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 (26%)

Live Outage Map

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

City	Problem Type	Report Time
Kalgoorlie	Website Down	6 hours ago
Newark	Errors	1 day ago
Greenfield	Errors	1 day ago
Marseille	Sign in	1 day ago
Saint-Rémy-de-Provence	Sign in	1 day ago
Gaillac	Sign in	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:

• 
  TheValueist (@TheValueist) reported 3 seconds ago

  $META $NVDA $MU $SNDK $LITE EXECUTIVE ASSESSMENT The Bloomberg report is strategically material because it changes the interpretation of Meta’s AI infrastructure buildout from a largely binary internal-consumption bet into a potentially monetizable infrastructure platform. The most objective read is that Meta is not yet becoming a full-stack hyperscaler in the AWS, Azure, or Google Cloud sense; rather, it is developing a real option to commercialize excess AI compute and potentially host model APIs if internal AI demand does not absorb the full capacity being built. This distinction matters for valuation. A cloud resale or model-access business could partially reduce the market’s concern that Meta is overbuilding uneconomic AI capacity, but it does not yet solve the core investment question: whether $125-145 billion of 2026 capital expenditures can generate returns above Meta’s cost of capital through ads, engagement, AI assistants, business agents, wearables, and possibly third-party infrastructure revenue. Meta’s Q1 2026 revenue rose 33% to $56.31 billion, operating income was $22.87 billion, and free cash flow was still positive at $12.39 billion despite $19.84 billion of capex, which confirms that the Family of Apps cash engine remains exceptionally strong. However, the updated 2026 capex guide implies a dramatic sequential ramp from Q1 levels and an 87% increase at the midpoint versus 2025 capex of $72.22 billion, making the durability and monetization of AI infrastructure the central underwriting issue for the equity. (Meta) The most important analytical point is that the proposed business would be more valuable as a utilization hedge than as an immediate new AWS-scale revenue line. If Meta can sell excess capacity at a premium to its acquisition cost, the downside case for overbuild improves. If Meta can also create a credible model API around Muse Spark and future Muse models, the company could add a higher-margin AI platform layer above raw GPU resale. However, the hurdle for value creation is materially higher than the headline suggests. Raw AI compute resale is a capital-intensive, depreciating-asset business with commodity risk, high customer concentration risk, power constraints, rapidly changing silicon economics, and limited differentiation unless the provider has superior cluster performance, software orchestration, model access, or committed demand. Mature cloud businesses generate attractive operating margins because they bundle compute with storage, data services, identity, security, governance, developer tools, partner ecosystems, enterprise sales coverage, and support. Meta has substantial infrastructure competence but does not yet have a comparable enterprise cloud control plane, broad service catalog, field sales organization, compliance posture, or installed enterprise workload base. The near-term investment conclusion is therefore balanced: the report is incrementally positive for narrative and downside mitigation, but insufficient on its own to justify underwriting a large stand-alone cloud revenue stream without disclosed customers, pricing, utilization, margins, SLAs, and contracted backlog. CAPEX CONTEXT AND WHY THE CLOUD OPTION MATTERS The cloud initiative should be interpreted against Meta’s unprecedented infrastructure commitment. Meta raised 2026 capex guidance to $125-145 billion from $115-135 billion, citing higher component pricing and additional data center costs for future capacity. Q1 2026 capex, including principal payments on finance leases, was $19.84 billion, which means the full-year guide requires roughly $105-125 billion of additional capex over the remaining 3 quarters, or approximately $35-42 billion per quarter. This is a major change in capital intensity for a company historically valued on high-margin advertising cash flow and disciplined capital returns. The balance sheet can support the investment for now, with $81.18 billion of cash, cash equivalents, and marketable securities as of March 31, 2026, but the trajectory is no longer a normal “growth capex” cycle; it is an infrastructure land-grab whose returns depend on model scaling, product monetization, and utilization. (Meta) Meta’s contractual commitments show that the company has already moved beyond an optional exploratory phase in infrastructure. As of March 31, 2026, Meta disclosed $182.88 billion of operating and finance leases not yet commenced, primarily for data centers, colocations, and network infrastructure, with lease terms extending from more than 1 year to 30 years. Meta also disclosed $237.67 billion of non-cancelable contractual commitments, mostly related to third-party cloud capacity, servers, network infrastructure, data centers, and Reality Labs hardware, with $42.25 billion due in 2026 and $47.65 billion due in 2027. In addition, Meta disclosed up to $14.72 billion of contingent cloud-capacity purchase obligations and stated that April 2026 infrastructure contracts increased non-cancelable commitments by another roughly $24 billion. These disclosures are central to the investment debate because they make excess-capacity monetization financially relevant. If the infrastructure is underutilized internally, a third-party compute business can convert otherwise idle or suboptimally used capacity into revenue. If internal demand remains high, the external cloud opportunity may remain theoretical because the highest-return use of capacity could still be Meta’s own ads, ranking, messaging, agent, and wearable AI workloads. (SEC) The market’s positive reaction to the Bloomberg report is understandable because it gives investors a potential answer to the “what if Meta overbuilds?” question. The same reaction should be treated with caution because the economics of overbuild monetization are asymmetrical. Selling excess compute is value accretive only when external pricing clears the fully loaded cost of power, depreciation, networking, site