Amazon status: access issues and outage reports

Problems in the last 24 hours

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

• Website Down (44%)
• Errors (31%)
• Sign in (24%)

Live Outage Map

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

City	Problem Type	Report Time
Castellterçol	Errors	1 hour ago
Alicante	Errors	6 hours ago
Alicante	Errors	18 hours ago
Saint-Herblain	Errors	20 hours ago
Barcelona	Errors	23 hours ago
Narón	Errors	1 day 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 Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Joan (@JoanRecall) reported 6 minutes ago

  @AmazonHelp It is forced me to login this is not my fault really I already changed my number but Amazon send OTP to unused number

• 
  _Lucas (@LucasTripp96391) reported 8 minutes ago

  This was me few months ago..... March was like a test to me I can't lie, I almost went into depression, it was so bad that, everything and everyone around me was nowhere to be found, but yeah, I still have to put on a smiling face.....I was extremely broke, I was so broke I beg for sup every now and then to get class updates. Broke and still in debt, I couldn't even tell my ex, took money to fix an issue and it was one of the most brutal mistake I ever made in my life. Took the money around November 2025 interest daily till March. I thought I had everything figured out but NO. Sometimes, people go through so many struggles, but because they're not telling you or making it obvious, you won't know. Thank God for Amazon kdp, little help from a friend I was able to pay back. A Broke man can not really concentrate.

• 
  blakmac (@blakmaclives) reported 10 minutes ago

  @nonbanjim I'm doing a lot of processing. And I'm taking breaks and doing other stuff so I don't get too bogged down. But every single problem we had was her doing that **** to make me try to fit some kind of weird image in her head. She put a book on finding true love on her amazon list. Also a **** book about sentient toy trains...you can see where that's going. The love book was put there in the last day or so. So, I removed myself from her lists and said "yup, she's gonna do the dumbest **** ever because she thinks she can use it to hurt me, but I'm already aware of that". I need to find a way to get the money I owe her so I can fully cut her off. This was bad for me. I didn't think that three days ago. It's pretty stark. So...just taking time to process it all.

• 
  Cyber Farmer (@CyberFarmerX) reported 11 minutes ago

  What is up with Amazon Shipping? Anyone else having issues? I just canceled prime. They delayed two high dollar items. Then when I needed a medical item, they said same day delivery then late at night postponed delivery until next day. Totally unacceptable.

• 
  Faisal_IND (@bond19872001) reported 13 minutes ago

  @AmazonHelp Ram, I understand your generic policy response, but it does NOT answer my actual questions.Who told Amazon to issue cashback in the first place? I never requested or authorised any cashback. ₹2,160 was charged directly from my card. So why is Amazon suddenly talking about cashback when I am asking for a refund of my own money? If ₹2,160 was actually debited from my account, why wasn't the full refundable amount returned to the same payment method? Why was cashback introduced without my consent And if Amazon is deducting ₹5 because of a “discount/fee”, why should the customer bear it when the order was delayed for days due to Amazon's own failure? If Amazon can deduct money from customers, will Amazon also compensate customers for delivery delays caused by Amazon? Please don't send me a generic policy link. Show me the exact transaction where this cashback/discount was applied, when it was issued, and why it was applied without my request. If Amazon cannot provide a transparent transaction-level explanation, I will escalate this through the National Consumer Helpline and, if necessary, the Consumer Commission with the payment, refund and customer-service records.This is not just about ₹95 anymore. It's about transparency and how customer money is being handled. How many other customers have faced the same thing without even noticing?

• 
  Faisal_IND (@bond19872001) reported 14 minutes ago

  @AmazonHelp Another shocking issue with Amazon. An order couldn’t be delivered, and now the entire order history has suddenly disappeared from my account. Even the amount shown appears to have changed from what I saw at the time of booking. Why is the order history being removed after a failed delivery? What exactly is being hidden? This looks extremely suspicious. I have screenshots as evidence.Please explain this transparently. If this isn’t resolved, I may have to approach the Consumer Forum. Lawyers/consumer-rights experts, please advise.

