Amazon Web Services status: access issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Amazon Web Services 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 Web Services users through our website.

• Website Down (71%)
• Sign in (14%)
• Errors (14%)

Live Outage Map

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

City	Problem Type	Report Time
Boca da Mata	Errors	3 days ago
Township of Evan	Website Down	17 days ago
New York City	Website Down	20 days ago
Ciudad Jardín	Website Down	1 month ago
Kyiv	Sign in	3 months ago
Chennai	Website Down	3 months 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 Web Services Issues Reports

Latest outage, problems and issue reports in social media:

• 
  TAPE Vector (@Tape_Vector) reported 9 minutes ago

  JPP-KY $5284.TW is not an AI chip company. It makes the precision metal infrastructure that surrounds the chips, power systems and cooling hardware inside modern AI servers. That distinction matters. JPP Holding designs and manufactures precision metal mechanical parts, enclosures, cabinets and structural components. Its products are used across: AI server racks Server chassis Power supply housings Battery backup unit enclosures Liquid cooling components CDU and manifold structures Telecom equipment Aerospace avionics Aircraft structural and cabin parts Medical equipment Industrial systems The company is headquartered through a Cayman holding structure and listed in Taiwan, but much of the manufacturing engine sits in Thailand through Jinpao Precision Industry. That Thailand base is important. JPP is positioning itself between Taiwanese and global technology customers that increasingly want manufacturing capacity outside China. The operating model is high mix precision manufacturing rather than mass production of one standardized component. A customer brings JPP a mechanical design or performance requirement. JPP can then handle several steps internally: Engineering and design support Metal cutting Stamping CNC machining Sheet metal forming Welding Surface treatment Painting Assembly Inspection Final integration That means the company can take a customer from drawing to finished enclosure instead of supplying only one small step. For AI servers, this can include the physical rack or chassis holding compute hardware, power equipment and cooling systems. For aerospace, it can include avionics housings, structural parts and cabin components that require much tighter certification and process control. This combination is unusual. AI infrastructure gives JPP growth. Aerospace gives it another technically demanding end market with different cycles. The company describes this model as a mix of European engineering capability and Thai manufacturing. The phrase used by management has been: French brain. Thai heart. That comes from the European aerospace companies JPP acquired and integrated with its Thailand manufacturing base. The aerospace side matters because the qualification barriers are much higher than ordinary sheet metal fabrication. JPP has Nadcap certified processes and has worked within the European aerospace supply chain. Company materials and industry reporting have referenced customers and programs connected to Airbus, Thales and Safran. Those relationships do not automatically mean every JPP aerospace product goes directly into those companies. But they show that the manufacturing system has passed qualification standards far above normal commodity metal fabrication. Then AI arrived. This has changed the financial profile of the company very quickly. FY2024 revenue was approximately NT$2.39 billion. FY2025 revenue jumped to about NT$3.73 billion. That is roughly 56% growth. Net income reached approximately NT$618 million. EPS reached NT$12.05. Gross margin stayed around 37.8%. That margin is one of the numbers I find most interesting. JPP did not double its business by becoming a low margin commodity manufacturer. The company expanded rapidly while keeping gross margin in the high 30% range. That suggests the current product mix still carries meaningful engineering and manufacturing value. Q1 2026 continued the trend. Revenue reached approximately NT$1.17 billion. That was about 45% higher year over year. Gross margin remained around 37.5%. So the 2025 acceleration did not immediately reverse once the calendar changed. This is now a real operating ramp. The AI server side has become the main growth engine. JPP manufactures server racks, chassis, power enclosures and increasingly components associated with liquid cooling. That last category matters. AI servers are becoming more difficult to cool. Higher power GPUs produce more heat. More compute density means more thermal load inside each rack. That is pushing the data center industry toward larger cooling distribution systems, manifolds, cold plates and liquid cooling infrastructure. JPP does not manufacture the GPU or the cooling technology itself. It manufactures some of the metal