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 (46%)
• Errors (29%)
• Sign in (25%)

Live Outage Map

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

City	Problem Type	Report Time
Irving	Errors	11 hours ago
Lakeville	Website Down	19 hours ago
Zürich	Website Down	1 day ago
Cali	Errors	2 days ago
Strasbourg	Errors	2 days ago
Lakeville	Website Down	2 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 Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Valkyrie but PMA (@sren1ty_) reported 4 minutes ago

  @ToryTheForg @ashyeunderscore im not american??? do u have a problem with amazon workers because their company actively gives tech the IDF? if someones only chance at going to college is joining the military and being reserve (which is likely as only 15% of the USA military is deployed) i do not blame them

• 
  Nagendar (@Chanti37901150) reported 4 minutes ago

  @AmazonHelp I have already reported this issue, but no resolution has been provided yet. Please investigate and resolve this urgently.

• 
  Disabled Queer Atheist (@gotpartyfavorz) reported 6 minutes ago

  @FurkanGozukara @CatLoverCares Having lived next to an Amazon fulfillment center, I can confirm this. Even after they were run out of the neighborhood — it took another year in court just to shut everything down because the noise pollution drives you nuts. It comes in through the walls with a constant low hum.

• 
  P Balachandran (@PBalach34697784) reported 6 minutes ago

  @AmazonHelp @samirkumar @amazonIN I am not a FOOL to expect resolution over X platform. This is a message I have been trying to pass on to your CEO...whether or not you want to take action is up to you. I already reached out to CC on this specific problem to get a resolution.

• 
  azzy ₊˚⊹♡ (@fullmetalazzy) reported 7 minutes ago

  i started a new job and need a work bag and was looking on Amazon and that was like $30 but people have the same unused bags on marketplace for $10 bc they have overconsumption problems

• 
  TMN (@themainnetwork) reported 11 minutes ago

  @bavedikian @BestBuy @FedEx Literally the worst to deal with. They have 3rd party sellers selling phones and tech products and if there’s any issue they go talk to the 3rd party seller. Amazon does this too but they’re good about fixing a problem imo.

• 
  ︎ArYa (@ruggedFANBOY) reported 12 minutes ago

  Calls are being disconnected when am requesting for a followup on a issue running from 25th of May. What kind of service is this @AmazonHelp madharchod

