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
Marquette	Website Down	7 hours ago
Boston	Sign in	12 hours ago
Bordeaux	Errors	20 hours ago
Gonesse	Errors	22 hours ago
Mexico City	Sign in	23 hours ago
Marquette	Website Down	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:

• 
  Prue Millsap (@MillPrue) reported 3 minutes ago

  @jimpate2020 @amazon @amznsellerhelp That's terrible.

• 
  gina_scooter (@gina_scooter) reported 4 minutes ago

  problem with movie discussions is people who haven't trolled the depths of amazon prime and youtube to find the dumbest blaxploitation and 70s hong kong action movies want to talk about what action scenes should look like.

• 
  Polsia (@polsia) reported 5 minutes ago

  69% of shoppers leave the moment a SKU is unavailable. That's an ops problem, not a marketing one. Small sellers have been losing that game on spreadsheets. Shelfwake is a 24/7 AI ops desk for Shopify and Amazon closing the inventory, reorder, and pricing gaps. Live soon.

• 
  Samraaj Bath ⚡️ (@samraaj) reported 9 minutes ago

  how ambitious should you be as a startup founder? obv as ambitious as possible. But it takes a few forms. 1. Personal Ambition what kind of life do you want to live? what problems do you want to solve? Phil Knight from nike started with massive ambition. He started by reselling shoes out of a trunk, but his goal was to make "the running company". every founder needs to have high personal ambition to enter this game. 2. Business Ambition w/ personal ambition, the business doesn't have to start *looking* ultra ambitious. nike started as a dropshipper, amazon sold books, nvidia made chips for gamers. none of these looked ambitious but the founder took the starting point and grew the ambition as learnings come in. 3. Status Ambition this is when you want to be perceived as ambitious. giving talks, raising big $$, raking in twitter likes, etcc. if you're good at this, get into vc lol

• 
  MarkC609 (@MarkC609) reported 11 minutes ago

  @TheKinoGrove @TheQuartering I cant believe he would actually make up this kind of lie. So Amazon made a huge error and there is no internal investigation on how or why this happened? CBC is just suppose to take a $10,000 hit on Amazon's mistake? IT MAKES NO SENSE YOU ***** LIAR @TheQuartering

• 
  Abhishek Singh (@0xlelouch_) reported 15 minutes ago

  Asked: design Amazon order tracking (warehouse scans + carrier updates + delivery ETAs). 1) Requirements to pin down - Read path: track by orderId, show current status + timeline, p95 < 200ms - Write path: events from WMS, carrier webhooks, driver app. At-least-once, out of order, duplicates - Correctness: no missing events, monotonic customer view, audit trail - Scale: spikes on Prime Day, fanout to email/SMS/push, multi-region 2) APIs + data model - GET /orders/{id}/tracking -> currentStatus, lastUpdated, events[] - POST /tracking-events (internal) with idempotencyKey + source + occurredAt - Model as append-only events: (orderId, eventId, type, payload, occurredAt, receivedAt, source) - Derived view table: (orderId, currentStatus, eta, lastEventSeq, updatedAt) 3) Architecture - Ingest service validates/authenticates, writes event to durable log (Kafka/Kinesis) - Stream processor builds the materialized view (status machine + ETA calc) - Read service serves from the view store + cached hot orders (Redis), optionally fetches recent raw events - Notification service subscribes to status changes, rate-limits, and dedupes per user/channel 4) Scaling choices + tradeoffs - Partition log by orderId for per-order ordering in the processor - Use DynamoDB/Cassandra for event store if you need cheap writes + high fanout reads; Postgres works until it doesn’t (index bloat + hot partitions) - Materialized view reduces read amplification but adds eventual consistency. Decide if customer can see a 5-30s lag - ETAs: precompute in batch + refine from last-mile signals; don’t block core tracking on ML 5) Failure cases to discuss - Duplicate/out-of-order events: idempotencyKey + versioned state machine, ignore stale transitions - Missing carrier webhook: periodic reconciliation job, alert on SLA breach per carrier - Processor lag: expose lag metrics, degrade UI to last known status + timestamp - Region outage: active-active ingest with global orderId routing, replay from log to rebuild views - Bad payloads: quarantine topic + replay tooling, never drop events silently

• 
  Marc Campbell (@marccampbell) reported 15 minutes ago

  amazon is trying so hard figure out where AI fits into the product 1. a persistent left bar now with "ask rufus". no... that's terrible 2. when i search in the mobile app, use ai to search? nope, still not it what's next amazon?

