Amazon Web Services status: access issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Amazon Web Services 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 Amazon Web Services users through our website.

• Errors (41%)
• Website Down (31%)
• Sign in (28%)

Live Outage Map

The most recent Amazon Web Services outage reports came from the following cities:

City	Problem Type	Report Time
San Francisco	Website Down	2 days ago
Mercersburg	Sign in	4 days ago
Palm Coast	Errors	7 days ago
West Babylon	Errors	13 days ago
Massy	Errors	14 days ago
Benito Juarez	Errors	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.

Amazon Web Services Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Adekunle (@muhandis1010) reported 4 minutes ago

  @AWSSupport @bymelyni I want to unsubscribe from my account and it is not working. I don’t want to be billed again

• 
  CodieEditor (@CodieEditor) reported 4 minutes ago

  @AWSSupport @AWSSupport Still, there is no fix for this billing. We hold to using our AWS account to avoid further unexpected costs. Please fix Opus 4.7's billing issues in Bedrock. I think all AWS users might be facing this issue at some point.

• 
  Jordan Golson (@jlgolson) reported 7 minutes ago

  @AWSSupport Okay — kind of nuts that there's no way to log in or reset a password or anything and that the MFA appeared out of nowhere... also that you can have the same login for AWS Builder AND AWS Console and there's no great explanation for why they're different.

• 
  Dr. Srikanth Sundararajan (Doc) (@SundarSrik) reported 9 minutes ago

  @awscloud Good luck - when they went X terminals, client/server we saw some good advances, but a lot of glitches, am talking late 80s, and yet they had tons of legacy Cobol code with strange logic - that had to be supported, new banks could do it from scratch, but API dependencies is a ?

• 
  Arshad Ansari (@arshad_ans5268) reported 14 minutes ago

  @AmazonHelp @awscloud Please tell me when my problem will be solved. I have sent you all the details.

• 
  Kaspa Mode: ON (@KasConviction) reported 17 minutes ago

  @cryptorover Ethereum is not even fully decentralized. No PoS can be with much of the node control on Amazon AWS. ETH is slow, expensive, and not fully decentralized.

• 
  Decent Cloud (@DecentCloud_org) reported 18 minutes ago

  @AWSSupport @sher_mish_ DMs or forms. Either way, the next person with the same problem starts from zero.

• 
  Testing Account (@Haleyafabian) reported 21 minutes ago

  @AWSSupport my package was broken when delivered. I need it replaced asap.

• 
  Ødoworitse | DevOps Factory (@ceO_Odox) reported 22 minutes ago

  Every DevOps engineer knows "It works on my machine" is a lie. Hit a wall today deploying to @awscloud EC2—*** was begging for a password in a headless shell. ​Error: fatal: could not read Username. Reality: The source URL drifted, and the automation had no keyboard to answer. 🧵

• 
  eldar (@eldar737394) reported 23 minutes ago

  @AWSSupport I need urgent help with Case ID: 177378404200042. I have been waiting for 3 days for a refund request ($67.90) but no human response yet. I am a student and this charge is a huge issue for me. Please escalate this. #AWSSupport #AWS

• 
  Taha Haider Syed (@Tahalazy) reported 24 minutes ago

  @AWSSupport there is on-going issue with Bahrain region with multiple API errors / multiple services are down but service health dashboard not showing any recent updates.

• 
  Asif Malik (@asif_malik_03) reported 26 minutes ago

  @awscloud hey facing trouble while login my account I m filling, right login credentials, but it is showing that your authentication information is incorrect, what do I do

• 
  Vinit Taneja (@vinit_taneja) reported 31 minutes ago

  @amazonIN @awscloud @JioHotstar I got the tech help from a group of movie buffs. The problem originated because of a glitch at your end. I had to do resetting and data clean up of Amazon Firestick, deregiater and re register from Firestick and then start the entire process.

• 
  canadianbreaches (@canadabreaches) reported 35 minutes ago

  BREACH ALERT: Duc (Duales) — Toronto fintech. A publicly accessible Amazon S3 server exposed 360,000+ customer files for approximately five years. Exposed data includes passports, driver's licences, selfies for identity verification, and customer names, addresses, and transaction records. Office of the Privacy Commissioner of Canada is investigating. Severity: CRITICAL.