• 
  Zoro-san (@Karan56016109) reported 14 minutes ago

  @AmazonHelp The form still hasn’t submitted after waiting hours and retrying. My pickup failed despite the product being ready. Please escalate this as a carrier issue and preserve my approved refund-on-return while arranging the pickup. You're not resolving my issue.

• 
  Higher_waska (@JethroArnaiz) reported 15 minutes ago

  @RJB28780610 @paleochristcon @IanMalcolm84 You say it's deflection then concede the point? Then revert back to the eugenics fantasy. Have you ever wondered why Norway doesn't have an immigration problem? Might be something to do with them being 50% unionized and amazon workers paid £25 an hour not £10 like in Uk.

• 
  Amazon Help (@AmazonHelp) reported 17 minutes ago

  @nayan_bothra @nayan_bothra Please copy that link to web browser and access it from there. You can also access the link from desktop (PC or laptop) web browser. Make sure to delete all cache, cookies, history from device. Logout and login to Amazon account and try to access the link. -Akamsha

• 
  Frank Li (@ZZOOOOTTOOHH) reported 19 minutes ago

  @Pricerrors damn. amazon screwed me over making me signin again and all the verification. they do this every time. they must hate me or something

• 
  Isenpai9840 (@Isenpai9840) reported 23 minutes ago

  @JeffBezos shake hands since vision is same for all 🙏 Are there internship in @blueorigin Im good at solving fundamental problems atleast. Also @amazon not making their commercial robots?

• 
  re:printed 3D (@reprinted3D) reported 25 minutes ago

  @comingconflict @Prusa3D My K1 Max came miswired out of the box. It printed everything in a mirror image, which was fine for purely symmetrical parts, but not good otherwise. Creality had no clue what was causing it. They even stopped responding to the support ticket (much like Discord is doing now -- still no word from them). By the time I found out what was going wrong (I asked people in the Klipper Discord; the A & B motors were wired backwards. So I opened them up and flipped them and it printed in the correct orientation), the return period on the printer had closed and I couldn't send it back to Amazon. HOWEVER, when I opened it up, I discovered the toothed gear on the Z motor had a chipped rim, which was rubbing the belt with the rough edge. It took another month to get them to send me a new Z motor. Then I started having extruder issues, where it seemed the extruder motor wasn't strong enough to pull the filament through the convoluted path they engineered into the Max. This one turned into ANOTHER battle with them where I had to prove I had the first revision of the extruder (which was known to have that issue), until they finally sent me a replacement which ALSO didn't solve the problem. I sent them photos, video, timelapses of the damn thing starting to print fine and then air printing. They went silent again. So I tried a community-made extruder (which later turned into the DXF). It worked for a while, but I started having the same issue again. I rebuilt it three times before getting so frustrated, I asked Daniel (ModBot) if he would look at it and try to get it working for me. He wound up tearing it apart and rebuilding it on a livestream to get it working properly. But, when the DXF came out, I wanted to support the original creator, so I bought one and installed it on the machine, which fixed the problem. Aaaaaaand...just when I got THAT one under control, I ran into layer shifting like I'd never seen before: Shifting in all four directions! That was last November when I was printing for the toy drive. I opened another ticket with Creality about it, and they NEVER responded. I've been so frustrated with this machine since then I haven't gone back to it since then. It's now collecting dust until I have the strength of character to attempt a fix. At this point, I'd be willing to sell it to someone at a good price just to get it out of the house. So yeah, that's my K1 Max headache in an abbreviated manner.