operations, customer support, financing, and opportunity cost. It is less value accretive if Meta is forced to monetize capacity during a broader industry oversupply period when AWS, Microsoft, Google, Oracle, CoreWeave, Nebius, Crusoe, and others are also trying to fill GPU clusters. The cloud option is therefore a downside buffer, not a full substitute for internal AI ROIC. Its importance rises in the bear case and falls in the bull case. In the bull case, Meta’s own products consume the compute because internal applications generate superior returns. In the bear case, external compute sales become a mechanism for recovering some capital cost, but pricing and margins may be weaker precisely when the option is most needed. MARKET STRUCTURE: LARGE, FAST-GROWING, BUT ALREADY INTENSELY COMPETITIVE The demand backdrop is favorable. Synergy Research estimated Q1 2026 cloud infrastructure service revenue at $128.6 billion, up 35% year over year, with a trailing 12-month revenue base of $455 billion. AWS, Microsoft, and Google held Q1 global cloud infrastructure shares of 28%, 21%, and 14%, respectively, and the top 3 providers accounted for 67% of public cloud. AI is explicitly changing market dynamics, with high-growth 2nd-tier providers including CoreWeave, OpenAI, Oracle, Crusoe, Nebius, Anthropic, and ByteDance. This supports the view that there is real demand for AI infrastructure capacity and model access. It also reinforces that Meta would be entering a market where incumbents already have massive scale, deep customer relationships, and rapidly expanding AI offerings. (Synergy Research Group) The competitive benchmark is high. AWS generated Q1 2026 segment sales of $37.6 billion, up 28%, with $14.2 billion of operating income, implying a 37.8% segment operating margin. Microsoft’s Azure and other cloud services grew 40% in fiscal Q3 2026, and management stated that customer demand continues to exceed supply, while also acknowledging that AI infrastructure investment is pressuring gross margin. Alphabet’s Google Cloud revenue grew 63% to $20.0 billion in Q1 2026, with $6.6 billion of operating income, implying a 33.0% operating margin, and cloud backlog nearly doubled sequentially to more than $460 billion. Oracle’s Q4 FY2026 cloud infrastructure revenue grew 93% to $5.8 billion, while remaining performance obligations reached $638 billion, but Oracle also reported negative $23.7 billion of FY2026 free cash flow as it continued investing behind OCI growth. These data points show both sides of the AI infrastructure opportunity: enormous demand and strong revenue growth, but heavy capital absorption and margin pressure even for established providers. (Amazon) CoreWeave is a particularly relevant comp for the “raw compute” version of Meta’s potential strategy. In Q1 2026, CoreWeave reported $2.08 billion of revenue, $99.4 billion of revenue backlog, more than 1 GW of active power, and more than 3.5 GW of contracted power. It also disclosed multiple new agreements with Meta, including a new $21 billion commitment signed in March. However, CoreWeave’s adjusted EBITDA margin was 56%, while adjusted operating margin was only 1% and net loss margin was negative 36%, reflecting the burden of depreciation, interest expense, scaling costs, and operating leverage that has not yet matured into GAAP profitability. This is directly relevant for Meta. GPU cloud revenue can look attractive at the EBITDA level, but the equity value depends on depreciation, financing, utilization, useful life, and terminal pricing, not simply gross demand. (SEC) PRODUCT STRATEGY: RAW COMPUTE, MODEL API, OR PLATFORM CLOUD The Bloomberg report implies 2 potential commercial products: raw AI compute capacity and access to hosted AI models. These have very different economics. Raw compute is faster to commercialize because customers primarily need GPU clusters, high-speed networking, scheduling, storage adjacency, and acceptable reliability. It can be sold to AI labs, enterprises, sovereign buyers, research institutions, and developers that are constrained by GPU availability. The strategic logic is straightforward: Meta has built or contracted massive capacity, customers are asking for access, and selling excess capacity at a premium can mitigate stranded-asset risk. The strategic weakness is equally clear: raw compute is the most commodity-like layer of the AI stack. It is exposed to spot-market pricing, rapid GPU obsolescence, power-cost variation, cluster utilization swings, customer concentration, and the risk that new architectures lower inference cost per token faster than legacy capacity can be depreciated. Model API access is more strategically compelling because it can attach software economics to infrastructure. Meta’s Muse Spark is relevant here. Meta described Muse Spark as the 1st model in the Muse family from Meta Superintelligence Labs, a natively multimodal reasoning model with tool use, visual chain-of-thought support, and multi-agent orchestration. Meta also stated that Muse Spark is available through Meta AI and that a private API preview is being opened to select users. A successful API business could monetize inference on a token basis, create a developer ecosystem, and improve utilization of Meta’s inference infrastructure. It could also give Meta a way to capture value from proprietary models after years of using Llama to shape the open-source ecosystem. The risk is that model API businesses are extremely benchmark- and trust-sensitive. Enterprise developers select models based on performance, latency, reliability, safety, context length, tool use, cost per token, governance, and integration with existing data and application stacks. Meta’s consumer distribution does not automatically translate into enterprise API adoption. (Meta AI) A broader Bedrock-like model marketplace would be the most ambitious path and the hardest to execute. AWS Bedrock provides a serverless way to use multiple foundation models through a unified API, with features around RAG, model evaluation, guardrails, security, compliance, and private connectivity. Microsoft Foundry Models provides a model catalog spanning Microsoft, OpenAI, DeepSeek, Hugging Face, Meta, and others, with tools for evaluation, fine-tuning, observability, responsible AI, enterprise SLAs for certain models, and Azure integration. Google’s Model Garden