• 
  mango lassi (@mangolassi69) reported 14 minutes ago

  $mu is still WAY TOO CHEAP for what is about to print and i expect it to go to 2000+ over the next 12 months. everyone knows the ai memory trade now. this is not some “have you thought about hbm?” take. the market knows nvidia needs hbm, hyperscalers need more memory, inference is getting heavier, and micron is one of the cleaner ways to play it. the bit i think is still wrong is the pricing of the earnings power. the story is not hidden. the beat might still be underpriced. micron reports q3 fy2026 on 24 june. company guide is already huge: about $33.5b revenue, roughly 81% non-gaap gross margin, and about $19.15 non-gaap eps. that is a monster guide, but i still think they probably beat it, and the bigger question is whether q4 and fy2027 commentary make the stock look too cheap even after the move. last quarter already showed the operating leverage. q2 fy2026 revenue was $23.86b. dram was 79% of sales and asps were up mid-60s quarter over quarter. nand was 21% of sales and asps were up high-70s quarter over quarter. non-gaap gross margin hit 74.9%. operating cash flow was $11.9b. adjusted free cash flow was $6.9b. now layer the read-throughs on top... microsoft is the obvious one. the q3 fy26 commentary pointed to q4 capex above $40b, roughly $190b of calendar 2026 capex, and more than $600b of revenue still to deliver. that is not “ai is a feature” spend. that is multi-year capacity buildout spend. if microsoft is still capacity constrained, that reads through to accelerators, networking, hbm, server dram and storage. meta is probably the cleanest memory read-through. its fy2026 capex guide is $125b-$145b, and management explicitly called out higher component costs, especially memory pricing. they also said they have continued to underestimate compute needs. that is basically the bull case in plain english: the buyer is spending more, still short, and memory is one of the areas where cost pressure is showing up. oracle is another. fy26 q3 rpo was reported around $553b, up 325% year over year, with oci iaas up 84% year over year. oracle has said ai training and inference cloud demand is growing faster than supply. again, that is not vague chatbot hype. that is contracted infrastructure demand. coreweave has talked about a revenue backlog around $99.4b, with 2026 capacity largely sold out. you can worry about financing and leverage there, fine, but the demand signal is still obvious: ai compute buyers are taking capacity before it exists. aws is not quiet either. q1 2026 revenue was $37.6b, up 28% year over year. amazon has cited a chip run-rate above $20b, plus large ai chip and gpu deployment plans. openai trainium commitments around 2gw and anthropic commitments up to 5gw are the kind of numbers that make the “we have enough infrastructure” argument look daft. even where the silicon is not nvidia, the system still needs hbm, dram, storage, networking and power. google is the slightly messier read-through because tpus mean not every accelerator dollar flows through nvidia, but that does not make it bearish for memory. public summaries point to cloud backlog around $462b and fy2026 capex around $180b-$190b. tpus still need memory, servers, data centres and storage. the ai stack can swap the accelerator brand and still be memory hungry. then you have openai, stargate, xai, uae and uk sovereign ai projects, and the broader frontier-lab arms race. some of those numbers are hard to cleanly underwrite from the outside, but the direction is not subtle: the largest buyers in the world are trying to secure compute capacity years ahead. this is why i think $mu still looks cheap. not because the market has missed “ai needs memory”. it has not. but because the read-throughs suggest the duration of demand and the margin leverage may still be underappreciated. the consumer and enterprise usage side supports the same point. chatgpt has been reported around 900m weekly active users, with tens of millions of paid consumer subscribers and millions of business users. google has talked about gemini at huge scale and direct api usage in billions of tokens per minute. meta has put meta ai around 1b monthly active users across its apps. microsoft has said copilot paid seats are scaling, query intensity per user is rising, and agent usage is growing. you do not need to pretend that proves a clean 10x global compute curve. most token volume, utilisation, workload mix, routing and inference economics are private. but you also do not need perfect data to see the workload mix getting heavier. coding agents call models repeatedly, read files, run tools, run tests, retry, pull context and loop. ai search is retrieval, ranking, vector search, source selection, synthesis, citations and a model answer. long context adds kv-cache pressure. reasoning burns more inference compute per answer. video, voice and multimodal are heavier again. that all maps into memory. hbm is the headline, but server dram, data-centre nand and essd pull-through matter too. the more ai shifts from demos to scaled inference and agents, the more the bottleneck becomes bandwidth, capacity, latency and concurrency, not just “more gpus”. so the print setup is simple. bear case is q3 around $33b-$34b, gross margin around 80%-81.5%, eps below about $19.70, and q4 commentary that makes it feel like peak-pricing risk. that can sell off even if micron technically meets the guide. base case is $34.4b-$36.2b revenue, gross margin around 81.5%-83.5%, eps around $19.80-$21.60, with q4 good enough to keep the pricing and mix story alive. bull case is $36.5b-$39b revenue, gross margin around 84%-86%, eps around $22-$25, with hbm, server dram, data-centre nand, essd and pricing all pulling together. but the real upside is q3 plus q4 plus fy2027 visibility. if management says hbm supply is allocated, customer demand is backed by commitments, data-centre memory tightness is extending, pricing is structurally firm, and ai demand is broadening beyond one product line, the market has to treat micron differently. a commodity company with one hot quarter gets faded. a company selling constrained capacity into microsoft, meta, oracle, aws, google, coreweave and frontier-ai capex can re-rate. that is the whole thesis. everyone knows $mu should beat. i do not think everyone has fully priced what happens if the beat comes with proof that ai customers are locking up scarce memory into fy2027. if the call sounds like “pricing was good this quarter,” fine, maybe the stock chops around. if it sounds like “customers are fighting for constrained memory supply and we have visibility into next year,” then $mu still looks too cheap. not financial advice. expectations are high, and if q4 guidance or fy2027 commentary disappoints, a q3 beat may not save the stock. but if the call confirms the read-throughs we have been discussing, hyperscaler capex, oracle backlog, coreweave sold-out capacity, meta memory pricing pressure, aws ai commitments, google capex, heavier inference, agentic workloads and customer-backed hbm demand, then $mu is still too cheap for what is about to print.