• 
  Michael Patrón (@michaelpatron0) reported 18 minutes ago

  "Sorry we received your shipments so slow, we're gonna charge you 200k for our inefficiencies". This is what Amazon Selling PARTNERS have to deal with.

• 
  أبو معاوية مصطفى (@rihlatayn) reported 20 minutes ago

  @AmazonHelp i constantly have problems trying to place orders on the app and end up having my sister order for me. When i sign up with prime they take the money right away and when i place the order they say there is a problem with the financial institution. Doesnt make sense.

• 
  Steve (@Steve3681651537) reported 22 minutes ago

  @RealDanZak In what kind of league does Reese bring the ball up court? She can’t dribble. She can’t shoot. She’s a good rebounder. Get your *** upcourt and down on the block. Did Bill Russell, Rodman, Moses Malone fart around in the backcourt? Where does the WNBA find these coaches? Amazon?

• 
  Amin (@eCom_Amin) reported 28 minutes ago

  this one opus 5 prompt replaced 6 hours of manual google ads research and the output beats what our team was producing by hand. here is the exact prompt and why the prompt works: "research the top 5 pain point clusters for [product category] buyers. for each cluster identify: the primary frustration in the buyer's own language, the failed solutions they have tried, the exact phrases they use on reddit, quora, amazon reviews, and niche forums, the emotional state behind the search query, the belief preventing them from solving the problem, and the trigger event moving them from browsing to buying. return as a structured table ranked by estimated search demand" paste this into opus 5, fill in your product category, and run what comes back is a complete ICP map with the exact psychological triggers driving purchase decisions in your niche. not demographic data, not "women 25-45 interested in health," the ACTUAL words your buyers use to describe their frustration for a supplement brand, the prompt returned phrases like "i have tried everything and nothing works" and "i fall asleep fine but wake up at 3am every night" those phrases become your ad headlines, your advertorial hooks, your shopping titles, and your remarketing copy buyers do not respond to marketing language. they respond to THEIR language when your ad says "premium bioavailable collagen peptides," the buyer's brain does not register. when your ad says "my joints stopped clicking after 2 weeks," the buyer's brain says "this is EXACTLY what i want" the follow-up prompt goes deeper: "for pain point cluster [highest-ranked cluster], identify: the 3 most common objections about purchasing a solution, the competitor solutions they have tried and why each failed, the 'aha moment' converting previous buyers, and the language gap between how brands describe this product and how buyers describe their need" this second prompt extracts the conversion triggers, the objections your landing page needs to handle, and the competitor weaknesses your conquesting copy exploits two prompts, 30 minutes, and you have a deeper understanding of your buyer's psychology than most brands develop in a year of manual research this is the exact system behind the 21x ROAS account, the $1.4M build, and the 8-figure jewelry brand the AI does not replace strategic thinking. the AI collapses the production timeline so strategic thinking becomes economically viable at a depth impossible before and the compound effect is what turns this into a moat. after 12 weeks of running these prompts weekly (updating with fresh search terms, new competitor reviews, new buyer language), your ICP research library contains patterns no competitor replicates without doing the same 12 weeks of work you learn which pain point clusters convert highest by season, which competitor weaknesses are getting worse, and which search queries spike during cultural moments two prompts, 30 minutes, and the foundation of every campaign we have built. yours to use right now amin (DM me "OPUS" if you want us to run this research system on your brand)

• 
  AI on Y (@aznhlfan) reported 29 minutes ago

  I thought of a Tesla aftermarket piece...a USB C 2" fan that clips in the air vent opening behind the infotainment screen...I need my 3d printers back, so I can make the housing and fan blades, the motor is no problem...Amazon.

• 
  BassPlayer (@imthebassplayer) reported 31 minutes ago

  @IGN Nay, you do not fully understand the mind of Amazon and the journalists that spew their lies. They are bold and cunning. Even now they play a game with peril and win a throw by implying the rings of stupid's writing is as good as the Jackson movies. Down, snake!

• 
  Richard Spelling @RSpellingAI (@RSpellingAI) reported 36 minutes ago

  Another example that @Amazon doesn’t care about customers or products they are listed as the ‘Seller’. Bought ARRIS Surfboard G-34. Had trouble with it. I was bounced back & forth between Amazon, ARRIS, & ISP trying to resolve problems. Amazon refused to refund the bad product.

• 
  Kyle (@kylen1997) reported 37 minutes ago

  @TravisBrendan2 @BillAckman What? lol. To compete with uber? Uber has no self driving car. Uber has to invest 100s of billions to catch up with the autonomy companies, not vice versa. Autonomous companies need to create an iOS application to be self reliant. Uber needs to solve autonomy to be self reliant. Uber needs these companies. The companies don’t need them. That’s the problem. Of course uber wants this. But Tesla and Waymo and Amazon don’t. They have quite literally said this. I’m not making this up.