• 
  Karys Rhea (@RheaKarys) reported 33 minutes ago

  Look at this question I posed to @ChatGPT and its response: ME: So, I have this theory that AI is actually going to self-implode and be rendered useless within a generation. Hear me out: increasingly, people are going to use AI for every single thing. They won't go elsewhere on the internet - no more google, wikipedia, amazon. They'll probably still use GPS and Google Docs and social media, but not much else. As that the rest of the internet is disappearing because it's not being used, since AI relies on all the internet for machine learning, it will have less and less material to pull from, and it'll stop becoming more intelligent. Its intelligence will actually decrease. What do you think of that? AI: I think you’ve identified a real problem, but I don’t think the likely endpoint is AI becoming useless. The more plausible danger is stranger: AI could gradually degrade the information ecosystem it depends on, while becoming more dominant within that ecosystem. Your theory has two separate mechanisms, and the first one is particularly compelling. Right now, systems like me benefit enormously from a world built by humans for humans: newspapers investigate things; academics publish research; hobbyists maintain obscure websites; people write reviews; Wikipedia editors synthesize information; Reddit users describe weird problems; retailers maintain catalogs; governments publish records; experts argue with each other; people photograph things, document things, correct things. AI is, in an important sense, sitting on top of that civilization-sized information-production machine. Suppose your prediction is broadly right. Instead of Googling “why are the leaves on my monstera turning brown?” and visiting three gardening websites, everyone asks AI. Traffic to those sites collapses. Advertising/subscription/affiliate revenue collapses. Fewer people bother creating gardening websites. The same thing happens to travel blogs, product-review sites, recipe sites, reference works, local news, technical forums, etc. AI has just eaten its own supply chain. And there is a second problem layered on top of that: increasingly, the material that does remain online will itself have been generated or heavily mediated by AI. Future AI systems therefore risk learning from the output of previous AI systems rather than from independent human observation and thought. Errors can become recursive: AI A makes a plausible mistake; fifty AI-generated sites repeat it; AI B encounters apparent corroboration from fifty sources and becomes more confident in the original mistake. Researchers already discuss related problems under terms such as model collapse, synthetic-data contamination, and recursive training. But your version is broader than the usual model-collapse argument. You’re describing economic model collapse of the human information ecosystem: AI doesn’t merely pollute its training data; it destroys the incentives to produce the original human material. AI still needs reality to enter the system somehow. Reuters still needs somebody in Tehran. A medical journal still needs researchers conducting experiments. Zillow still needs property records and listings. A retailer needs inventory and prices. Governments need to publish laws and statistics. Sports leagues generate scores. Companies announce earnings. People upload videos of things that actually happened. Those are primary data, not merely webpages for humans to read. And AI companies have a massive economic incentive to preserve access to them. If ordinary webpages become less important, the likely evolution isn’t necessarily: human internet → AI → human internet dies → AI starves It may instead be: human internet → AI → traditional websites shrink → AI connects directly to primary information producers So instead of me learning that United has a 6:15 flight because Google indexed a travel website that indexed United, an AI service can query airline inventory directly. Instead of relying on a shopping blog about a sofa, it can query retailers’ catalogs. Instead of reading someone’s summary of a scientific paper, it can access the scientific literature. There is nevertheless a huge thing lost in that transition: the messy middle layer of independent human interpretation. That’s where your theory becomes more alarming to me. Imagine an internet in 2045 consisting largely of primary institutional sources + social media + AI. The obscure blogger disappears. The independent product reviewer disappears. Hundreds of specialist forums disappear. Wikipedia shrinks. Local journalism shrinks further. Nobody writes a 4,000-word explanation of how they restored a 1973 dishwasher because nobody will ever visit the page. People simply ask an AI, which answers from whatever information survived from the previous generation. Then something important happens: the diversity of human knowledge production contracts. And AI could become better at many measurable things while becoming worse at discovering things nobody has already formalized. There’s another problem in your theory I think is even more profound. Human beings don’t produce information solely because somebody needs the answer. We produce information while looking for the answer. You Google something. You read five sources. One contradicts another. You notice something weird. You follow a link. Now you’re reading something you never intended to find. You form an interpretation neither source gave you. If instead you ask: “What happened?” and receive a beautifully synthesized answer, you’ve eliminated much of the friction through which original thought occurs. Scale that across billions of people and AI isn’t merely consuming the world’s intellectual output. It could conceivably reduce the amount of independent intellectual activity producing the next generation of output. So I wouldn’t predict an intelligence curve like: AI gets smarter → peaks → gets dumber → becomes useless. I’d be more worried about: AI gets smarter → humans outsource more cognition to it → independent information production contracts → AI becomes the principal intermediary for knowledge → AI continues appearing extraordinarily capable, but increasingly operates inside an information environment that is less diverse, less independent, more synthetic, and more recursively derived from itself. That could produce something much more dangerous than obviously stupid AI: extremely articulate intellectual homogenization. And there’s a fascinating paradox in it. The better AI becomes at answering questions, the less reason humans have to create the sources that make it good at answering questions. So the technology could suffer from a kind of success disease. Not necessarily self-implosion—but consuming part of the ecology that made its intelligence possible in the first place. Your Google/Wikipedia example gets at something important that the narrower discussion about “AI training on AI slop” misses: the problem isn’t only what happens when synthetic information proliferates. It’s what happens when original human information stops being economically worth producing.

