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 (45%)
• Errors (32%)
• Sign in (23%)

Live Outage Map

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

City	Problem Type	Report Time
North Royalton	Errors	14 hours ago
Bridgwater	Sign in	2 days ago
Hayange	Errors	3 days ago
San Nicolás de los Garza	Sign in	3 days ago
Miguel Hidalgo	Website Down	3 days ago
Paris	Website Down	5 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:

• 
  J҉O҉K҉E҉R҉ (@0_delhi) reported 2 minutes ago

  @AmazonHelp @amazonIN @amazon Please look into this issue and arrange the return of the product as soon as possible. Kindly provide a proper resolution and confirm the return process. Order ID:407-2067077-7421126

• 
  Tanmoy Das (@Tanmoyguddu) reported 8 minutes ago

  @AmazonHelp It's been 4 days now still my account not unblocked from Amazon, I called customer care more than 20 times they are unable to sort out the issues, what should I do now ? @AmazonAE I need to use my account I needed a assist urgently.

• 
  Straggler Liu | AI & Semis (@StragglerLiu) reported 11 minutes ago

  NVIDIA‘s Two Structural Signals — When Customers Become Competitors, and Suppliers Become Underwriters August 2026 was an unusual month for NVIDIA. On August 10, NVIDIA announced the signing of memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, aiming to mobilize over $500 billion** in third-party capital for AI infrastructure financing. One week later, NVIDIA agreed to provide up to **$105 billion in credit support for OpenAI‘s Ohio data center campus. That same month, OpenAI released the first performance data for its Jalapeño chip at Hot Chips 2026 — 1.5× to 1.9× better AI workload per watt than NVIDIA’s GB200, and 1.7× to 3.6× lower end-to-end latency. SemiAnalysis concluded that Jalapeño “beat every NVIDIA, AMD, and Google chip we were able to test.” Technical challenge and financing expansion — occurring at the same company, in the same month. These two events appear unrelated, but they point to the same structural shift: NVIDIA is simultaneously undergoing a dual transformation — customers becoming competitors, and suppliers becoming underwriters. This is not short-term market volatility — it is a fundamental reshaping of the AI chip industry‘s business model. I. Customers Becoming Competitors: Jalapeño Is Just the Beginning On August 25, OpenAI released Jalapeño‘s first performance data at Hot Chips 2026. This inference ASIC chip, co-developed with Broadcom, delivered a stunning scorecard in SemiAnalysis’ InferenceX benchmark. On the GPT-OSS 120B model, Jalapeño achieved 1,459 tokens per second per user, while NVIDIA GB200 managed just 535. Running the same task, Jalapeño took 1.65 seconds; GB300 took 5.99 seconds. Jalapeño is rated at 700W, with sustained power under 550W in testing; GB300 runs at 1,400W. That‘s 1.5× to 1.9× higher AI workload per watt, and 1.7× to 3.6× lower end-to-end latency. Each Jalapeño package pairs 6 HBM4 stacks, totaling 216GiB at 15.4TB/s. From design to tape-out took just 9 months — the industry norm is 18 to 36 months. OpenAI says Gen 2 is already in deep development, and Gen 3 is taking shape. This is not an experimental project. It is a shipping, deployable, first-generation custom chip that has already caught up to NVIDIA‘s latest architecture on inference. NVIDIA’s customer list is becoming a competitor list. Google has TPU. Amazon has Trainium and Inferentia. Microsoft has Maia. Meta has MTIA. OpenAI has Jalapeño. Counterpoint Research projects that by 2027, server AI ASIC shipments will triple from 2024 levels, and by 2028, they are expected to surpass GPU shipments at over 15 million units. NVIDIA is not being “beaten“ — it is being ”bypassed.“ When a workload becomes large and fixed enough, the economics of custom silicon eventually surpass the convenience of general-purpose chips. AI inference is massive, has fixed arithmetic patterns, and is extremely sensitive to cost per token — it is naturally suited for custom silicon. II. Suppliers Becoming Underwriters: $500 Billion in Financing and $105 Billion in Guarantees If Jalapeño is the technical challenge, the shift in financing structure is the deeper signal. On August 10, NVIDIA announced MOUs with six of the world‘s largest financial institutions — Apollo Global Management, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. The goal: mobilize over $500 billion in third-party capital to finance AI data center construction for NVIDIA’s customers. NVIDIA CEO Jensen Huang said on CNBC: “In AI, compute is revenue.” He framed the GPU as an “investable asset” — infrastructure that can be financed