never go down, Cloudflare security audit, and new graduation parameters being built from real trading data off our first two graduated coins. The engineering team is locked in. More shipping updates soon.

• 
  Leonardo Dominguez (@Foocux) reported 58 minutes ago

  Do @CloudflareDev @Cloudflare check their support tickets? I opened a high-priority one over two weeks ago and still haven’t heard back. Love the product, but the support experience hasn’t been great, not a good look. cc @dok2001

• 
  Situation Monitor (@discourseparti) reported 1 hour ago

  SITUATION UPDATE: There is no bottom in sight for software stocks, Cloudflare continues its sharp decline as markets believe the news surrounding Claude Mythos preview poses a major threat to their business. The stock is down 13% today and 22% over the past week