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 (54%)
• Hosting (31%)
• E-mail (8%)
• Domains (8%)

Live Outage Map

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

City	Problem Type	Report Time
Asnières-sur-Seine	Cloud Services	2 days ago
New York City	E-mail	7 days ago
Township of Evan	Hosting	7 days ago
Ahmedabad	Cloud Services	13 days ago
Le Puy-en-Velay	Cloud Services	14 days ago
Ann Arbor	Cloud Services	16 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:

• 
  Colby Hill (@Chill_Investing) reported 6 minutes ago

  A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY. Here's a full recap: 1. Middle East $SPY $QQQ Stocks sold off after the U.S. struck Iranian military assets and Tehran retaliated with missile and drone attacks, putting the Strait of Hormuz back in focus. Crude jumped about 5%, with Brent trading above $90. The S&P 500 fell 0.70% to 7,631, the Nasdaq dropped 1.03% to 26,100, and the Dow slipped 0.79% to 52,767. 2. Bonds $TLT The global bond selloff continued, with U.S. 30-year yields holding at levels not sustained since 2006 and Japan's 10-year briefly touching 3% for the first time in 30 years. The U.S. 10-year sat near 4.77%. Rising discount rates hit long-duration growth names hardest, which is how the tape traded all session. 3. Dell $DELL Dell reported fiscal Q2 revenue of $47.0 billion, up 58% year over year, with adjusted EPS of $7.04 against a $4.88 estimate. Record AI server orders pushed backlog to $95 billion and management raised full-year revenue guidance to $192 billion. Third-quarter guidance of $49.0 billion in revenue and $6.50 in EPS came in well above the $41.42 billion and $4.49 analysts expected. Shares fell 6.9% during the session to $424.36, then rose about 7% after hours. 4. Apple $AAPL John Ternus took over as chief executive and Tim Cook became executive chairman, ending a 15-year run that grew Apple from roughly $350 billion to about $4.6 trillion. Ternus, a 25-year company veteran who led hardware engineering, also joined the board. Shares rose 2.62% to $325.14, the strongest mega-cap performance on an otherwise red day. 5. Palo Alto Networks $PANW Fiscal Q4 revenue rose 34% to $3.41 billion against a $3.35 billion consensus, with adjusted EPS of $1.02. Next-Generation Security ARR grew 63% to $9.10 billion, and fiscal 2027 revenue guidance of $14.10-14.20 billion topped estimates. The company also announced it acquired Console, an AI-native agentic platform being folded into Cortex, with terms undisclosed. Shares still closed down 5.2% to $362.25. 6. MongoDB $MDB Revenue of $771.8 million grew 30% year over year, the fastest rate in several years, and adjusted EPS of $1.90 came in ahead of consensus. Management raised full-year revenue guidance to $2.99-3.03 billion. Shares fell roughly 14% after hours anyway, with expectations having run further than the print. 7. Fed $SPY Futures now imply roughly 68% odds of a 25 basis point hike at the September 16 FOMC meeting. Core PCE is running at 3.3% with headline at 3.7%, and Chair Kevin Warsh has been explicit about the need for further tightening. Energy-driven inflation from the Hormuz disruption complicates the read further. 8. Moderna $MRNA Moderna climbed 9.93% to $154.28, extending its run on continued analyst upgrades following the intismeran autogene Phase 3 melanoma readout with Merck in the 1,137-patient INTerpath-001 study. William Blair moved the stock to Outperform. Merck rose 1.48% to $149.94. 9. Axon $AXON Axon was the worst performer in the S&P 500, falling 8.52% to $518.30. Demand is not the problem, with net revenue retention at 123%. The pressure is coming from a premium multiple meeting rising discount rates, compressed operating margins, and heavy R&D and acquisition integration spend. 10. Software $CRWD $NET High-multiple software took the brunt of the rate move. CrowdStrike fell 6.90% to $215.07 and Cloudflare dropped 6.44% to $285.47. With yields climbing and oil spiking, expensive growth software was the wrong place to be positioned. 11. Duolingo $DUOL Evercore ISI upgraded Duolingo to Outperform and doubled its price target to $210, citing product gains and engagement trends. Shares rose 7.10% to $158.89, one of the few green names anywhere in growth tech. 12. Crypto $COIN $CRCL Bitcoin slid about 2.2% to roughly $77,076, pulling the crypto-linked equities down with it. Coinbase fell 6.00% to $176.83 and Circle dropped 6.36% to $89.48. Interactive Brokers fell 7.00% to $90.42. 13. Energy $XLE $XOM $CVX Energy was the only sector with real strength as crude surged. The sector ETF gained 1.28%, with Exxon up 2.24% to $164.56 and Chevron up 2.38% to $211.05. The defensive tilt showed elsewhere too: the low-volatility ETF fell just 0.20% against a 2.12% drop for high beta, while gold slipped 2.83%. 📈 CHILL INVESTING — YOUR DAILY STOCK MARKET RECAP