structures and precision components that allow those systems to be installed inside the rack. That places the company several layers beneath the visible AI names. $NVDA and $AMD create demand for increasingly powerful accelerators. Those accelerators require more complex server systems. $DELL and $SMCI integrate servers and racks around those accelerators. $VRT and $ETN operate in the power and cooling infrastructure around the data center. JPP sits further inside the physical manufacturing chain. It produces some of the metal cabinets, chassis, housings and structural components required by this infrastructure. These are ecosystem comparisons. They are not all disclosed customer relationships. The most interesting potential US connection is the company's major cloud customer. Management commentary and Taiwan reporting have repeatedly described a major US cloud service provider as one of JPP's largest AI customers. That customer has widely been reported as Amazon AWS. If correct, that creates an indirect connection to $AMZN. But I would keep the wording disciplined. JPP has not provided enough English primary disclosure for me to treat the identity and exact revenue contribution as completely settled. The important hard fact is that a major US CSP has become a very large customer. Recent commentary has indicated that this customer may account for roughly 30% of revenue during parts of the AI ramp. That is both the opportunity and the risk. A customer that large can transform a small supplier. It can also transform the income statement in the opposite direction if orders slow. Another major relationship is in Thailand. JPP has been expanding production around a large power and server customer widely identified as Delta Electronics Thailand. That customer makes power supplies, thermal systems, data center equipment and related electronics. The geographical relationship matters because both companies operate major production facilities in Thailand. Shorter logistics. Faster delivery. Closer engineering cooperation. Just in time production. Dedicated manufacturing capacity. Those factors can make a supplier harder to replace once a large program is running. But they also deepen customer concentration. JPP is effectively investing ahead of these customers. The company has been adding production capacity in Thailand. One important bottleneck has been painting and surface treatment. JPP is expanding automated paint capacity. It is also investing in larger stamping capacity and dedicated production areas for AI server and power related products. The logic is simple. More AI server racks require more metal structures. More power density requires more sophisticated power housings. Liquid cooling adds additional structural parts. If JPP remains qualified inside those programs, each generation of AI infrastructure can increase the content opportunity per rack. That is the bull side. The risk is that the company adds capacity for demand that later slows. AI infrastructure spending is strong now. It will not grow in a straight line forever. A hyperscaler can change server architecture. An ODM can move a program. A customer can dual source. A competitor can cut price. If one large customer represents 25% to 30% or more of revenue, those decisions matter immediately. That is why I want the exact customer concentration table from the latest annual report. The aerospace business gives JPP some diversification. Before the AI acceleration, aerospace represented a much larger part of the company. That business went through a difficult period around the pandemic and the following aerospace supply chain disruption. It has been recovering. The company has continued obtaining certifications and expanding its European aerospace capabilities. That creates a useful second engine. AI server demand is fast and capital intensive. Aerospace is slower, qualification heavy and built around longer product cycles. The two businesses have different risks. Together they can potentially produce a more balanced manufacturing platform. But right now AI is clearly driving the growth rate. The financial question from here is not whether revenue can grow. It already has. The question is whether the current margins survive the next stage of scale. High 30% gross margins are strong for a precision metal manufacturer. I want to know how much of that comes from: AI server racks Power enclosures Liquid cooling components Aerospace Specialty low volume work New customer programs I also want the operating cash flow behind the reported earnings. Fast manufacturing growth consumes working capital. More orders require more raw material. More capacity requires more equipment. More inventory sits between production and customer delivery. Receivables rise. So a company can report excellent earnings while cash is being absorbed into expansion. That is not automatically bad. But the return on that capital has to remain high. JPP ended 2025 with roughly