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 (48%)
• Errors (27%)
• Sign in (25%)

Live Outage Map

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

City	Problem Type	Report Time
Kefar Yona	Errors	5 hours ago
Monterrey	Website Down	6 hours ago
Monroe	Website Down	17 hours ago
San Jose	Website Down	19 hours ago
Santa Cruz	Sign in	23 hours ago
Paris	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:

• 
  Umang Khosla (@ukhosla) reported 25 seconds ago

  @amazonIN @AmazonHelp - I have written a mail yesterday to customer service asking about my refund & got no answer. When I raise a question here, I am told to write to customer service via mail & even then I get no answer or solution to the problem. Pathetic service from Amazon

• 
  Babatunde E. Daramola (@RiteChoice23) reported 8 minutes ago

  Within minutes, 85 percent of Fastly's global network started returning errors. Amazon, Reddit, the New York Times, the UK government's own website, all down at once, not because of an attack, but because of how tightly coupled the edge network was.

• 
  miru (@subaruyourichi) reported 10 minutes ago

  @help_delhivery @DelhiveryLtd This is getting absurd. Yet again my Amazon parcel AWB 13372288728993 is marked "delivery attempted — incomplete address" with NO call, NO message from your agent. Same address has received 3 other parcels without issue. Peak carelessness from your agent!

• 
  Rinku Malik (@rinkumalik297) reported 10 minutes ago

  @AmazonFresh I am trying to order by Amazon fresh delivery pin code 143001 Amritsar Punjab receive error message all location, is We're temporarily unable to deliver orders please help and provide reason for this error