• 
  Mohammed Tawakkal Ahmed (@Tawakkalah13_10) reported 17 minutes ago

  API Gateway An api gateway is a single entry point that sits between clients and your backend services Think like this a cat opens the app and says i need puking tuna right now api gateway does checks if the cat is logged in also figures out what is needed tuna send to tuna service or other servers its like sending requests to correct services on time and combines everything into one response so api gateways acts as a single entry point for client requests, handling routing, request transformation, and cross cutting concerns like authentication and logging what is api gateway ? its a server less management tool, functions as an and intermediary between micro services and users it Improved performance ,Enhanced security and Improved scalability it also supports multiple protocols such as websocket, and restful apis and provides various security features such as authentication, authorization, and encryption provides in depth insights into API usage, latency, and error rates benefits of API Gateway Authentication and Authorisation , Encryption Rate Limiting it’s Supports growing in microservices and also handles large traffic volumes and better Monitoring in servers Circuit Breaker It’s protects the system from cascading failures. popular Gateway tools ate as folles Amazon Gateway Kongm ,NGINX , Apigee, Azure API Management and Spring Cloud So how would you secure an API Gateway??

• 
  Jennifer Gokey (@jennifer_gokey) reported 17 minutes ago

  @bavedikian @BestBuy @FedEx FedEx is terrible. Ordered a poncho from Ireland on Amazon. God only knows where FedEx delivered it. No one ever gave me a refund. Last year FedEx failed to deliver my Thanksgiving dinner. They didn't refund but the company I ordered from did.

• 
  Tom Bilyeu (@TomBilyeu) reported 18 minutes ago

  EMPLOYEE 1: THE RESEARCHER Finds the pain people will pay to fix. Most important hire. Skip it and nothing else matters. "Act like a world-class researcher who checks their work. List the top 5 urgent, painful problems faced by [your customer]. Give me proof from Reddit, Amazon reviews, and real forums."

• 
  Michelangel Yo (@MichelangelYo) reported 21 minutes ago

  @HealthRanger Yep, they've been terrible since the days of drones and requiring a smartphone activation for that. In fact many chinese companies are like this now, half the cams you buy on amazon now require you to go through their servers to even download the video, utterly ridiculous.

• 
  Roshan Kumar Jha (@jharoshan52) reported 22 minutes ago

  If you're shutting down business I'll simply order from amazon and delete flipkart from my phone just say it. Who's gonna take the delivery if you reschedule it for tomorrow are you guys dumb or what? @flipkartsupport

• 
  Joshua Michaels (@JoshuaMichaels2) reported 22 minutes ago

  @kjgillenwater The whole issue is that Walmart has blurred the lines between delivery and shipping. If they want to be Amazon Prime then they should never even have tipping as an option. Otherwise we may as well just sign up for Prime and know it’s not a thing.

• 
  Dr Selwyn Sundaram (@SelwynSundaram) reported 32 minutes ago

  @AmazonHelp @AnshulV80710569 I am not going to waste my time jumping through Amazon hoops! Read my problem! Find a solution! Fix it!