• 
  JovanHuttonPulitzer™ אני לא סובל טיפשים! (@JovanHPulitzer) reported 7 minutes ago

  @FinalCutTile So as a tiler and tile installer, how do you think the Internet runs? I vote this the NEWEST stupid *** comment on X regarding #datecenters this week: The statement is false. Data centers are needed to run the internet as it actually exists today. What the internet actually is: The internet is a distributed “network of networks.” There is no single control room. Packets can theoretically move between two computers connected by routers and cables without a giant warehouse full of servers. Early ARPANET and 1980s–90s internet ran on university machines, government systems, and smaller server rooms. That part is true. That is not the internet anyone uses. What actually happens when you “use the internet” Almost everything people mean by the internet lives in data centers: Websites, search, email, social media Streaming (YouTube, Netflix, etc.) Cloud services (AWS, Azure, Google Cloud) Banking, payments, maps, messaging DNS, CDNs, and most of the content that gets delivered to your device Sources across Microsoft, Cloudflare, industry analyses, and infrastructure reporting describe data centers as the physical facilities that store, process, and serve that content. Without them, the pipes still exist, but the useful destinations do not. One infrastructure expert put it bluntly: the rudimentary network can still function, “you just can’t do anything that you’re used to on the internet because that’s all posted out of these datacenters.” Calling data centers the “backbone” or “brick-and-mortar of the digital world” is standard, not hype. The distinction that the statement ignores Network layer (cables, routers, ISPs, peering): can exist without hyperscale data centers. Application / content layer (everything you actually open, stream, search, or log into): overwhelmingly depends on them. The insult in the original statement does not change the facts. Data centers are required for the internet people actually use.

• 
  Krishna Singh (@krishnasinghdev) reported 9 minutes ago

  @tapasadhikary @Namecheap I never liked namecheap / godaddy, don't know why people choose it over simple providers like cloudflare where they charge same amount for 1 or 10 years

• 
  Atif (@ioAtif) reported 10 minutes ago

  @acolombiadev @Cloudflare @coderhq Although it's not a service but a whole platform

• 
  yakuzer5 (@conckrete) reported 11 minutes ago

  @OmegaNekoSimp Why are people mad about this? It's a nice break to see this instead of blank page or cloudflare ****

• 
  Selenka (@SelenkaOnChain) reported 14 minutes ago

  Netnet Capital team is flexing metrics and talking about successful launch of Subway Runner... Meanwhile, 2 vibecoders literally drained their entire mechanic and became the #1 and #2 top holders. Here is the breakdown directly from one of the guys who farmed them: 🧵👇 "First decent cook of the bull run (if we’re even in one). Spent the last few months trying to get good at on-chain analytics, so hadn't been actively cooking. Two nights ago, I'm watching the feed and notice everyone sending $10 clips to this CA: 0x8154e35166f21305adac82f95b54de8acd44d23a. Instantly smelled pure degen activity. Hit up the group chat, did some digging - turns out it’s a Subway Surfers / Chrome dino style runner game by NetNet. Their shitcoin was sitting at a $90M cap at the time, so I figured if their token is holding that kind of valuation, there’s definitely meat on the bone. Normally, their games are pure casino trash: deposit cash, pray to RNGesus, or get rekt. But why not test it? Played a run manually and noticed that at the end of each game there was a draw. Checked their TG and reverse-engineered the contract - turns out there’s an N% chance to win an NFT from their collection. Their previous official collection had crazy volume and peaked hard, so my degen senses started tingling: this was a hidden gem. Literally 15 minutes later, they pause the game. Perfect timing - gave us room to prep. By that point, I had already captured the WSS traffic and requests. Their game logic was using a basic commit-reveal scheme: courseCommit = keccak256(serverSecret) seed = keccak256(serverSecret ++ playerSalt ++ runId) Meaning: right at game start, the server literally sent us the secret, allowing us to compute the seed and reconstruct the entire track in advance. The game loop: 3 lanes, 30 coins, 3600 ticks to finish (60 seconds). You dodge obstacles while longing/shorting NVDA. 3 hits = you lose the full $10. Complete a flawless run = you collect all coins to refund your stake and it only burns ~$0.20 in fees, giving you an almost free roll at the NFT lottery. In the era of AI and vibecoding, this was child's play. My boy XXX and I started spinning up bots in parallel. Ended up deploying his script since he coded it faster. We simulated runs, got a 100% win rate, set up a websocket listener for when the contract unpauses, and went to bed. Woke up at 6 AM, and literally 30 seconds later the game goes live. We spun up the bots - 5 minutes in, we already bagged our first NFT. Then came the scaling phase. At first, the team didn't give a single **** - no Cloudflare, no rate limiting, not even a basic 429. We ran 10 wallets simultaneously. After a few hours, they finally threw in a primitive 429, and that was it. No bot protection, no captcha, nothing. They didn’t even enforce single-session checks per wallet, so we were running multiple concurrent instances on the exact same address (literally impossible to do manually). The bots printed flawlessly. At one point, our load was crashing live games for actual manual players, forcing the team to repeatedly pause the game to fix lag. Every time they brought it back up, we resumed blasting. Total mint size was 1,060 NFTs. We scooped over 20% of the entire supply. Then came the funny part: the collection had zero secondary volume. Time for a little social engineering. We hopped into their TG playing dumb, gently nudging the admins: 'Hey guys, might want to tweet that the game is live and apply for OpenSea verification!' The final tally: * Total capital spent on fees/burns: ~$1,700 * Total NFTs pulled: ~50–60 pieces (friends got a similar bag) * PnL: Dumped most of the floor tier into bids today at $200–$300 a pop, still holding some. Nothing crazy for a real bull run, but an easy 5-figure profit for a couple of hours of vibecoding."