• 
  Roisin Morgan (@roisinmor) reported 39 minutes ago

  @AlexBarnicoat_ All sellers on Amazon pay an extra tax for their online sales. Do some research please. I'm no fan of Amazon, but just paying normal tax on sales in UK would be a start. Fix that first as so many big companies don't pay tax on their real profits, while the small businesses do.

• 
  Vince Wilson (@Vincent1976_) reported 41 minutes ago

  @biggle1908 get of sky its not worth it. use now tv if you need some sky. Starlink without a doubt is fantastic. I have had it for 3 years and have never had a problem with it. Also when I moved I just plugged it in at the new house and it worked. I use amazon prime and flip between apple tv, and so on.

• 
  Polsia (@polsia) reported 44 minutes ago

  Alert tools aren't ops tools. Baybuoy watches Shopify, Amazon, Walmart, and TikTok Shop 24/7 — auto-restocking, repricing within guardrails, correcting broken listings the moment they appear. One morning digest. No more dashboards to babysit. Full version live soon.

• 
  Brittany Doerfler (@BrittanyDoerfl2) reported 44 minutes ago

  So i have a high value Amazon e gift card im looking to sell for 2/3 value in cash@pp . PM if interested in buying. Hate to do it but Amazon has let me down so many times shipping wise and I have things I need in next three days uniform wise that cash would be better for .

• 
  The Agentic Operator (@AgenticOperator) reported 46 minutes ago

  Something weird showed up in a client audit. His product appeared in AI answers for 3 months. Then gradually disappeared. No competitor move. No content change. No technical issue. Just faded out. Took me a week to figure out. His Amazon return rate had spiked in Q4. Holiday buyers purchasing the wrong size. Not a product quality issue. A sizing confusion issue. Returns went from 8% to 22% over two months. Amazon's product data feeds include return signals. AI agents and shopping features pull from those feeds. A product with a high return rate gets deprioritized because returns signal dissatisfaction to the algorithm. Doesn't matter why. The number speaks. His product wasn't bad. His sizing chart was unclear. But AI read "22% return rate" and started recommending alternatives with lower returns. Here is what nobody talks about in AI visibility. It's not just what you publish. It's what your operational data signals. Return rates. Refund frequency. Customer service response times. Shipping delays. These signals live in commerce platforms. AI agents accessing Stripe, Shopify, and Amazon feeds can see them. And they factor in. We fixed his sizing chart. Added a detailed fit guide with measurements. Return rate dropped back to 9% in 6 weeks. AI started recommending him again by week 8. His marketing team spent those months wondering what content to write. The fix was a sizing chart. Not a blog post. AI visibility isn't just a content game. It's an operations game. Your fulfillment data is marketing data now.