• 
  LearnInvest (@LearnInvest2026) reported 11 minutes ago

  ⚡ Customers are funding part of the GPUs. Why does Oracle still face credit pressure? 【Daily · 2026-07-22】 Oracle now presents a striking paradox. The prepaid and customer-supplied hardware portions of its large AI contracts total $75B. That can include customers prepaying Oracle to buy GPUs or supplying GPUs themselves. Yet on July 9, S&P Global Ratings cut Oracle's long-term issuer credit rating from BBB to BBB-, with a stable outlook. The rating remains investment grade, but it now sits at the lowest rung. Meanwhile, a 902MW AI campus in Wisconsin has exposed Oracle to a conditional financial-security estimate of more than $7B. S&P's action addressed Oracle's overall rising business risk and weaker cash flow. It was not a ruling on Wisconsin or the $75B hardware structure. The value of placing these clues together is different: They show that customer funding can remove one layer of capital risk without removing the others. The short answer is that an AI data center is not one GPU purchase. It is a stack of long-term contracts across four balance sheets. 〔Chart 1: One AI campus, four balance sheets〕 ━━━━━━━━━━━━ 1. The AI customer It may sign a multiyear compute contract, prepay for hardware or supply GPUs. Its risk is long-term payment and demand. 2. The cloud operator It integrates GPUs, networking, software and facilities into usable compute. Its risks include execution and utilization. At Lighthouse, Oracle subsidiary OACS is the entity taking electric service. 3. The data-center developer It prepares land, buildings, cooling and long-term facility capacity. Vantage describes Lighthouse as a 902MW development with roughly $15B of investment. 4. The utility It may arrange generation, grid and dedicated interconnection assets before the load arrives, then recover those costs over years. At very large, dedicated and power-first AI campuses, technology investment increasingly needs to be understood through the financing logic of infrastructure. Contracts can move risk. They cannot make it disappear. ━━━━━━━━━━━━ The $75B and >$7B solve different problems Oracle's $75B addresses who funds part of the hardware upfront. It is not all cash received by Oracle, and it is not specific to Lighthouse. Wisconsin's >$7B estimate addresses a different question: If a data center ultimately uses less power than promised, who pays for the infrastructure built for it? Oracle says Lighthouse could need more than $7B of financial security if the current tariff rule applies and OACS receives no exemption. The amount would cover unrecovered asset costs for subscribed generation and dedicated distribution facilities, plus certain tariff charges. Bank letters of credit would be the more likely form of support. This is not a bill, fine or cash already paid. If Oracle performs, it does not pay $7B as a bill. But it still pays for the support and uses credit capacity. So the two numbers cannot be netted: ▸ $75B shifts part of the hardware funding burden ▸ >$7B addresses exit risk tied to dedicated utility assets ▸ Oracle sits between the two, receiving customer support while potentially committing its own or bank credit to remaining risks 〔Chart 2: Hardware funding and utility exit protection are different layers〕 ━━━━━━━━━━━━ The real risk is not high spending alone. It is a mismatch between revenue timing and asset recovery. Amazon says AWS must lay out cash for land, power, buildings, chips, servers and networking gear 6 to 24 months before customer billing begins. But those assets do not expire together: ▸ Chips, servers and networking: about 5 to 6 years ▸ Oracle uncommenced data-center leases: 15 to 19 years ▸ Certain WEC wind and solar resources: 20 years under its May 2026 presentation, then still subject to the Wisconsin PSC's final written order ▸ Data centers: more than 30 years Microsoft also had $196.6B of additional leases, primarily for data centers, that had not yet commenced at the end of March. Their terms range from one to 21 years. Oracle's fiscal 2026 10-K disclosed $260B of additional data-center lease commitments that had not yet commenced. Neither number is current borrowing or cash due today. But both show why current capex and operating cash flow miss long-term facility commitments that have already been signed. A mismatch appears when: ▸ Rent and power obligations begin before the workload is ready ▸ GPUs need another refresh while facility and power costs still have years left to recover ▸ The original customer exits while the site's scale, location or design slows alternative use The reverse can also be true. If contracts, prepayments and minimum payments cover the main recovery period, and other creditworthy tenants can use the site, the mismatch may remain a construction-phase timing issue rather than become a credit problem. 〔Chart 3: Billing, hardware, facilities and power have different recovery periods〕 ━━━━━━━━━━━━ Why does Oracle's pressure reach the credit market faster? This is not a story in which Oracle carries risk and the rest of Big Tech does not. Microsoft's uncommenced leases and Amazon's cash-cycle explanation show that even cash-rich hyperscalers lock in years of facilities and customer commitments. The difference is how much operating cash flow can absorb the gap, and how much risk must be placed with bondholders, shareholders, landlords, customers and banks. Oracle generated about $32.0B of operating cash flow in fiscal 2026, spent about $55.7B on capex and reported negative $23.7B of free cash flow. It has also used debt and equity financing to support the expansion. This does not mean Oracle cannot raise capital. The $75B customer-hardware structure is a real buffer. But if power is delayed, customer terms change or financing costs rise, and customer payments or asset reuse cannot absorb the gap, pressure is more likely to reach the next campus. The BBB- downgrade highlights a crucial distinction: Having a large order book and safely carrying the balance sheet until those orders become cash are two different achievements. Investors should ask three questions: ▸ Who pays upfront? ▸ Do revenue and cost timelines align? ▸ If demand changes, do cancellation payments, guarantees or reuse arrangements cover unrecovered costs? The next AI infrastructure winner may not be the company announcing the most gigawatts. It may be the one that converts announced capacity into financeable, monetizable megawatts with clear exit compensation or reuse arrangements. 〔Chart 4: Announced MW is not necessarily bankable MW〕 Sources: Oracle, Microsoft, Amazon and WEC Energy Group official materials; Wisconsin PSC and Oracle regulatory affidavit; Vantage Data Centers; S&P Global Ratings. Charts: LearnInvest.

• 
  Pietro Mappers (of Success Mappers) (@SUCCESSMAPPERS) reported 11 minutes ago

  Amazon SEO isn't just stuffing keywords; it’s diagnosing your funnel. No impressions? You have an indexing or bid problem. Impressions but no clicks? You have a Main Image or Price problem. Fix the exact bottleneck before changing random variables. #AmazonSEO

• 
  Tom (@tdharris20) reported 13 minutes ago

  Bezos is a terrible person. The way he runs Amazon nobody should want him anywhere near LFC but as usual, all the idiots on here are just going "omg he's so rich!" 🤤🤤

• 
  Daniel Kinley (@FearlessVoicee) reported 14 minutes ago

  @AmazonHelp I’m gonna reply really quick. I did replace the order by updating my drivers license information. I think I fixed the problem, but I will send a message and have you take a look just to make sure. hold on one second

• 
  Bobby (@Bobby1479644) reported 14 minutes ago

  @nyy10453 @pittsburghfan68 @thehdroom Amazon lost mine too! Was lucky enough to get in on second round. But that’s been first real issue I’ve had with amazon in over a decade!!