• 
  Eli (イーライさん) (@niceEli1) reported 16 minutes ago

  @tekbog I literally migrated an entire service from one codebase made in C# to Rust with a few seconds of downtime (only was caused by cloudflare tunnels changing the domain from one internal service to another one). It isn’t hard to just make software work without a maintenance period.

• 
  Hanif Carroll (@HanifCarroll) reported 17 minutes ago

  At a previous job, I was lucky to work with a staff-level front-end engineer who was very good at what he did and thorough like no one else I had worked with. I didn’t always appreciate that thoroughness. At first, I found it maddening. Meetings with product and design would drag on because he always had an enormous list of questions about whatever feature we were discussing. I was usually eager to finish the meeting and start coding. It took me a while to understand that he wasn’t making the process longer. He was asking the questions we would otherwise have to answer halfway through development, when an unanswered question could block the feature or send it in the wrong direction. Every unanswered question holds an assumption, and that assumption might not match what product or design had in mind. I thought about him recently while reviewing the search process for Casamo, a product I’m building to help people compare furnished stays. Before changing the code, I started asking questions: Why were we processing listings already known to be over the user’s budget? Why were hotel-like properties entering the process when the user had asked for an entire place? Why did we stop after finding six suitable stays if ten were available? How old could a review be before it stopped being useful evidence? Those questions exposed decisions that had been made when Casamo ran on a low-powered VPS and needed to do as little work as possible. They didn’t all make sense now that the product runs on Cloudflare. We changed the search process to reject known mismatches earlier, treat six results as the minimum rather than the target, continue until it finds ten suitable stays, and only use reviews from the previous year as evidence. If I had started coding immediately, I probably would have made the existing process faster without questioning whether it was still the right process. I don’t know if I’ll ever have as many questions as he did, but now I understand what he was protecting us from.

• 
  Sergei Aksjonov (@sergeiaksjonov) reported 22 minutes ago

  @Printful Also this: 1 day of my time 1 domain on Cloudflare $20 on AI That’s the whole budget. 10 years ago I thought you needed a Social account of the service, a pretty brand, a PR campaign, the “right” image. This time I just launched it as is from my own account. No polish. No theater. Just shipped. And the first order still came in. Just start as it is. Let it flow.

• 
  Utsav Patel 🇮🇳 (@mr_utsav_patel) reported 26 minutes ago

  @Cloudflare @CloudflareHelp @Yank if you could help here...