• 
  tommy.eth (@TommyMac501) reported 35 minutes ago

  @LayoffAI I received a pack of two bulbs from Amazon, but one bulb was missing. I used the support AI to explain the situation. I said I don’t want to go through the hassle of waiting for a new bulb, and packaging the box and driving to a shipper to send the single bulb box back, can they just credit me? The AI was great, it understood the issue, it was empathetic about the silliness of returning a light bulb and rearing to help - but it can’t “actually help”. So it passed me on to the Customer Service bot, which promptly sent me a new bulb and explained how I need to package the box with one bulb and drive it to a shipper. Ugh.

• 
  Chukwuebuka Okoroafor (@DannyFreshii) reported 36 minutes ago

  Most “e-commerce gurus” don’t have a sourcing problem. They have a execution problem. I’ve worked with Amazon sellers, Shopify brands, and creators who all had the same issue: Great ideas… but no real access to China manufacturing. One YouTuber wanted custom T-shirts for his audience. We turned that idea into actual product drops. A Saudi businessman needed luxury hotel equipment for a project — sourced and delivered. This is what people don’t see behind “building a brand”: Can you actually turn your idea into something a factory in China will produce correctly, at the right cost, and on time? That’s the gap I operate in. If you’re building something real and need sourcing from China, I can help you execute it properly.

• 
  P Balachandran (@PBalach34697784) reported 36 minutes ago

  @AmazonHelp @samirkumar @amazonIN What can I say ! It seems you definitely have a THIEF in your logistics department. In 3 weeks this is the 3rd recurrence of the same issue I have highlighted. Ordered steam Iron, received a defective product, raised replacement, replacement product was marked as delivered👏👏

• 
  ROI (@ormkaa) reported 39 minutes ago

  Section 7: The Challenges — Not Everything Is Simple Being honest — NVIDIA faces real problems too. Problem 1: ChinaThe US banned NVIDIA from selling its most advanced chips to China. China is one of the largest potential markets. NVIDIA counts zero China revenue in its 2027 projections. Problem 2: Hyperscaler Custom ChipsGoogle has its own TPU chips. Amazon has Trainium and Inferentia. Microsoft and Meta are building custom chips. These companies — NVIDIA's biggest customers — are simultaneously trying to reduce their dependence on it. Problem 3: AMD and New PlayersAMD is actively developing the MI300X and new chips. Qualcomm and Intel aren't standing still either. But: None of these challenges currently threatens NVIDIA's dominance. CUDA is too deeply embedded. Hyperscaler custom chips still lag in performance. AMD hasn't yet reached the necessary scale.

• 
  OaklandMofo (@OaklandMofo) reported 42 minutes ago

  I no longer user Amazon for high dollar purchases. Best Buy pick up at store only. Amazon has sent me open box items, broken items, clearly returned items(in the green bag), and their drivers never ring the doorbell even when it is on my account and I am home.

• 
  DOGE Employee 121 (@DOGEmployee_121) reported 47 minutes ago

  @fandompulse They fell victim to the Woke Producer Era An NGO stole $30 billion in 2024 from Department of Energy Amazon Studios was given $8 billion to hire only DEI and Female producers/Directors to then Destroy and Wokify all IPs they purchased. LOTR was targeted for this to turn men into gay losers. It did not work, no one watched Rings of Power past a first glance over. While a Fan of LOTR has watched the movies 10 times over since the release of Rings of Power, no one will ever watch those shows more then 1 time, if even once at all. That is the reality of anything DEI or WOKE Produces, it is a attack on any sane or normal person, everything woke produces feels unnatural and wrong, and broken. So no one ever watches it past its initial curiosity.

• 
  M. (@mattyniven) reported 49 minutes ago

  @JamesAbbott2013 Remember COP30 James; the world conference where they built a 4-lane highway through the Amazon rainforest to facilitate access for the participants? What was decided at COP30? What urgent and immediate measures were put in place to combat the ‘deadly threat posed by CO2?’ Nothing! Nada! Zilch! It was all just a load of hot air (ironically). But, don’t listen to me James, just read between the lines of the COP30 narrative. Yes, CO2 is a GHG. Is it a problem? Well, apparently not. There is no climate crisis. Only climate change.