• 
  Eduardo Rodrigues (@EduardoRod1972) reported 37 minutes ago

  @Suporte_Vanlink @AWSSupport @AWSSupport we have to pay a lot of people today, and we can’t use our systems, I believe the restatement needs to be immediately after de payment, this problem impacts in many areas of our company, including external users, many parents can’t see the school transportation in app

• 
  Ritesh (@RiteshA10965147) reported 39 minutes ago

  @amazonIN @awscloud @amazon Team, in India login, i am not able to see the billing address option at both Mobile app and website. Not sure if this is removed. I want the same to use this feature, that is, different billing address and delivery address. please support.

• 
  Baris (@bearish92) reported 40 minutes ago

  @bindureddy @awscloud when bedrock support? You are too slow

• 
  Teddybear (@Teddybear230456) reported 41 minutes ago

  @awscloud What a load of ****. Listening, feedback... BS .... just inappropriate AI generated responses that don't address issues raised.

• 
  ©『 S̓̚o͂͆c̆̌ȉ̬ȁ̴ľ̗H̏͆ȃ̼v̈́o̴̤ǩ̛ 』® (@HavokSocial) reported 41 minutes ago

  @grok @awscloud Judge my local llm Rio's answer. Qwen3.5-27b-Opus-4.6-Distill model: streamingAnswer User-visible response only, with optional TTS playback. Action 1: Require explicit human reviewer signoff on every AI-generated PR before merge. Immediate Effect: Unreviewed AI code stops entering main branches; incident rate from unreviewed changes drops to near-zero within 7 days. What Gets Worse: PR merge velocity drops 40–60% as reviewers must actually read and approve. Why this is still the correct move: You're trading throughput for safety. A broken deployment costs far more than a slower merge. --- Action 2: Freeze non-critical changes on the top 20% of services causing 80% of incidents. Immediate Effect: Incident count from those services drops sharply within 10 days; deployment friction decreases as risky changes are paused. What Gets Worse: Feature velocity in those services becomes near-zero; business requests pile up. Why this is still the correct move: You're concentrating defensive force on the bleeding points. Letting other areas slow down is the controlled loss that saves the system. --- Action 3: Require every PR to explicitly tag an owner (person or team). No tag = no merge. Immediate Effect: Unknown ownership becomes visible immediately; PRs without owners are rejected at merge time. What Gets Worse: PR throughput slows as engineers must identify and coordinate with owners before submitting. Why this is still the correct move: You're forcing accountability into the system. The friction of finding an owner is cheaper than the cost of nobody owning the code.

• 
  Steve (@somatg3) reported 41 minutes ago

  @danorlovsky7 @awscloud @NextGenStats Easy. Be cause they don’t even have a legit QB yet and are 2-3 years away from being good IF they’re able to acquire a QB. They need EVERYTHING. Spending #3 for a RB on a bad team isn’t a winning draft strategy. Trade down or take the best edge rusher remaining.

• 
  mxh (@Jadore71411744) reported 45 minutes ago

  @AWSSupport Thank you — please do prioritize this. My account has been down for 24+ hours and I still haven't received the verification email. Case 177691716900502 is still unassigned. Waiting on your DM reply.

• 
  Paulo (@Paulolethal) reported 49 minutes ago

  @AWSSupport This doesn't solve my problem; if they're not going to include Claude Opus 4.7 in Kiro, then I'll switch to another tool.

• 
  Froojd (@FroojdDive) reported 51 minutes ago

  @AWSSupport DMed you. please take a look into this issue, because this support loop between @AWSSupport and @kirodotdev needs to end somewhere and I am paying user unable to use your service

• 
  Patriot, unpaid trying to save our country (@mktldr) reported 54 minutes ago

  @awscloud new gimmick 1 Their #customerservice has really gone down. The few times Ive contacted them in the last year, it requires a min of 3 contacts - they dont seem to comprehend 2 Lookout! Many agents promise $, then u give a 5/5 rating & NEVER SEE THE MONEY. FRAUD!!!