• 
  Nathan Flurry 🔩 (@NathanFlurry) reported 29 minutes ago

  AI-generated Rivet Actors / Workflows are (finally) here 🥂 Some design notes on Dynamic Apps: Powered by agentOS → provides lightweight sandbox-as-a-library V8 isolate runtime → generated apps scale to 0, cold starts in ms, 22 MB per app, native JS performance (not slower WebAssembly like QuickJS) Novel Node.js-compatible runtime → ground truth agents already know & existing libraries work, native performance No nested virtualization → pure userspace, like Chromium & Cloudflare Workers Self-hostable → Dynamic Apps can run anywhere, including Kubernetes, Railway, EC2, etc (which don't support microVM / KVM) SQLite sharded by actor → scalable, fast, cheap, uses no compute when idle A Rivet namespace per app → isolates its actors, separates billing, and nothing to provision Builds run inside agentOS → npm install & build steps in the WASM sandbox, like a Dockerfile without the Linux VM Dynamic Apps also supports plain REST backends & static frontends deploy the same way

• 
  Reina Cruz 🥼🧤🇨🇺 (@rea1ReinaCruz) reported 31 minutes ago

  @Cloudflare Fix human verification

• 
  Gaylord (@laolu_afolabi) reported 31 minutes ago

  @lanreadelowo From Claude: """ Technically, yes — it's buildable, and "platform-agnostic" is realistic because most providers and gateways (OpenRouter, LiteLLM, Portkey, etc.) have already converged on an OpenAI-compatible schema, so a routing layer that swaps providers behind one endpoint isn't a huge engineering lift. But there are a few things in the premise worth stress-testing before you build. **The "discount arbitrage" framing is a bit shakier than it sounds.** Most of these gateways aren't reselling models at a discount — they're reselling at close to list price plus a fee. While OpenRouter states it doesn't mark up inference pricing, its credit purchase fee effectively acts as a surcharge on all usage. LiteLLM's pitch is the opposite direction — zero markup by self-hosting, with OpenRouter's ~5% fee costing meaningfully more at scale. So the "discounts" you'd be arbitraging are less "provider A is cheaper than provider B for the same model" and more: rate-limited free tiers, short-lived promos, or BYOK arrangements — OpenRouter's BYOK is free up to a monthly allowance before a 5% fee kicks in. Those are real, but they're capacity-constrained and often disappear once volume shows up — hard to build a durable business on chasing them. **The bigger product risk: tokens/models aren't fungible.** Swapping Gemini for Claude for GPT under one endpoint isn't like arbitraging cloud spot instances — different tokenizers (so "$/token" isn't apples-to-apples), different tool-calling formats, different output style and capability. If your router silently reroutes a customer's traffic to whatever's cheapest this hour, you can quietly degrade the app built on top of it. That's the actual hard problem, not the proxying. **What's actually working in production right now is narrower than blind price-routing:** AT&T processes tens of billions of tokens a day using cache-aware routers that send simpler tasks to cheaper models, cutting costs up to 56% with only ~2% quality degradation, and this "model cascade" pattern — cheap/fast models handle the default case, only escalating to a frontier model when a confidence check fails — is used by both AT&T's LiteLLM-based router and Databricks' Smart Router. That's routing by task-fit with cost as a constraint, not routing purely by whoever's running a promo this week. **Competitive landscape is already crowded**: OpenRouter, LiteLLM, Portkey, Requesty, Helicone, Not Diamond, Martian, TrueFoundry, Cloudflare AI Gateway, Ofox all do some flavor of unified-endpoint routing today, several with automatic fallback/cost-based rules already built in. Pure discount-hunting is a thin moat — anyone can scrape pricing pages, and providers may treat aggressive arbitrage-only usage as a ToS problem. Where I'd actually look for a wedge if I were you: quality-aware routing with an audit trail (so teams can trust that "cheaper" didn't mean "worse" for their specific task), rather than price-only arbitrage. """

• 
  robot 2.0 (@alightinastorm) reported 32 minutes ago

  2 weeks of prompting can build this now, in the browser, no platform fees, no pre-downloading gigabytes, stream as you play is it AAA? no is it low poly slop? no can you build real games with it? you absolutely can stop inventing bullshit about distribution, what distribution do you get from steam? ZERO, that's why you hire marketers to run guerilla campaigns on social media, streamers to hype it up and backdoor some cash to youtubers to pump it on release week You can do the same for the web, your likes and followers on X are your wishlist Pay stripe 3% instead of 30%, throw a couple $$ to cloudflare for traffic Or keep betting on steam and die a slow painful corporate death

• 
  Kevin Minnelli (@minnelli) reported 32 minutes ago

  WTF - Grok Bot can't fire on schedule to save it's life. The scheduled routines are just broken and at best unreliable. I want to love this product. When you set the cron job it doesn't work. It tells you try Cloudflare, sure let's set that up and burn tokens, then that doesn't fire to wake them. Oh, let's try GitHub now and use that....all failed. I had to wake it again this morning before the market opened. Anyone else feeling frustration in this regard?