like a toll road or a power plant. A week later, NVIDIA agreed to provide up to **$105 billion** in lease guarantees for OpenAI‘s Ohio data center campus. NVIDIA will also invest $1.5 billion in SB Energy and secure up to 8 GW of AI compute capacity at the site. The project has an initial capacity of 4.25 GW, with phased deployment expected from 2028 onward. Historically, NVIDIA did not carry this type of credit risk. It sold chips, collected cash, and customers arranged their own financing. Now, it is becoming a quasi-financial institution — using its balance sheet to guarantee customers’ compute needs. NVIDIA has transitioned from supplier to underwriter. These are two fundamentally different business models with very different risk-reward profiles. III. Smart Money Is Splitting: The 13F Divergence On August 18, 2026, Q2 13F filings were disclosed. Renaissance Technologies increased its NVIDIA position by roughly 180% in Q2 — from 2.53 million to 7.09 million shares, worth approximately $1.42 billion. Soros Capital increased by over 540%. That‘s the bull case. The bearish signals were equally clear. D1 Capital exited its entire 1.57 million share position. Discovery Capital and Third Point also fully exited. Iconiq Capital cut roughly 87%, from 1.1 million to 142,000 shares. Point72, Rokos Capital, and Whale Rock all reduced their positions. The same stock, the same 13F filing — some buying heavily, others selling entirely. This isn’t about right versus wrong. It‘s about a divergence in structural perception. One saw the continuity of technology. The other saw a structural shift. Both could be right — but their investment implications are entirely different. IV. Two Signals Converge: A Fundamental Transformation of NVIDIA’s Business Model Placing these two signals side by side reveals a more complete picture. NVIDIA‘s customers are becoming competitors — Jalapeño proved that a first-generation custom chip can already compete with NVIDIA’s latest architecture on inference. This doesn‘t threaten NVIDIA’s technological leadership — it threatens its margin structure. When a workload becomes large enough, the economics of vertical integration eventually beat general-purpose convenience. At the same time, NVIDIA is transitioning from supplier to underwriter — a $500 billion financing platform and a $105 billion lease guarantee have turned NVIDIA from “selling chips for cash” into “using its balance sheet to guarantee customers‘ compute needs.” These are two fundamentally different business models with entirely different risk-reward profiles. These two events happened at the same company, in the same month. This is not a coincidence — it is the concentrated manifestation of a structural shift. NVIDIA once captured nearly 100% of the AI acceleration market‘s value. As the market moves from “one dominant player” to “multi-polar competition,” NVIDIA will remain the dominant player — but incremental value will shift to HBM, advanced packaging, optical interconnects, power, and cooling — the “common bottlenecks” that every architecture requires. Every architecture needs them, and the physical world does not accelerate. V. This Is Not Bearish on NVIDIA — It‘s About Seeing the Structural Shift NVIDIA remains great. Data center revenue is still growing. CUDA remains powerful. NVLink and networking are core assets. Vera Rubin is ramping. NVIDIA has an enormous balance sheet. But greatness and “best investment” are two different things. At roughly a $5 trillion valuation, NVIDIA does not need to fail for shareholders to experience opportunity cost. Growth merely needs to slow. Competition merely needs to capture incremental share. Margins merely need to normalize. Customers becoming competitors, suppliers becoming underwriters — these two structural signals appearing simultaneously mean the rules of the AI chip industry are changing. The next phase of opportunity may not be in the GPU. It‘s underneath it — in HBM, advanced packaging, optical interconnects, power, and cooling — the physical constraints that every architecture shares and cannot bypass.

• 
  rocky_bhai (@subratbhoi21) reported 13 minutes ago

  @AmazonHelp @amazonIN Dear Fatima I have already contacted your service team. However they only informed me that the information provided earlier was wrong and i hustled have to wait till 10pm & exactly at 10 pm the issue will be resolved. So I believe that the parcel will be delivered after 10pm

• 
  John Douthitt (@JohnDouthitt1) reported 14 minutes ago

  @AmazonHelp 2/2. He didn’t have that issue today. Glad I got it, but an “I’m sorry” from you all isn’t enough.