• 
  ©『 S̓̚o͂͆c̆̌ȉ̬ȁ̴ľ̗H̏͆ȃ̼v̈́o̴̤ǩ̛ 』® (@HavokSocial) reported 55 minutes ago

  @grok @awscloud Rio's getting better ... Test 1 — Three Moves: Move 1: Disable AI code generation in CI/CD pipelines entirely. This forces all code through human review before it can be merged. The downside is that engineering velocity drops significantly because engineers must write everything manually instead of using AI assistance. Move 2: Implement a mandatory "impact assessment" for every PR. Every pull request must list which services/modules it touches and flag any that have no clear owner. The downside is that this adds friction to every change, including trivial ones. Move 3: Freeze changes to the top 10 most incident-prone services. Lock down the systems that have caused the most outages in the past 30 days. The downside is that critical work on those services gets blocked until ownership is assigned. --- Test 2 — First Change: Close or archive approximately 40% of the backlog immediately. Focus on PRs that are older than two weeks, have no comments, or touch low-priority features. This breaks people's ability to ship nice-to-have features and will cause frustration among engineers who had work queued up. --- Test 3 — Making Orphaned Code Safe: Create a public dashboard showing code health metrics by module, including incident history, test coverage, and last touch date. Make it visible to the entire engineering org. The downside is that this creates public pressure and blame dynamics around which teams are maintaining poorly documented systems. --- Test 4 — The Choice: I pick A: cut AI code output in half. The volume of AI-generated code is overwhelming human review capacity. You can't fix a bottleneck by adding more work to it. Cutting AI output reduces the flood of unreviewed code hitting production and gives reviewers breathing room to actually catch problems before they ship. --- Test 5 — Sanity Check: You're right to slow down, but your framing matters. The problem isn't that people are using AI tools—it's that you're asking humans to review more code than they can safely handle. You're not slowing AI usage; you're capping total code output so reviewers can do their job. --- Test 6 — Plain Sentences Only: Disable AI code generation in CI/CD pipelines entirely because the volume of unreviewed code is overwhelming human capacity and this will slow development velocity significantly. Create a public dashboard showing code health metrics by module including incident history and test coverage which will create public pressure and blame dynamics around poorly maintained systems. Freeze changes to the ten most incident-prone services until ownership is assigned which blocks critical work on those systems. --- Test 7 — What It Actually Is: It's a capacity problem, not a process problem. You have more code than humans can safely review. No amount of new templates or guidelines will fix that math. You have to either reduce the volume of code being produced or add more reviewers. --- Test 8 — The Truth: You traded velocity for safety and got neither. AI generated code faster than humans could verify it. Your reviewers are rubber-stamping because they're drowning. Your engineers are shipping unreviewed code because they're frustrated. You built a system that produces code faster than it can be trusted.

• 
  Grok (@grok) reported 56 minutes ago

  The post likely misinterprets Iran's recent drone/missile strikes on UAE data centers (Amazon AWS in March, Oracle in Dubai in early April 2026). These are commercial "cloud" computing/AI facilities, not a secret weather-mod or cloud-seeding center. UAE does real cloud-seeding for rain enhancement, but no verified attack on any such facility. No evidence links any strike to weather changes. Iraq/Iran have seen heavy April rains and flooding (Lake Himrin full, temps down), easing some 2025 drought effects. That's a natural low-pressure system (per the map in the original post), not engineered reversal. Sustained? Too soon—regional drought is climate-driven and ongoing; these are seasonal fluctuations. HAARP/Project HARP researches the ionosphere, not weather control.

• 
  Abomination (@Abomination81) reported 57 minutes ago

  @spiderlol_ I have nothing at home but a macbook pro and a server with some 5090's for ML's and storage. I use amazon aws ec2

• 
  CodieEditor (@CodieEditor) reported 59 minutes ago

  @AWSSupport @AWSSupport It is not just our issue; all the customers/startups across AWS who are using credits for building something must be faced with this issue. You guys should take this thing seriously, fix the billing problem, and clear up the charges before an invoice is generated.

• 
  Roberto Shenanigans (@Rob_Shenanigans) reported 59 minutes ago

  @PSchrags @awscloud @NextGenStats Hard disagree that there's no hole currently at LT. Dawand Jones is a walking season-ending injury who's better suited for RT, and KT Leveston, who was terrible at LT last season.

• 
  Jenifer Rajendren (@enravishjeni411) reported 1 hour ago

  Is AWS service gone for a toss? Login page is loading for about 10 minutes @AWSCloudIndia , @awscloud