similarly provides a place to discover, customize, and deploy Google, open, and third-party models. Meta could theoretically replicate pieces of this architecture, but the value of these incumbent platforms is not just model hosting. The value is the surrounding enterprise fabric: identity, data access, compliance, observability, billing, procurement, private networking, professional services, and integration into existing enterprise cloud footprints. A Meta model-hosting product without this enterprise fabric would be a narrower developer service, not a full hyperscaler substitute. (Amazon Web Services, Inc.) STRATEGIC FIT WITH META’S CORE BUSINESS The strongest strategic argument for Meta entering AI compute and model APIs is not that it can displace AWS, Azure, or Google Cloud. The stronger argument is that Meta’s internal AI stack has reached a scale where external monetization can be layered onto an existing asset base. Meta already operates huge training and inference workloads for ranking, ads, recommendations, Reels, business messaging, and AI assistants. Its Family of Apps generated $55.91 billion of Q1 2026 revenue and $26.90 billion of operating income, implying a 48.1% segment operating margin. That business provides a massive internal demand anchor, which is structurally different from a pure neocloud business that must externally sell capacity to survive. The internal anchor reduces utilization risk if Meta’s AI products continue to scale. It also provides a demand prioritization problem: third-party revenue is attractive only if it does not crowd out higher-return internal workloads. (Meta) The most natural commercialization path may therefore be “AI tools for businesses on Meta,” not generic cloud infrastructure. Meta has direct relationships with millions of advertisers and businesses through Facebook, Instagram, WhatsApp, Messenger, and Threads. AI agents for ad creation, customer service, catalog management, lead qualification, messaging automation, content generation, and performance optimization are more strategically aligned than selling undifferentiated GPU hours. In that model, cloud infrastructure revenue may not be reported as cloud at all; it may appear as higher ad pricing, better conversion, higher messaging revenue, AI business subscriptions, or usage-based AI features embedded in Meta’s existing surfaces. This would likely carry stronger strategic defensibility than competing head-to-head for generic enterprise cloud workloads, because Meta’s proprietary social graph, ad data, business messaging surfaces, and distribution are differentiated. The article’s reference to selling access to “excess” compute is therefore important. It suggests that Meta’s hierarchy of uses remains internal first and external second. That is rational. If an incremental GPU can improve ad ranking, conversion modeling, video recommendation, content generation, agent quality, or consumer retention, the internal ROI may exceed external rental economics. External sales become most compelling when capacity is temporarily idle, mismatched to internal workload timing, or built ahead of demand. The strategy should be viewed as dynamic yield management for AI infrastructure. The ideal outcome is high internal utilization with a residual external marketplace that monetizes spare capacity at attractive rates. The weaker outcome is a forced external push because internal product monetization disappoints. ECONOMIC MODEL AND RETURN THRESHOLDS The economics must be framed around utilization and capital recovery. At the midpoint of Meta’s 2026 capex guide, capex would be $135 billion, up $62.78 billion from 2025. Offsetting that incremental capex through external cloud profit would require enormous scale. At a 30% operating margin, $62.78 billion of annual operating profit would require roughly $209 billion of annual external revenue, which is not a credible near-term target. A more realistic framework is that third-party monetization could offset depreciation and operating costs on excess capacity rather than fully offset capex. For example, if a $60 billion subset of AI compute assets were depreciated over 5 years, annual depreciation would be roughly $12 billion before power, support, networking, and financing costs. At a 30% operating margin, $12 billion of operating profit would require $40 billion of annual revenue. That scale is still very large, but it is more plausible over a multi-year period if demand remains supply-constrained and Meta can secure long-duration customers. The key difference between accounting profit and economic profit is silicon useful life. Goldman Sachs has noted that AI silicon turns over on much shorter cycles than data center buildings or power infrastructure, making replacement cadence a critical assumption in AI buildout economics. This point is central to Meta’s potential cloud strategy. If GPUs retain economic value for 5-6 years through an inference “value cascade,” Meta has more time to monetize excess capacity. If effective useful life is closer to 2-3 years for leading-edge workloads, then external pricing must recover capital much faster. A raw compute business is particularly sensitive to this assumption because customers will demand lower prices for older GPUs as new architectures improve cost per token. A model API business can partially abstract the hardware generation from the customer, but Meta still bears the underlying cost curve. (Goldman Sachs) The margin comparison is instructive. AWS and Google Cloud are mature platforms with broad service portfolios and Q1 operating margins of 37.8% and 33.0%, respectively. CoreWeave, despite extreme growth and a 56% adjusted EBITDA margin, reported only 1% adjusted operating margin and a negative 36% net loss margin in Q1 2026. This suggests that Meta should not be awarded mature cloud margins on any prospective raw compute revenue. The appropriate underwriting range depends on product mix. Raw reserved GPU capacity deserves a lower margin and lower multiple. Model API revenue deserves a higher margin and higher multiple if performance, developer adoption, and retention are strong. Business-agent revenue embedded into Meta’s ad and messaging stack may deserve the highest strategic value because it can reinforce the existing Family of Apps flywheel rather than create a stand-alone infrastructure P&L. (Amazon)