• 
  🇺🇸 Chris Trapp ₿ 🟠🌎 (@cjtrapp) reported 16 minutes ago

  @amazon Seeing a new trend with orders of anything sold as a "multi pack" showing up as only one item. Happened 3 times in the past 2 months after years of zero issues with this. Anyone else experiencing this?

• 
  Amazon Help (@AmazonHelp) reported 17 minutes ago

  @rinkumalik297 We are sorry to know that you are facing an issue with us on ordering products from Amazon Fresh at your location. Kindly check periodically and place an order once the service restores. -Siddarda

• 
  Arpan Jain (@jainarpan13) reported 20 minutes ago

  Between all this got a call from Mohit again today because I gave him 1star rating in everything. Now he is insisting me to remove that and then he will be able to resolve the issue. I am surprised by this behaviour of yours @AmazonHelp how can you insist customer to give goodr

• 
  randomtexasdude (@curraheevet) reported 20 minutes ago

  @PlumbNick Florida workers are getting ****** over so Amazon can shut down and FULLY AUTOMATE two warehouses for the next two years. (FIFY) Once fully automated, they won't hire the employees back.

• 
  BlackwellBoy (@Blackwellboy) reported 21 minutes ago

  for all of those following along this bs scenario I had to endure. I finally have an outcome. amazon: we need the police report police: we legally cannot give him that report amazon ecr team : we need the police report my bank: here's your money back case closed i guess. not by amazon though. amazon is still waiting on the report. only issue is now, i dont trust to order through the same place again and the prices have went up by several thousand dollars where i live. why dont they have a service that records high value items as they are packaged to avoid issues like this? my guess is it by not doing so, it saves them a lot of money by being able to drop the items back in to new stock that were returns. @NVIDIAAI best price you can ship me a new founders? haha.

• 
  DeusExMachina (@ExMachina498) reported 21 minutes ago

  @rupertsmom1 @ImMeme0 @amazon How did you find a number for Amazon? I’ve only been able to chat with bots whenever I had an issue

• 
  Darin Pace (@Ogrelord73) reported 22 minutes ago

  @AmazonHelp Waste of time! Amazon is 100% computer run. No humans can make decisions or tell me anything other than ask me to call and when you call they say there is no way for them to see what the issue is and there is no way to get a person to resolve the issues. WHAT A JOKE!

• 
  tea stoner (she/her) 📚 57/50 (@bookbeduion) reported 23 minutes ago

  @scliest Yep I've mostly never had problems with Amazon books except once the spine was broken lol but I didn't mind

• 
  mi (@overcatbe) reported 26 minutes ago

  @Borussenfan527 My Greek TV is cckblocking me to amazon prime US, which means I cannot login with my amazon DE account there and I'm locked out of all the movies/series I've bought.

• 
  Tom | The Investing Tiger (@SavingsCaptain) reported 27 minutes ago

  A person asked, what is an example of a stock you sold when your thesis breaksdown $UBER is a great example I originally bought post-IPO and had a cost basis in the $30's and sold at $74 about two years ago I was attracted to their story of disruptive scale, a couple of years ago I felt the story was changing and they would no longer be the disruptor, but responding to it on compressed margins. I may be wrong, but wasn't willing to risk my 2x to stick around and find out. Here is what I see for their key segments Rideshare: Waymo, Tesla, Cruise, Zoox, and others are each rolling out their own robotaxi services. They won't need Uber's platform. They will undercut on price, own the customer relationship, and serve the most profitable dense urban corridors. When scale hits Uber’s network moat and take rate and margins face pressure. I know few businesses that thrive on margin pressure. This will be a slow spiral. There is no friction, next week I'll be in a Cybercab city and will tap my Robotaxi app just as easily as Uber. Delivery: In food and parcel delivery, the battlefield is different but just as crowded. Amazon (with its logistics network and Zoox), DoorDash, and emerging AV startups are all testing autonomous delivery—robots, drones, and driverless cars. These rivals can bundle AV delivery with their own ecosystems (Prime, subscriptions, loyalty programs) and use AVs to lower costs on predictable routes. Uber Eats then has to compete on speed, price, and reliability while still managing a human-heavy fleet, which compresses margins. A longer term risk, high frequency last mile is the hardest to solve for Freight: Freight is where the competition is more industrial but still intense. Amazon, dedicated trucking AV firms, and large OEM-backed initiatives are all pushing autonomous trucking. Their advantage: control over lanes, depots, and shipper relationships. If AVs prove reliable on key corridors, Uber Freight risks losing high-value lanes to players who can offer cheaper, more predictable capacity without relying on Uber’s broker network. Visit/listen to Freightwaves, this is real Again, I may be wrong on all three and erosion will be gradual. This is just an example of my thesis 10 years ago breaking down in real time, and I took my money and moved on.