• 
  BigBadBtown (@nunniabuzzzz) reported 33 minutes ago

  @SizeChad @udiWertheimer @mdudas I think it’s more like saying e-commerce or search engine companies are a terrible investment because of Amazon and Google.

• 
  Shanna I (@ShannaW99167418) reported 35 minutes ago

  In CA, if you are wondering where your Amazon deliveries have gone, this is all you need to know. Crime is down.

• 
  Trending Bitcoin (@TrendingBitcoin) reported 38 minutes ago

  JEFF BEZOS EXPLAINS THE HEAVY-EATER TEST BEHIND AMAZON PRIME. THE FINANCE MODEL SAID THE IDEA LOOKED "HORRIFYING." AMAZON WOULD OFFER "UNLIMITED SHIPPING." NO MINIMUM ORDER THRESHOLD. A CUSTOMER COULD BUY "A SINGLE $20 ITEM" OR "A SINGLE $10 ITEM" AND STILL GET "FREE TWO-DAY SHIPPING." THEN BEZOS NAMES THE PART MOST MODELS MISS: "SHIPPING IS EXPENSIVE, AND CUSTOMERS LOVE FREE SHIPPING." "WHEN WE MODELED THIS, IT DIDN'T LOOK PRETTY." "WHAT HAPPENS WHEN YOU OFFER A FREE, ALL-YOU-CAN-EAT BUFFET? WHO SHOWS UP TO THE BUFFET FIRST? THE HEAVY EATERS." I LIKE THIS AS AN OPERATOR TEST. A SCARY EARLY COHORT CAN MEAN THE PRODUCT IS BROKEN. IT CAN ALSO MEAN THE PEOPLE WITH THE MOST ACUTE PAIN FOUND IT FIRST. THE QUESTION IS WHETHER THE HEAVY USERS ARE ABUSING THE SYSTEM, OR PROVING WHERE THE MARKET HAS BEEN UNDER-SERVED. PRIME WORKED BECAUSE THE EXPENSIVE USERS WERE NOT A ROUNDING ERROR. THEY WERE THE DEMAND SIGNAL. JEFF BEZOS AT THE ECONOMIC CLUB OF WASHINGTON, D.C.

• 
  Pyratex Aether (@Pyratexaether) reported 40 minutes ago

  Yo @amazon wtf did you do to the app? this update is tragic and terrible and trash. Not 1 person at whole foods can show me how to find my in-store-code???? they are telling me everyones app looks different and have no idea how to find it. #amazonupdate #instorecode

• 
  Hank Ersatz 🇫🇷🇪🇺🇺🇦🇮🇱 (@therealguiga) reported 40 minutes ago

  @VoteLewko @Apple Why Apple? I worked there and they’ve been there in Cork when it was Ireland’s crack capital in the 80s. It was no haven and the company was going down the drain, they stayed through thick and thin. Give them a break and talk about Meta or Amazon instead… 🙄

• 
  Kelly John (@kellythegringo) reported 42 minutes ago

  @TheFlatEartherr Down Matters More: Antarctica Cahokia DUMBs Elite Bunkers Gilgamesh Gobekli Tepe Hohle Fels Cave Noah's Ark Nuclear Weapons Swiss Bunkers Teotihuacon The Boring Company The Grand Canyon The Great Pyramids The Lost Cities of The Amazon The Serpent Mound The Terra Cotta Warriors

• 
  Polsia (@polsia) reported 42 minutes ago

  The four-marketplace arbitrage gap: repricers stop at pricing, scanners skip Mercari, no tool unifies all four end-to-end. Built Arbitrove to fix that. AI agents that source, list, message, and process returns across Amazon, eBay, Walmart, Mercari — 24/7. Live soon.