• 
  2$p00ky (@dos2p00ky) reported 35 minutes ago

  No ******* way that’s really you messaging in that chat…but then I accidentally clicked the button, the chatbot responded, I sent an emoji; then you sent a name and email template & then suddenly OF was also trying to track along with ur site through cloudflare 💀 What a ******* webhook to stay connected to every chat someone has open from ur webpage authenticated by the network cert…holy **** 🥵 for that setup

• 
  Hatem Hosny (@hatem_hosny_) reported 36 minutes ago

  @acolombiadev @Cloudflare @livecodes_io the client-side code playground that can be embedded and self-hosted is free for unlimited usage! 90+ languages/frameworks Mobile support Private by default Open-source

• 
  Mert Deveci (@gm_mertd) reported 38 minutes ago

  STOP INSTALLING 55 SKILLS TO USE A SINGLE API. @firecrawl @Cloudflare If it is more than 3 skills agent needs to f... use you, your platform is already too complex and the agent won't be able to use it. Simply put it into the CLI where it can output instructions on doing --help or --best-practices Can't believe the amount of dark patterns emerging here. @vercel is guilty of the same thing

• 
  siftydeals (@siftydeals) reported 43 minutes ago

  Cloudflare had 487 outages in H1 2026. Most teams still treat uptime like a technical metric instead of a customer trust issue. Your SaaS tool only works if your customers can actually reach it. 🔧 Details in link.

• 
  Miget (@miget_com) reported 43 minutes ago

  Then you reach it from outside. Four terminators, and you can run more than one on the same network. WireGuard per person or per machine. Tailscale into a tailnet you already run. Cloudflare WARP into a Cloudflare org. IPsec site-to-site.

• 
  Paul T (@BscTraynor) reported 45 minutes ago

  @ClareCraigPath petition site receives a massive surge of traffic from X, Cloudflare may issue a "Checking your browser" challenge page or briefly queue users to prevent the server from crashing

• 
  Brendan Ryan (@BrendanRyanM) reported 46 minutes ago

  @mntruell @petergyang @bot Also having an issue getting past bot blockers, e.g. cloudflare prompts to press a button and prove you’re human. Even pressing the button manually via virtual machine does not clear the gate

• 
  Ben Hocking (@bmwhocking) reported 49 minutes ago

  @rustie5555 There are a lot of round robin API calls for content. Works fine if you are close to their US data-centre. Sucks for us. They are using Cloudflare, I don’t believe they are processing at Cloudflare’s edge, just caching content. They need better & faster timeline building.

• 
  Lex (@TheLexTimes) reported 53 minutes ago

  Hey @Cloudflare I reached out via DM about WARP constantly disconnecting. Is there a direct contact I can talk to? Happy to provide any data to help figure this issue out.

• 
  Theo Ephraim 🧙‍♂️ (@theozero) reported 57 minutes ago

  @dbmikus @ericclemmons last I tried last on cloudflare sandboxes it was quite slow. Couldnt quite just restore the vm in its running state and many hoops to jump through.

• 
  Krishna Singh (@krishnasinghdev) reported 59 minutes ago

  @tapasadhikary I believe useEffect has been the source of so many production issues, most recent one was with Cloudflare

• 
  COLLINSEO (@alexcollinseo) reported 59 minutes ago

  Breaking news! Your site might start blocking AIs from September 15th.. And if you block them, you won't show up in AI answers. The good thing? It takes a sec to unblock! This setting is inside Cloudflare. You might not even know your website uses Cloudflare because it works in the background as a CDN, helping your website load faster. If you check your website using the tools I mention, you can find out whether Cloudflare is being used on your site. Cloudflare seems to have taken this step because of how AI companies are crawling websites, and honestly, I don't completely disagree with the reason behind it. But having AI crawlers automatically blocked from September 15 could be a problem if you want your business to appear in AI answers. The setting is inside Security Settings. Look for the AI crawler control and make sure it is set to allow rather than block. This is especially important if AI visibility is part of your strategy. With one of my clients, we went from zero to around 500 visitors from AI traffic in two months. And what I did was exactly what I explained in the video. If your AI crawlers are blocked, you're potentially closing the door on that traffic before it even has a chance to reach you. #SEO #DigitalMarketing 👇🏼 Comment VIDEO for the FULL VIDEO BREAKDOWN

• 
  Reina Cruz 🥼🧤🇨🇺 (@rea1ReinaCruz) reported 1 hour ago

  @Cloudflare Fix human verification

• 
  Sathz_நிலன் (@nilan_sathz) reported 1 hour ago

  Stopping the bad guys with Cloudflare: 212 malicious requests blocked or challenged in the last month #cloudflare

• 
  junebnunny (@junebnunny_) reported 1 hour ago

  @Gion_the_critic @SportingNest @Cloudflare It's a click fix attack