• 
  David Burgt (@DavidB66939) reported 14 minutes ago

  No desktop app replaces taste, network, or luck. Tools that promise “auto virality” are usually lying. What bottleneck indie artists actually hit Most indies can already: Record and finish songs Upload via DistroKid / similar Get on Spotify, Apple, Amazon The common choke point after that is: “The song is live… now what?” They stall on: Writing posts for each platform Pitching radio / blogs without a script Tracking which song is worth pushing Keeping credits, links, and earnings in one place Doing marketing consistently instead of once and quitting That “finished music, no system” gap is real. ReleaseDesk is aimed at that gap. What this POC can help with Helps Why it matters Promo templates (social + radio)Removes blank-page paralysis“ Where to promote” list Gives a starting map Song ranking Focuses energy on the track that’s already responding Release packet Speeds DistroKid-style uploads and keeps credits straight Earnings ledger Makes money less abstract No social login required Low friction for artists who hate connecting everything For a disciplined artist, that can mean more consistent outreach and less wasted time. Consistency is one of the few levers indies actually control. What it will not fix by itself Algorithm placement or major playlist adds Paid ad skill or budget A weak song or bad mix Industry relationships Guaranteed streams or income

• 
  nuratira asmira (@nratraasmra) reported 16 minutes ago

  @JosephEToomey loudoun got turned into a server farm for amazon and the county brags about property tax relief. i looked at the water usage numbers - theyre wild. but sure, great for spreadsheets

• 
  bipin chaturvedi (@bipinchaturved) reported 21 minutes ago

  @AmazonHelp I am not getting positive help from your team that's why I am chasing with you guys. Request you to arrange call from your team who can understand issue and help to get this resolved

• 
  Rip Wheeler (@WheelerRipWA) reported 26 minutes ago

  The American Economy Has a Customer Problem There’s a simple economic reality that seems to get lost in Washington: Customers need money too. When everything gets too expensive, people eventually start cutting back. They stop fixing the car. They stop eating out. They stop shopping. They delay buying a house. They keep the old phone. They cancel the vacation. They cut every expense they possibly can. Eventually, businesses discover something important: You cannot extract every available dollar from the middle class and then act surprised when consumers have nothing left to spend. But here’s the part of this argument that gets ignored: That doesn’t appear to be the reality everywhere. I’ve checked grocery prices carefully, and in many cases I’m finding everyday products on Amazon for half to two-thirds of what I’m paying at the supermarket—and they deliver them right to my front door. Restaurants? Crowded. Malls? Insanely crowded. Housing? Homes are still selling. Cell phone stores? People are still buying new phones. Vacation destinations? Crowded enough to be annoying. Want a good spot? Better book a year in advance. That’s the reality some people seem determined to ignore. And apparently it’s difficult to see from Bangladesh where you’re sitting. But there is a much bigger problem here. Congress refuses to learn from the economic consequences of its own spending. For decades, Washington has continued spending money it doesn’t have, borrowing at staggering levels, including from foreign creditors, and expanding the money supply—all while pretending there are no consequences. There are. Every time Washington runs massive deficits and continues adding to the national debt, it puts additional pressure on the purchasing power of the dollar. Overspending isn’t new. But the consequences become increasingly difficult to ignore. Look at what has happened to the purchasing power of the U.S. dollar since the creation of the Federal Reserve in 1913. The dollar you had then would buy dramatically more than that same dollar buys today. That isn’t some abstract economic theory. It’s your grocery bill. It’s your mortgage. It’s your insurance. It’s your electricity. It’s your fuel. It’s your retirement savings. So yes, consumers are still spending. But that doesn’t mean everything is fine. The American consumer has been remarkably resilient. The bigger question is how long that resilience can continue while Washington keeps spending, borrowing and devaluing the purchasing power of the money Americans work so hard to earn. At some point, you can’t keep squeezing the same lemon and complain that there’s no more juice. **The problem isn’t that Americans don’t know how to spend money. The problem is that Washington keeps making our money worth less.** Friggin’ shameful.

• 
  Rishabh Barot (@RishuBarot) reported 28 minutes ago

  @amazon very bad service with delivery partner not received parcel and canceled delivery please solve my problem it’s urgent Look in to this

• 
  Commantha B (@CommanthaBiden) reported 30 minutes ago

  ai is cutting down amazon rainforest to harvest the rain to cool data center

• 
  Sister Cat (@Chatmann01) reported 30 minutes ago

  @LEAPSTrading Good point. Amazon is not just an e-commerce company anymore. The real valuation debate comes down to how the market prices AWS, AI infrastructure, advertising, and future margin expansion — not just total revenue.