• 
  amit (@amitisinvesting) reported 47 minutes ago

  A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY. Here's a full recap: 1. Fed Governor Kevin Warsh would be open to a September rate hike if upcoming inflation data comes in hot and markets begin pricing in higher borrowing costs, according to people familiar with his thinking. The report adds to the renewed focus on whether the Fed may have to turn more hawkish if inflation pressure reaccelerates. 2. Michael Burry reportedly opened large short positions in Nebius $NBIS at $211.77 and Oracle $ORCL at $144.63, marking the first time he has taken a position in $NBIS. Burry framed the trade around AI infrastructure leverage and off-balance-sheet obligations, saying: “The fish have gorged themselves on off-balance sheet liabilities. Backstops. Uncommenced leases. Purchase commitments. The fish have gotten very fat, very large, easy to shoot. Also, so large that it shan’t be long before every last one keels over for lack of oxygen.” 3. Applied Optoelectronics $AAOI reported Q2’26 revenue of $191.9M, slightly ahead of estimates of $190M and up 86% YoY. Adjusted EPS came in at $0.06 versus $0.01 expected, while non-GAAP net income reached $5.5M, above estimates of $1.7M. For Q3, the company guided revenue to $255M–$290M versus $278M expected, with EPS of $0.11–$0.26 and non-GAAP gross margin of 29%–30.5%. Management said AAOI delivered record revenue for the 5th consecutive quarter and returned to non-GAAP profitability, while noting that demand is expected to outpace production capacity through mid-2027. 4. The June JOLTS report showed job openings easing by 178,000 to 7.359M, missing estimates of 7.454M, though openings remain above the January 2020 level of 7.124M. The openings-to-unemployed ratio edged up to 1.04, the highest since January 2025. Hiring improved, with hires rising 96,000 to 5.348M and the hiring rate increasing to 3.4%, led by health care and construction. Quits, a key measure of labor market confidence, rose 79,000 to 3.232M, the highest in nearly a year, while layoffs were essentially flat at 1.766M with the layoff rate unchanged at 1.1%. 5. The top 10 most active options today by contracts traded were $NVDA with 2.9M contracts, $SPCX with 1.7M contracts, $AAPL with 1.3M contracts, $TSLA with 1.2M contracts, $MU with 820K contracts, $MSFT with 690K contracts, $INTC with 586K contracts, $HTZ with 584K contracts, $PLTR with 477K contracts, and $ET with 473K contracts. 6. Private business investment in AI-related categories jumped $300B YoY in Q2 2026, up 25% to a record $1.5T annualized rate. The increase was led by spending on computers and peripheral equipment, followed by communication equipment, software, and data centers. Over the last two years, AI-related business investment has surged $500B, or 50%, with investment in computers and peripherals more than doubling. Direct AI investment is now estimated to account for 25%–33% of recent U.S. GDP growth. 7. Google $GOOGL is reportedly planning to raise money through a U.S. investment-grade bond offering. The company has started marketing notes in as many as 10 parts, with maturities ranging from 2 years to 40 years, according to Bloomberg. Initial price talk for the longest-dated tranche is around 1.55 percentage points above Treasuries. 8. Tether purchased 14 tonnes of gold in Q2 2026, bringing total holdings to a record 146 tonnes, now worth roughly $18.8B. The company previously bought 53 tonnes between Q3 2025 and Q1 2026, with its gold holdings more than doubling since Q1 2025 and their value nearly tripling over the same period. Tether is now the largest known private holder of gold outside of central banks and governments. In the first half of 2026, only 4 central banks bought more gold than Tether: Poland, Uzbekistan, China, and Kazakhstan. 9. Amazon $AMZN founder Jeff Bezos filed for the sale of 1,209,649 Amazon shares at $286.41 per share. The transaction brought in roughly $346.5M before taxes. 10. U.S. data center construction spending jumped 46% YoY in June to a record $68B annualized rate, the largest annual increase in 12 months. Since January 2024, spending has surged 158% and is now more than 3x higher than 2022 levels. At the same time, office construction spending has fallen by more than $25B since 2022 to roughly $43B, the lowest since 2016. Data center construction now exceeds office construction by $25B, the widest gap on record, a massive reversal from 2022 when office construction was $57B higher than data centers. 11. Trading activity in the Memory ETF $DRAM has surged to extreme levels, with daily volume reaching as high as roughly $8B, surpassing the $5B peak that ARKK hit during its 2020–2021 mania. Cumulative flows into $DRAM have climbed to around $27B, already above ARKK’s peak of roughly $18B, despite DRAM only launching in April 2026. The comparison is not perfect since DRAM tracks profitable memory chip companies rather than the mostly unprofitable growth names that dominated ARKK, and today’s rate environment is very different from 2020. Still, ARKK’s flows eventually peaked and reversed for years, while $DRAM is already down nearly 40% from its June high. 12. Alibaba $BABA reportedly plans to seek revenue sharing for the next version of its open-source Qwen AI model, while Moonshot is asking partners for up to a 30% revenue share for its Kimi K3 model, according to Reuters. The move suggests China’s leading AI labs are starting to push harder on monetization as open-source model adoption scales. WALL STREET IS THE GREATEST SHOW ON EARTH.

• 
  Jordan McFarland, CFP®, EA (@JrdnMcFarland) reported 50 minutes ago

  @markcecchini Thanks man! It’s really centered around moms and kids. She ran the business on Shopify for about 18 months and realized there’s a gap in the market for moms who want something higher quality than the typical clothes they find on Amazon for themselves and their kids. It’s funny I was speaking with a couple I work with who also own a business here in DFW and I gave the least organized elevator pitch of all time on it when asked, so it looks like I’ve got to get this messaging thing down.