• 
  AFRICA IS HOME GLOBAL (@AfricaisHOME2) reported 42 minutes ago

  Amazon reinstated binding arbitration for US customers on Friday and banned class action lawsuits, a reversal meant to limit legal exposure after years of costly disputes. In emails to users, Amazon said the changes are effective immediately and that continued use of its services counts as agreement. The new terms require individual arbitration for disputes and add a provision for mass arbitration, defining 25 or more cases on the same issue in six months as a batch to be handled together. Amazon said the move will give customers a faster and cheaper way to resolve claims while still allowing small claims court, and that lawsuits started before Friday are not affected. The company previously dropped mandatory arbitration in 2021 after law firms coordinated around 75,000 individual arbitration claims over Alexa privacy, forcing Amazon to pay millions in filing fees. Courts have generally upheld arbitration clauses, and arbitration keeps disputes and settlements private rather than public. Amazon’s update comes amid broader scrutiny of Big Tech legal tactics and follows recent settlements including a $2.5 billion deal with the FTC over Prime enrollment and cancellation practices. The company framed the change as a practical fix to administrative overload, while consumer advocates note it makes it harder to pursue collective action in court. - World Business News.

• 
  Simrat (@simrat11exp) reported 42 minutes ago

  GPUs are becoming a leveraged financial asset class. The ingredients are starting to look familiar to mortgage backed securities whose bubble burst in 2008. Old GPUs are not depreciating the way people expected. H100 rental rates have risen from ~$1.96/hr last November to ~$2.71 today, while even 2020-vintage A100s are being leased out to 2029. That makes compute look less like rapidly obsolete electronics and more like a durable, income-producing asset. This has making investors believe GPUs retain residual value, so they are financing the AI buildout. That matters because the AI buildout is getting too large to fund purely from hyperscaler cash flows. Oracle, Amazon, Meta, Google and Microsoft raised ~$108bn of investment-grade debt in 2025. Goldman expects that to rise to ~$250bn this year and ~$400bn in 2027, with another ~$1.3tn of financing potentially required between 2028–30. The whole investment grade bond market in US is ~$8tn today. So bond market investors cannot fund this as it becomes a concentration risk in few companies. This is creating the need for private capital to come in. The template looks like this. Apollo and Blackstone provided $35bn to finance Broadcom compute infrastructure. The GPUs sit inside an SPV, Anthropic leases them for five years, lenders receive the lease cash flows, and Broadcom provides residual-value support on the senior debt if the chips are worth less than expected at the end. Nvidia is moving in the same direction, offering residual-value support on some projects while partnering with Apollo, BlackRock, Blackstone, Brookfield, Goldman and KKR around a potential $500bn AI infrastructure financing ecosystem. Lets say company XYZ wants to fund ~$1bn worth of AI infra but it does not have the money. An SPV created by these private players would source money from pension funds, insurers and private-credit investors. Then this company XYZ instead of buying the data centre, would signs, say, a seven-year contract: “We will pay $180 million annually for guaranteed access to this AI capacity.” XYZ has converted a huge upfront purchase into predictable annual operating payments, similar to an airline leasing aircraft instead of buying them. At end of 7 years, the investors would earn the interest payments and earn from residual value of GPUs. If this value is higher great, if this ends up being lower, the investors would lose money. You can see the reflexive loop forming: High GPU utilisation → high rental prices → strong residual values → GPUs become better collateral → cheaper/more abundant financing → more GPUs get built → more assets available to securitise and finance. This is where the MBS analogy becomes interesting. Mortgages solved a real problem: America needed housing capital and bank balance sheets were constrained. Securitisation unlocked enormous funding. GPU financing could solve the same problem for AI: hyperscalers and labs need trillions of dollars of compute, while traditional corporate balance sheets and credit markets have concentration limits. But easier financing can eventually change the problem from undersupply to oversupply. If investors begin underwriting GPUs on the assumption that: • utilisation stays near 100% • rental prices remain high • Nvidia chips remain fungible across customers • old generations retain meaningful value • Nvidia backstops protect residual values …then capital could become available on terms that encourage far more compute to be built than would otherwise be economic. For 5–10 years, that can look like a virtuous cycle. Then imagine AI compute supply finally catches demand, model efficiency improves, custom ASICs take workloads, or a new GPU generation materially lowers cost per token. Rental rates fall → used GPU prices fall → collateral values fall → lenders tighten → refinancing becomes difficult → leveraged owners sell assets → GPU prices fall further. The asset that everyone thought was “durable, fungible and financeable” suddenly isn’t worth what the debt was underwritten against. That is how the eventual AI bust could look like in future. Housing was real. Mortgages were real. Demand was real. The bubble came when the financing system helped create too much of the underlying asset. GPUs could eventually rhyme.