• 
  Manny L. (@mancan76) reported 4 minutes ago

  @JoshRing4UBI @Colteastwood The problem here, is the way digital rights work - you own nothing. While most people haven’t been affected by this yet, one day, that will change… There is no physical market to control prices, so Sony can charge whatever they want, and you can’t head to Amazon or eBay

• 
  sephtiger | Unshadowbanned! (@sephtiger) reported 6 minutes ago

  @AbdallahNATION preordered a physical copy from Amazon for $49.99, skill issue

• 
  THE FINAL COUNTDOWN (@THIS_TIME_X) reported 7 minutes ago

  @Indiana_TJ THIS IS PART OF THE PROBLEM WITH @amazon AND OTHER DELIVERY BASED CONPANIES ANYONE CAN DELIVER THESE DAYS ITS LIKE UBER, YOU NEVER KNOW WHO IS GOING TO PICK YOU UP

• 
  Sherry White (@sherrybordeaux) reported 8 minutes ago

  @Lefty_Kitty @PieterVand37940 There are also Amazon lockers in some cities that you can have items delivered to. This would solve the problem of no permanent address.

• 
  MegaNuggs (@McNuggs5000) reported 10 minutes ago

  @DanNeidle Ooooooorrrrrr. And heres a crazy thought. He could fix the tax avoidance bullshit pulled by Amazon, Costa et all...