• 
  👑 (@quionicaniasjia) reported 27 minutes ago

  I was up at 3 for my Amazon block, finished by 6:30, came home and laid down for about an hr and back up making breakfast for me and Pay before her therapy 🫠

• 
  Bibhas Pandey (@bibhaspandey) reported 29 minutes ago

  @AmazonHelp I put in all the details on the given link. Don't think you guys are willing to resolve this or even investigate the issue. That much is Your cust care is asking of screen recording. As a customer I feel like I've been stone walled. That's the worst part of online shopping.

• 
  Abhishek B (@itsmeabhi69) reported 33 minutes ago

  @AmazonHelp Other Accounts are also not logging in, what's the issue in my device check and Resolve

• 
  tylerswriting (@thewrittensoul) reported 35 minutes ago

  @RhysMerilot I don't trust corporations to do the right thing when the chips are down. Imagine them keeping up with the servers in a recession, or a depression, or in fifty years... Not owning the product is the big one for me. I've had digital 'assets' on amazon disappear (always sunny episodes for starters), and who knows what extreme censorship could take place in the future either altering or removing your titles? I don't think digital is bad, but I think ONLY digital definitely is for people like me who share my values. I also like video game stores like gamestop and browsing titles in a real world environment. But to each their own. By all means, enjoy your digital.

• 
  Fozzy Trooper (@TrooperFozzy) reported 36 minutes ago

  @AmazonHelp Have a look on this post and see how many times I have posted. Clearly nothing gets done and Amazon couldn’t give a ****. Otherwise it would not keep happening. Also the post I’ve made about issues on this thread are not even the only ones.

• 
  Aswanth P S (@GadgetLens2025) reported 36 minutes ago

  @Siva_Bharani Yes recently I purchased a anti dandruff shampoo from Amazon but it delivered only having amazon package no product package and seal on the bottle . So I requested replacement on the app but it's not returnable item and it got cancelled I just tweeted about this issue

• 
  Nav Toor (@heynavtoor) reported 40 minutes ago

  You bought a Kindle. You bought books on it. You paid real money for them. You do not own them. In 2009, Amazon reached into every Kindle in America and deleted the book 1984 by George Orwell. People had paid for it. Amazon took it back. The New York Times covered it. In February 2025, Amazon removed the button that let you download the books you already paid for onto your own computer. Before this button existed, you could save your library. Back it up. Move it to another device. Now you cannot. Your Kindle books live on Amazon's cloud. If Amazon deletes them, they are gone. If Amazon changes them, you get the new version. If your Amazon account gets closed, you lose everything you paid for. You do not own a library. You rent it. There is a free app that fixes this. It is called KOReader. KOReader is not made by Amazon. It is made by volunteers around the world. It runs on your Kindle. On your Kobo. On your PocketBook. On any Android phone. On reMarkable. You install it on the device you already own. Then your Kindle stops being an Amazon store. It becomes a book reader you own. Here is what changes. You can put any book on it. From Project Gutenberg, which has 70,000+ free classics. From Standard Ebooks, which has beautifully typeset public domain books. From your library through OverDrive. From Anna's Archive. From a friend who emailed you a PDF. You can read PDFs. EPUBs. DOCs. Comic books. Textbooks. Anything. You can tap any word to look it up in Wikipedia. Or in a dictionary you download for free. Or translate it with Google Translate. You can save articles from the web with Wallabag. Read them on your Kindle later, offline, no ads. You can turn a page faster than the Amazon software does it. The volunteers who built KOReader measured this. On some older Kindles, page turns are twice as fast. None of this needs Amazon. None of this sends data to Amazon. None of this can be taken back by Amazon. Right now: 27,935 stars on GitHub. 1,798 forks. AGPL-3.0 license, which means it is free forever and no one can lock it down. Version 2026.03 shipped in March. Pushed to yesterday. 13 years of development. Amazon Kindle Unlimited: $12 a month. That is $144 a year for a rented library. Kobo Plus: $10 a month. Apple Books DRM: locked to your Apple ID. KOReader: $0. Runs on the device you already bought. Reads books you actually own. Your Kindle was designed to make you buy from one store forever. KOReader turns it into a book reader that reads any book. Amazon deleted 1984 from Kindles in 2009. Nobody can delete a book from KOReader. That is the point. (Link in the comments)