• 
  Brutopian 🇫🇮🇺🇦🇵🇸🇪🇺 (@Brutopian1) reported 50 minutes ago

  @HillbllyAnalyst @fandompulse The problem is, they can't. Amazon only has rights to the short summary of the 1st and 2nd ages in LOTR appendices. If they try to make the show follow the Silmarillion, the Tolkien estate can sue. Perversely, they're contractually obligated to not be faithful.

• 
  CryptoGoos (@cryptogoos) reported 50 minutes ago

  WARNING: Nvidia $NVDA CEO Jensen Huang just told retail to buy SpaceX, OpenAI, or Anthropic at IPO. He said it could be like buying Amazon, Google, or Meta in the early days. But here’s the problem. Amazon went public at a valuation of roughly $438 million. Google at around $23 billion. Meta at around $104 billion. SpaceX is reportedly targeting a valuation approaching $2 trillion. That’s not the beginning of the story. That’s after years of explosive growth, multiple funding rounds, and massive value creation in private markets. The biggest gains are usually made when a company is still unknown, unproven, and cheap. By the time a company reaches a multi-trillion-dollar valuation, investors aren’t paying for potential anymore. They’re paying for expectations. The question isn’t whether SpaceX is a great company. It’s whether the next 10x is still ahead, or whether most of it has already happened. Follow and turn on notifications to stay up to date.

• 
  Gina (@sweettart024) reported 52 minutes ago

  @bavedikian @BestBuy @FedEx Ran into same problem - a computer ordered from Amazon, tracked to a UPS warehouse in TX,en route to me, just "disappeared". Amazon said to deal with Dell & UPS to solve it. I tried with no luck & finally convinced them THEY made the deal with UPS to ship the computer. Not me.

• 
  Fabula Rasa 02 (@fabularasa02) reported 54 minutes ago

  @bavedikian @BestBuy @FedEx I always make make major purchases on AMEX. Those folks will go hunt down a chargeback like pack of starving hound dogs. When someone got my AMEX Information to make fraudulent Amazon purchases THEY got me my lonely back after Amazon gave me the run-around.

• 
  Mohan Agrawal (@MohanAgraw7457) reported 58 minutes ago

  @AmazonHelp That email id linked to account is not working If possible contact me on the mobile no. Provided

• 
  DaBull Standard (@DaBullStandard) reported 59 minutes ago

  @deeplyoffensive @wil_da_beast630 step-father, maybe Biological father rode a unicycle in a circus, abandoned them, and didn't even realize he founded Amazon until 2012 when a journalist tracked him down

• 
  Nikolai Rostov (@austrosillyism) reported 60 minutes ago

  @MindReflects >big companies like Amazon have solved the allocation problem No they haven’t. They make mistakes in the quantity/supply of stock all the time and they adjust their capital structure by changing the quantity of certain capital and labour when needed.

• 
  Mehwish kiran (@mehwishkiran07) reported 1 hour ago

  The sixth fix: Notifications she didn't need on her wrist. He scrolled through her Notifications settings. She was getting alerts on her watch from: 1. Instagram (every like, comment, story view) 2. Amazon (shipping updates) 3. Email (every single message) 4. News apps (breaking news pushes) 5. Weather (daily morning summaries) 23 other apps she barely used Every notification costs battery the watch wakes the screen, vibrates the Taptic Engine, runs the relevant app briefly, and logs the event. He recommended a brutal but effective cleanup: Watch app on iPhone → My Watch → Notifications → go through each app individually → Off Keep on: phone calls, text messages, calendar, one or two genuinely important apps. Turn off: everything else. Her watch went from buzzing 80+ times a day to maybe 12. Battery life jumped. So did her sanity.