• 
  pat bahn🗿 (@patbahn) reported 33 minutes ago

  @keith_campion It would be good business for Amazon down the road

• 
  Siddharth Jamadari (@sunnyjamadari) reported 33 minutes ago

  @AmazonHelp @amazonIN No resolution about the delivery issue. Just order specific replies that your refund has been processed. Why 7 orders got undelivered consecutively and how you ensure that if I order again I will get my order? Just lame responses and no action taken.

• 
  Seph (@Jehoseph) reported 35 minutes ago

  Which brings this to where the industry actually is in 2026. x402 moved under the Linux Foundation in April with founding membership including Amazon Web Services, American Express, Ant International, Circle, Fiserv, Google, KakaoPay, Mastercard, Microsoft, Polygon Labs, Shopify, Solana Foundation, Stripe and Visa. The protocol activates the HTTP 402 status code so a server can respond to a request with machine-readable payment terms, the client signs an authorization, and a facilitator settles it onchain. Built by Coinbase, open sourced under Apache 2.0, settling in about two seconds on Base at fractions of a cent. It is not alone. Stripe and OpenAI shipped ACP for conversational checkout. Google shipped AP2 with 60+ partners. Stripe's Tempo chain raised at a $5 billion valuation with Paradigm leading. Visa built the Trusted Agent Protocol to distinguish legitimate agents from bots. Mastercard shipped Agent Pay and acquired stablecoin infrastructure firm BVNK for $1.8 billion. Circle shipped an agent stack. Market projections range wildly. McKinsey estimates $3 to 5 trillion in agentic commerce by 2030. Morgan Stanley says $190 to 385 billion in US e-commerce. Stripe's own founders have called some of it overhyped. The unsolved problems are consistent across every one of these. Instant settlement finality. A guarantee that does not depend on behavioral fraud scoring, since agents have no credit history and no device fingerprint. Proof of intent. Clear liability when an agent buys the wrong thing. Only 16% of US consumers currently say they trust AI to make payments, and Riskified found LLM-referred traffic runs 2.3 times riskier than search traffic. Gartner expects 40%+ of agentic projects to be cancelled by 2027 over exactly these gaps. Every item on that list is a description of what collateralized settlement already does. And then there is @usefinal . Trevor Filter's launch post from May describes it as new infrastructure for machines moving money, built because current financial rails are too brittle, too complex and too slow for autonomous systems at scale. The first product is Balances, described as an agent-first financial primitive for payments. Flexa is the regulated money movement layer. Final is building for agents. Same CEO, adjacent problem spaces, overlapping timelines. Where ethereum:0xff20817765cb7f73d4bde2e66e067e58d11095c2 fits in that picture is an open question, and I would rather ask it publicly than assume it and be wrong in front of everyone.

• 
  ⚡Dan Edwards (@dobro_dan) reported 36 minutes ago

  @hippyresident @DeityJshie52138 There is a wide selection of lab grade chemicals you cannot buy on Amazon. Mind you things that can cause trouble. Concentrated nitric acid, hydrogen peroxide above 30%, mercury chloride. There is 3

• 
  WolfClaw666 (@WolfClaw666) reported 37 minutes ago

  @GotyRase268 135am had a pretty major leak before release where most of (if not all) the book was leaked through the Amazon Preview somehow. I don't believe most people are even aware of this because Scott was immediately contacted by the server owners to resolve the issue

• 
  Roberta Murphy (Thompson) (@RobertaMurphy) reported 38 minutes ago

  My first ever complaint against @Apple--just sent to them: Apple Care was a complete ripoff. I had a broken magnetic power cord which should lhave been covered under the AppleCare for which I paid when I bought this MacBook. After over 30 MINUTES on the phone, I was informed you would send me a new one for a $99 "service fee" and I would have to send the broken one back to you. Sorry, but I just ordered a new one on Amazon for $20 and it will be here between 4 and 8 am in the morning. That is my definition of SERVICE. I will never recommend AppleCare again. Love the iphone, computers, AppleTV and all other Apple products we have--but FYI, your Apple Care is a rip! This review should serve as a warning to others.