• 
  Shanu Mathew (@ShanuMathew93) reported 41 minutes ago

  Some of the better insights from the MAD Podcast with Sachin Katti - Head of Industrial Compute at OpenAI. -$50B of OpenAI compute spend this year is “directionally about right.” Katti says demand still far exceeds supply and anything brought online is consumed immediately. -Their evidence against the overbuild thesis: OpenAI tripled compute and tripled revenue. The bigger risk in his view is underbuilding and later discovering they should have built faster. -AI doing AI research compounds the demand problem. Experiment volume was historically capped by scarce human researchers. AI removes that cap, which increases research compute demand. -Inference may already be the majority of compute. Synthetic data, post-training and test-time compute are all inference, even when grouped under “training.” -The key metric is tokens per watt. OpenAI’s Jalapeño chip is designed around known model workloads, allowing hardware-model co-design. Design to tape-out took ~9 months with a former Google TPU team, Broadcom and AI-assisted chip design. -AI is already helping design AI chips. Katti thinks the point where AI designs the systems needed to train the next generation, including the chips, “is not that far.” -Cooling is now a compute constraint. Chips, interconnects, cables and even parts of the electrical system require cooling. Better heat transfer supports higher power, memory bandwidth and floating-point operations. -Power procurement is increasingly full-stack. OpenAI says it funds incremental generation, transmission, transformers and substations where grid capacity is unavailable. Behind-the-meter power today is primarily gas; the constraint is turbine supply. Nuclear “can’t come soon enough.” -The most important structural point: OpenAI is generally the tenant/offtaker, not the infrastructure owner. Microsoft, Amazon, Google, Oracle and other partners finance and build the assets, while OpenAI commits to consume the compute. >That pushes the financing off OpenAI’s balance sheet, but the economic obligation remains in long-term offtake contracts. The real underwriting questions are the duration and strength of the guarantee, hardware residual value and who absorbs technology risk. -Stargate is now the umbrella compute strategy, not one Oracle/SoftBank joint venture. The portfolio includes hyperscalers, neoclouds, chip partners, partner-built campuses and potentially OpenAI-designed facilities. -At ~100,000-GPU scale, networking reliability becomes as important as chip supply. Their new multipath protocol is designed to mask constant link and switch failures from the training workload. -The bottlenecks are increasingly physical: permitting, turbines, transformers, substations, electricians, plumbers and construction capacity.

• 
  Tyler (@bluegrass99ky) reported 53 minutes ago

  @ABUnofficial100 @RMBee As a HUGE fan of the first film and the novel, I would gladly sit through all 3 hours and 6 minutes of this movie. I wonder if Amazon MGM is gonna ask Micheal Mann to trim it down by a few minutes. 👍🏻

• 
  ratnajeetsarkar@gmail.com (@RatnajeetSarkar) reported 54 minutes ago

  @AmazonHelp I am going to raise this issue of negligence to sev 1. A customer is reporting a fraud. Instead of looking into this thoroughly, you guys are playing the “reach out to this and that” game. Good job @amazon for not looking into this serious flaw in your system.

• 
  ayesha (@ayesha_fatiima) reported 58 minutes ago

  Just heard someone got laid off at Amazon. 8+ major projects 600+ LeetCode problems 1800+ rating 3× hackathon winner Open-source contributions Still got laid off. Is this really the end of Computer Science? 😭