• 
  Halmunde (@Halalmin) reported 10 minutes ago

  @chrisacko13 @thehonestletter We already have tax years. Amazon already avoids tax in the UK that way by setting up in Brussels and is part of the problem. Stock markets determine stock value. Estate agents can look at land/property. There aren't new things people aren't already doing. We have dual tax

• 
  hakimuddin burhani (@hakimuddinburh1) reported 13 minutes ago

  @AmazonHelp Idont want any assistance from you. I already deleted AMAZON.. I KNOW YOU DONT CAPABLE TO SORT OUT ANY PROBLEM

• 
  Emeka (@OInnocxnt) reported 15 minutes ago

  They are not the same. Amazon built AWS because a real internal need existed first. The infrastructure solved a problem Amazon already had, and it turned out thousands of other companies had the same problem. Demand brought AWS into existence. Meta is building compute on a bet that demand for its own LLMs will show up later. Nobody is building or asking for Llama the way early customers were asking to rent Amazon’s infrastructure. That is supply looking for a demand story, not the other way around.

• 
  Ramakanth (@xramakanth) reported 16 minutes ago

  Uber's able to handle massive traffic because they optimize their software for the underlying hardware. Mechanical sympathy is the practice of creating software that's aware of its hardware, using principles like predictable memory access and cache line awareness. For example, Netflix uses this to stream smoothly, Amazon optimizes its database queries, and Google does it to speed up its search results. To identify if you need this, check if your app is slow despite having powerful hardware. If so, it's time to apply mechanical sympathy principles.

• 
  bibi — aaron girl’s (@jhyrubyes) reported 16 minutes ago

  @AmazonHelp Hello! I need help with my Kindle Paperwhite Kids. It has become very slow and is showing severe screen ghosting. I can’t contact Amazon through chat or phone because of my region. Could you please help me with the warranty process? 😭😭😭

• 
  Sandy M.🇺🇸 🖖🌻 (@SandyMcInturff1) reported 17 minutes ago

  @BladeoftheS @AveryBa68752542 The warehouse where my son works, not Amazon, does have AC. But it will heat up in this extreme heat anyways. I gave him the heat exhaustion run down, just in case.

• 
  Melanie 🐍 (@MDJO) reported 18 minutes ago

  @weneedsunlight @NoFarmsNoFoods Are you a Farmer? Then extract yourself from a biz model that is at the behest of government , whose sole objective is to shut you down. Are you a consumer? Buy direct from farmers (and no they don’t do Amazon deliveries so it’s probably gonna be a bit inconvenient) Otherwise…

• 
  deputydog357 (@deputydogblitzn) reported 20 minutes ago

  @AngieNixon The only way to make things affordable is mass deportations and breaking up monopolies, Amazon, big insurance, corporate healthcare systems and big tech, unfortunately you nor anyone else in Congress will do their job to bring down the monopolies, lobbyists pay all u off

• 
  Anand Karthikeyan (@ak_2wter) reported 22 minutes ago

  @AmazonHelp Its been 11 pm and your delivery agent marked as expected delivery status. This has out for delivery by 5pm today and he is here in hsr layout only. But not delivered why??? Why the f** this always happens?? I have raised 200 times for the same issue. Amazon is really worst.