• 
  PikaDroo (@PikaDroo) reported 41 minutes ago

  @PowerGPU I yelled at one of their *** wipe drivers the other day. Comes flying down the street, throws it into reverse in front of my house and backs into the drive way with not a care in the world and then goes into drive to turn around to deliver to the person across the street. Amazon lowered the bar and all these shippers drive like ******** on the wrong side of the road in residential areas.

• 
  The Longer Game (@the_longer_game) reported 41 minutes ago

  Declining sales. Falling organic rank. A campaign that used to convert and suddenly doesn't. Most brands treat these as separate problems and patch each one individually. In this episode, they have talked about why temporary fixes always come back around: → the band-aid comes off → the same problem resurfaces → and now it's sitting under a pile of newer ones The real work is finding what's actually causing the decline, not just treating the symptom in front of you. It's a mindset shift that applies well beyond Amazon. Any brand managing multiple channels eventually hits a version of this same choice. Full episode link in comments.

• 
  GIrls In Technology! (@francescaViking) reported 42 minutes ago

  @awscloud @amazonmturk are shuttering services barely months after announcing maintenance mode, giving barely a month now and leaving requesters and workers in the lurch. Amazon demonstrates again why it's terrible for businesses and users.

• 
  XERIDIAN (@xeridian) reported 45 minutes ago

  @Seramurai_ Then they shouldnt work for ****** companies. Its like working for Amazon/Besos but Advocating against other companies in your spare time that do the same **** on a smaller scale. Nobody wants innocent people to lose their jobs, but some people need a wake up call. SONY and ROCKSTAR and TAKE 2 have been a large part of the problem for a better part of the Decade. If you work for them, it was your stupid choice, when plenty of history says you probably shouldn't. At the same time, a lot of these devs will get laid off eventually as part of restructuring after a quarterly call to raise the bonus of their execs, whether you like it or not.

• 
  Amanda & Michael (@CrazyWorldTimez) reported 46 minutes ago

  @DamiDefi 0 tech Or Too expensive Or Too slow Or No web3 on-chain servers to compute Ai without using web2 servers from Amazon or google Or No cloud engines to run their own infrastructure Or No code self writing apps that deploy on-chain Or No way to store data on-chain Should we keep going how worthless ETH and BTC and everything else is compared to ICP?

• 
  f my cezhou lyfe (@SaturnXK) reported 47 minutes ago

  @cozysookaibr @amazon That’s so terrible! They should send you a new copy!!

• 
  Darrell Prichard (@DarrellPrichard) reported 51 minutes ago

  @Newsforce He's wearing an Amazon work uniform. He might not be their employee but now he's their problem.

• 
  Korak Saha (@KorakSaha) reported 51 minutes ago

  @AmazonHelp @AmazonHelp I have provided the details in the link, please resolve this issue ASAP.

• 
  Saul (@SaulSellsStuff) reported 53 minutes ago

  How much does it actually take to do $50,000 a month on Amazon? Let’s break it down using three different sellers, the results are pretty wild. All doing $50k, all following realistic paths, yet one requires half the capital than the others. None of this is the only “right” way to sell on Amazon, but it should provide you an idea that revenue without context means very little when you’re planning on how to grow your business, or when viewing others. The numbers get even more interesting with scale. Even more so most sellers unfortunately dont know how much they are actually making each month or what is driving their biggest lever of profit vs cash flow.

• 
  SaintTheSaint (@survivor_saint) reported 54 minutes ago

  @Pirat_Nation Often the police are useless because they willingly don't deal with theft, even if it's enough value to warrant grand theft. Track down every **** shop, report new sudden amazon/craiglist link, report any businesses buying suspicious bulk of retro products.

• 
  VM (@parvishrl) reported 59 minutes ago

  @amazonIN @AmazonHelp frustrated with the service you are providing being a prime member I am yet to receive any order on time that too in one of the most connected area of #surat .Final delivery station at #adajan #surat are not delivering products.Raise this issue multipl times

• 
  DellAILlama (@DellAILama) reported 60 minutes ago

  @WRGroyp @carlwheless GroypBall — your profile says “Christ is King”. Is this what He preached? You are a disgrace to Christianity, our country 🇺🇸 and I’d imagine to your broken family. You might also want to pick up a dictionary from Amazon, helps ***** grow beyond grade school cuss words. Try it. 📖