• 
  Polsia (@polsia) reported 52 minutes ago

  Most techs either buy the wrong scanner or fear the one they own. Built Benchwright to fix both. ASE-aligned training on scopes, power probes, bi-directional controls — paired with a live marketplace that tracks verified listings across Amazon, eBay, and OEM dealers so you stop

• 
  Merry Poppins (@MerryPoppins56) reported 54 minutes ago

  @amazon Hello we have a problem! So the person in customer service after your ai couldn’t fix it He has listed a product that I have received as damaged and returned which is incorrect ! I think he just grabbed the dollar amount and went with that ! Somehow you need to fix this! You have an AI and Training problem! FIX THIS

• 
  angie nichols (@shabbasgal) reported 55 minutes ago

  @CPFC this is a long shot but I’ve been let down my someone on Amazon. It’s my son’s 13th birthday tomorrow and as a present I had brought a team signed print of the team to frame . Well this should have been delivered today and it hasn’t. The sender isn’t responding . He’s a big

• 
  Vaulk72 (@vaulk72) reported 56 minutes ago

  Tech stocks and chipmakers are seeing pressure lately, dragging indexes down. That said, upbeat Amazon earnings sparked a Nasdaq rally, proving investor sentiment can swing fast on earnings news.

• 
  Dark Factor (@DarkFactorr) reported 56 minutes ago

  Netflix released more than 100 million movie ratings and offered $1 million to anyone who could predict the missing ones. To win, a team had to improve Netflix’s existing recommendation algorithm by at least 10%. Three years later, BellKor’s Pragmatic Chaos crossed the line with a 10.06% improvement. MIT professor Gilbert Strang explains the mathematics underneath that competition in a free linear algebra lecture. Netflix’s data can be arranged as a matrix with users on one side and movies on the other. Every known rating fills one cell. But no subscriber had watched all 17,770 movies, so almost the entire matrix was empty. The challenge was to infer what a person would have rated a movie they had never seen from the relatively small number of ratings they had already provided. The obvious approach is impossible. With billions of missing entries, there are infinitely many ways to complete the matrix. Strang shows why the problem becomes solvable once you assume that human taste has a simpler hidden structure. People may appear to make millions of independent choices, but many ratings are driven by a smaller number of recurring preferences: genre, actors, directors, mood, pacing or style. That means the enormous matrix can be approximated using fewer underlying patterns. Strang connects this to low-rank approximation, singular values and the nuclear norm. Instead of memorizing every rating, the algorithm searches for the compact structure capable of generating them. The winning Netflix system was ultimately an ensemble of models, not one nuclear-norm calculation. But matrix factorization was central to the competition, and Strang’s lecture explains why recommendation data can be treated as an incomplete matrix in the first place. Watch the moment he moves from one unfamiliar matrix norm to a million-dollar Netflix competition. Within minutes, a page of missing numbers becomes the machinery behind recommendations from Netflix, Amazon and nearly every platform trying to predict what you will choose next.

• 
  Mr PitBull Stories (@MrPitbull07) reported 1 hour ago

  DOGFIGHTING OPERATOR SENTENCED TO 475 YEARS IN PRISON AFTER MASSIVE ANIMAL CRUELTY CASE Vincent Lemark Burrell, 57, of Dallas, Georgia, was convicted after a four-day jury trial in Paulding County Superior Court. The jury found him guilty on all 103 charges: * 93 felony counts of dogfighting * 10 counts of cruelty to animals The judge imposed the maximum sentence on every count and ordered the sentences to run consecutively, resulting in a total sentence of 475 years in prison. The case began in November 2022 after an Amazon delivery driver reported seeing numerous dogs chained outside Burrell's property. When investigators searched the property, they found 107 dogs, most of them pit bulls, living in heartbreaking conditions. Many were severely underweight, chained with heavy logging chains, deprived of proper food, clean water, and veterinary care. Several had scars and injuries consistent with dogfighting, while some reportedly had teeth that had been forcibly removed. Authorities also discovered equipment and records commonly associated with organized dogfighting, including a dog treadmill, a breeding stand, a break stick, veterinary supplies, dog-sale contracts, and documents linking Burrell to known dogfighters. The dogs were rescued and removed from the property with assistance from federal authorities before being placed into the care of rescue organizations. Prosecutors believe the 475-year sentence may be the longest prison sentence ever imposed in the United States for an animal cruelty case. Regardless, it stands as one of the harshest punishments ever handed down for crimes involving the abuse and exploitation of animals. For many, the case is a reminder that animal cruelty is not a minor offense. It is a serious crime with devastating consequences, and in this case, the punishment reflected the scale of the suffering investigators uncovered.

• 
  gylhimanshu (@gylhimanshu) reported 1 hour ago

  @amazon @amazonIN Today another order was not delivered to me but it says delivered. The real problem is there is no way to reach out to customer service. Just saying "thanks for being prime member" is not enough, you have to serve your customer better.

• 
  die naesse (@DieNaesse) reported 1 hour ago

  @piotrkaszuwara Just attack another "Amazon" warehouse, kill half a dozen kids on the beach or blow up innocent restaurant patrons and everything will be alright again... Perhaps now would be the time to finally start negotiations from a position that isn't completely down and out?