• 
  Abhishek Singh (@0xlelouch_) reported 22 minutes ago

  The interviewer asked me to design Amazon order tracking. I fumbled. I jumped into Kafka topics before asking what tracking even meant. So I restarted with requirements. Who’s the user: customer, seller, support, carrier ops? Read path dominates. Needs a timeline view, current status, and where-is-my-package accuracy. Updates are eventually consistent, but the status must never go backwards. GDPR deletes for customer data. Audit trail for disputes. Latency: sub-second reads, writes can lag. Then I framed the surface area as APIs. GET /orders/{orderId}/tracking returns currentStatus + ordered event timeline. GET /shipments/{shipmentId}/tracking for multi-box orders. POST /carriers/{carrier}/events for inbound scans (idempotent). Support-only: GET /tracking/{orderId}?includeRawEvents=true. Data model next. Order -> Shipment(s) -> TrackingEvent(s). TrackingEvent is append-only: (shipmentId, eventId, occurredAt, receivedAt, status, location, source, rawPayloadHash). CurrentShipmentState is a derived row: lastStatus, lastLocation, lastUpdatedAt, version. Enforce monotonic transitions with a state machine and version checks. Architecture: ingest carrier events into a durable log, validate/dedupe, then write to the event store. A projector builds the read model for fast queries. The tracking page hits the read model, not the write path. If I need search by tracking number, that’s a separate index keyed by carrierTrackingId -> shipmentId. Scaling: partition by shipmentId or carrierTrackingId to keep ordering per shipment. Hot partitions happen on big launch days, so shard keys need entropy (hash prefix). Cache current status in Redis with short TTL, but never cache raw timelines too aggressively because customers spam refresh during delays. Reads can go to replicas. Writes can be buffered, but the log must be the source of truth. Tradeoffs I called out: strict ordering vs throughput. If you require global order, you pay. I only need per-shipment order. Strong consistency on reads is expensive; eventual is fine if we show lastUpdatedAt and avoid status regressions. Storing raw payload helps debugging, but it’s a compliance and cost problem; keep it in cold storage with retention. Failure cases: duplicate scans, out-of-order events, missing events, carrier downtime, retry storms from bad webhooks, poisoned messages, projector lag, and a bad deploy that changes the state machine. I’d add idempotency keys, dead-letter queues, backpressure, and an admin tool to replay from the log and rebuild the read model without touching the source events.

• 
  Matt Edmundson (@mattedmundson) reported 35 minutes ago

  Walmart is far less competitive than Amazon. Higher margins too. Most sellers just don't know it yet because they copied their Amazon listings across, it didn't work, and they gave up. That's not a platform problem. That's a setup problem.

• 
  Greig. (@Scot_Gaijin) reported 36 minutes ago

  @ROS5IHD @PlayStation The fact that PSN store charges double for titles that are physical on Amazon is a problem.

• 
  Scymet  (@scymetricall) reported 37 minutes ago

  @NoctreSharp @smartmuki @notgrubles MS owns minecraft, that's a totally different issue and indeed they do it for the free work. I'm talking about Google, Netflix, Amazon, and Microsoft taking control of already open source tech once it becomes popular. They vastly outwork the free workers in these.

• 
  Udaya Singh (@udayah_veracity) reported 40 minutes ago

  @godrejsecure @godrejsecure @GodrejLocksS I have not got any call back. Pls check my mail n initiate the refund immediately or let me know where to escalate this issue so that I can get a refund. I need to buy a new lock urgently. Also mention in Amazon that your locks won't work with JIO

• 
  Ray Sharradh (@RaySharradh) reported 40 minutes ago

  Is @amazon bricking first gen Alexas? Widespread outages reported on Downdetector over the past 2-3 days. My old Echo Dots are still working fine. Anybody else with first gen Alexas having issues? @AmazonAlexa

• 
  Rohit Ahuja (@rohit_ah) reported 41 minutes ago

  In 2014, US debt was $18 trillion. Today it's $39 trillion. Watch what that money did on its way through the system. It's a story about what happens to every asset the moment the pool of dollars chasing a home gets bigger, and bigger, and bigger. Follow the train. First stop, the frontier nobody respectable would touch. 2017. Crypto was a curiosity, a magic internet coin your nephew wouldn't shut up about. Then the liquidity found it. Bitcoin, Ethereum, the entire complex went up 30x in a single year. The most speculative, longest-duration, hardest-to-value asset on earth got repriced first, because that's always where the newest money goes hunting for the biggest return. Next stop, quality at scale. As the debt kept climbing through 2019, 2020, 2021, the same force rolled up the risk curve into the safest growth on the planet. Apple, Microsoft, Nvidia, Amazon, the Magnificent Seven. A handful of companies absorbed trillions in market cap and came to define the index itself. The biggest pools of capital in the world deciding that scarce, dominant, compounding tech was where the debt-fuelled liquidity belonged. Current stop, the new frontier. AI and space. The train has pulled into the station where a single company can add the GDP of a mid-sized country to its market cap on a keynote. Where compute buildout is measured in hundreds of billions. Where rockets and models get funded at valuations that would have been laughed out of the room in 2014. Same mechanism. New frontier. The money got bigger, so the ambition got bigger. There was exactly one moment this decade when the train braked. 2022. Rates went up, liquidity got pulled, and every single asset on this train crashed together, crypto down 64%, the Nasdaq down 33%, in the same year debt was still rising. That wasn't a failure of the thesis. That was the proof of it. When the train slows, everything on it stops. When it accelerates again, and it did, everything reprices higher. As long as the debt train keeps accelerating, get ready for new booms to keep forming. The liquidity doesn't disappear, it rotates, from crypto to Mag7 to AI to whatever frontier is next, space, robotics, fusion, take your pick. Each stop looks like a bubble. Each stop also looks like it did right before it tripled. The sustainable bust everyone keeps predicting does not arrive while the train is still moving. It arrives when the train stops. Not when valuations get high. Not when it feels absurd. When the debt stops accelerating. Watch the train, not the ticker. What's the next station, space, robotics, or something not on the radar yet? Drop it below.