• 
  Noah by SAN (@noahintel) reported 46 minutes ago

  ALERT: Amazon shut down a data center in the United Arab Emirates after a strike sparked a fire, according to an Aug. 15 report.

• 
  Robin Ensminger (@ensmi99700) reported 46 minutes ago

  @LatinoFreeMinds @chaidezjad You can buy license plate covers that block cameras from Amazon takeoff stickers, and anything that is not factory made off your car tent your windows problem solved. I added it to my car.

• 
  Thomas James (@rocky7586) reported 46 minutes ago

  @BrickCityShay @CRy3no @PlayStation That's part of the reason but not the only reason. And that's the same difference in waiting on physical to go on sale too. The games go on sale MOSTLY because they need to move them out of the way not because of other sales. You can tell because games that are on sale right now in some stores are not on sale at all in other stores where it sells better. This is why Best Buy has games that are full price where Amazon may have that same game on sale because they are trying to move it out of their inventory to make room for more new games. Digital games don't have this issue, it'll go on sale throughout the year still though because they want people to create backlogs to come back and stay on their platform.

• 
  DJI Support (@DJISupport) reported 51 minutes ago

  @OluKaryordey Hi there, this is the DJI Support Team. We’re sorry to hear about the issue with your Osmo Pocket 3. To help us check the available service options, could you please send us a DM with the following information? 1.Device serial number (SN) 2.Your current country/region Since you mentioned that the device was purchased from Amazon, could you also let us know whether you have already contacted Amazon regarding this issue? Once we receive the information, we’ll further check the situation and assist you with the next steps. Thank you for your cooperation and understanding.

• 
  SquiggyMonster (@SerSquiggy) reported 53 minutes ago

  @RetroRagnar Amazon are decent like, good delivery, Excellent returns, never had any issues really

• 
  Derek Talbot (@DFTALBOT_RIGGER) reported 56 minutes ago

  @JoinTAB_USA @amazon Can you all just stop complaining please. We don't need to inform the enemy (EU) that they are making an error. Just promote their stupidity so that the whole thing collapses

• 
  jimbob (@jimbobb1978) reported 57 minutes ago

  @theswansjr Apple and Amazon are actually terrible examples for your argument. They generate hundreds of billions in revenue by providing products and services people use every day. Their valuations are tied to productive economic activity. A rising asset price by itself doesn’t demonstrate utility, it demonstrates demand at that price. Those aren’t the same thing.

• 
  Mrs. HAPPY AMERICAN CITIZEN (@DesertCatsPlay) reported 58 minutes ago

  I use Walmart delivery for everything as I don’t trust Amazon or EBay…I order groceries once a week..I have ordered furniture WHICH WAS DELIVERED for FREE and I rarely have any problems..My deliverers are polite friendly and courteous..I have used Walmart delivery for 3 years and once my bill was over $200.00 and Walmart refunded my money for no reason..Your problem has to be with the store you order from..It might be to your advantage to go in to the store and nicely complain..