• 
  Bryce Joe (@yobrycejoe) reported 41 minutes ago

  The fastest way to sell on Amazon is to say the exact problem they typed into the search bar. Lead with “it leaks” or “it hurts your hands.” Then show their day, the fix, and proof. Comment “message” and I’ll send you a resource on this. What problem do you lead with now?

• 
  Sir Smidgeon 🇻🇦 (@SirSmidgeon) reported 41 minutes ago

  @AmazonHelp That's quite insulting. How about you fix your app instead?

• 
  Aaron Cordovez (@Aaron_Cordovez) reported 41 minutes ago

  On Amazon, fraudulent and fake trademarks are used as a war to take down honest sellers. They hide with FAKE names, and submit "anonymous" reports and have full legal protection (You can't sue in China from the U.S. for less than $200k). In U.S. Law, if you enforce a Trademark or other Intellectual property WRONGFULLY, you are legally liable and can be counter-sued. However, Amazon currently allows for random, unknown and not verified owners of "Intellectual property" register and attack your listing. This needs to be a basic reform. Intellectual Property claims must be done by the VERIFIED OWNER and there needs to be someone IN THE US that can be held accountable for false claims. @amznsellerhelp @AmazonHelp Please confirm ownership of IP before taking down major Amazon products.

• 
  anthony (@ant9179) reported 42 minutes ago

  @PGC1a_RB @Nomaan_2003 @SSavson Stopper buying their mag last year when someone on here broke down how bunk theirs was, wonder if my amazon purchases qualify for the settlement, everything from them is expected to be less potent to me but price lets me just use more

• 
  johan1093 (@johanhappyhouse) reported 44 minutes ago

  @PUMACare Am I talking to a bot: As mentioned earlier, have you contacted Amazon? Please get back to us if you haven’t been able to resolve the issue with our retailer, and we’ll examine your shoes to determine whether this is indeed a manufacturing fault or simply normal wear and tear.

• 
  आनंद गुप्ता (@aanandsg) reported 47 minutes ago

  @amazon has become worst to work with. Amazon is harrassing employees with micro management. Downfall of Amazon started with managers putting pressures to fix their performance every month or week or they will put on a plan. #MentalHealth of employees have started deteriorate. Please take action sooner @ajassy

• 
  Adam Wren (@aswren) reported 47 minutes ago

  Old friend of mine from school has never had a real job, but he’ll diligently set an alarm & wake up at 7am to deliver amazon packages on a GTA V RP server

• 
  Normal Guy (@Normal_2610) reported 50 minutes ago

  GST law says exporters get 90% of refund within 7 days, 9 years in, Deloitte surveyed 1096 executives and 77% still named refund delays their worst problem Cash that should flow back to factories sits parked in government accounts while exporters borrow at 10% to keep shipping. Every month of delay adds cost that no tariff schedule captures. India wants to grow exports but nobody in policy seems bothered that tax office is slowing them down Supreme Court already ruled that GST refund on capital goods is not right Parliament owes, and Rule 89 blocks any input tax credit on machinery from refund calculation. Manufacturer buying equipment to make goods for export gets zero back on that purchase. Same manufacturer importing finished parts for assembly gets full credit. Policy ends up pushing exporters toward assembly instead of manufacturing from scratch. Vietnam refunds across entire production chain including plant and machinery and that matters when companies decide where to put their next facility. E-commerce platforms collect 1% TCS on every sale, meaning small exporters on Amazon or Flipkart see cash deducted before they even get paid. Credit comes back only after filing returns and reconciling portal data, which for small seller takes months. Then add mandatory GST registration regardless of turnover for anyone selling through platforms. System was designed for large firms and got applied wholesale to sellers running businesses from their phones. Small exporters spend more time on compliance than on finding new buyers abroad Well new Startup idea Who can solve this Issue Charge Fee :)