NT$7.4 billion in assets and around NT$3.7 billion in equity. The balance sheet does not currently look distressed. There is no obvious heavy dilution story. The primary capital allocation issue is expansion. Paint lines. Stamping equipment. Factory capacity. Dedicated customer production. Those investments are being made because demand already exists. Now they need to earn acceptable returns. For US market context, I see several useful layers. $NVDA and $AMD are demand drivers. More accelerator shipments can mean more server racks, more power density and more cooling hardware. $DELL and $SMCI represent the server integration layer. They assemble computing systems around GPUs, networking, storage and power. $VRT and $ETN represent the data center power and thermal infrastructure layer. $ANET sits in the networking layer connecting increasingly large AI clusters. $AMZN is relevant because AWS is widely reported as the major US CSP associated with JPP's AI server business. Again, I would treat that specific customer identity as reported rather than fully disclosed until the primary customer note confirms it. The aerospace familiarity is different. $BA is the obvious US listed aerospace reference. JPP is not primarily a Boeing supplier story. Its known aerospace footprint is more European. But the same qualification logic applies. Aircraft components require traceability, process control and long certification cycles. That experience can strengthen the overall manufacturing discipline of the company even when the fastest growth is coming from AI infrastructure. This is what makes $5284.TW more interesting than a generic sheet metal company. The metal itself is not scarce. The capability stack can be. A customer needs a supplier that can: Meet tolerances. Pass qualification. Build tooling. Handle design changes. Scale capacity. Deliver consistently. Maintain surface quality. Control welding and assembly. Locate production close to the customer. And do it without disrupting a multibillion dollar server or aerospace program. That creates switching friction. It does not create an unbreakable moat. Large customers still have enormous negotiating power. The company remains small relative to the customers it serves. That means the power relationship still favors the customer. The current strengths are clear. 2025 revenue grew about 56%. EPS reached NT$12.05. Gross margin remained near 38%. Q1 2026 revenue grew another 45%. AI server exposure is already producing real revenue. Liquid cooling adds another content opportunity. Thailand capacity is expanding. Aerospace is recovering. The balance sheet is supporting expansion without obvious distressed financing. The risks are also clear. Customer concentration is high. The largest AI programs are project driven. Formal long term volume commitments are not well disclosed. The company is investing heavily into capacity during an AI spending boom. Margins could compress as volume rises. Aerospace recovery could stall. And the current growth rate depends heavily on continued data center capital spending. For me, the next proof is not another monthly revenue record. I want to see: Exact top customer concentration. How much revenue now comes from AI server products. How much comes from liquid cooling. Whether the major CSP relationship is widening into additional products. Whether the large Thai power customer is gaining share of revenue. Utilization of the new painting and stamping capacity. Operating cash flow after expansion capex. Return on invested capital from the Thailand buildout. Aerospace revenue and margin recovery. Whether gross margin can remain above the mid 30% range as the company scales. Real manufacturing. Real AI infrastructure exposure. Real earnings growth. Real high margin execution so far. But also real concentration risk. jpp-KY $5284.TW does not need to invent the next GPU. It needs to remain the qualified company manufacturing the physical structures around the companies that do. If AI racks become larger, hotter and more complex while JPP keeps winning more content per system, the opportunity can grow much faster than the underlying server unit count. The question now is whether that position is durable enough to survive the inevitable cooling of the AI capital spending cycle. That is what I want to understand next. My investing journal, not financial advice.

• 
  D (@D8118199174282) reported 14 minutes ago

  @awscloud No, you still haven’t compensated users for the cost bug error

• 
  golden gate bridge (@_ggbridge) reported 18 minutes ago

  @litteralyme0 this happens to me everytime Amazon AWS has an outage

• 
  Pixi Dust (@Bethco2257) reported 19 minutes ago

  @FmrRepMTG @amazon @awscloud walk into any Amazon office building at Seattle HQ and South Lake Union Amazon Seattle offices and instantly you can see with your own eyes… nearly 80% of staff are HB1 visa holders!!!! So yes, we have a big problem.

• 
  GettingMyGlitchOff (@GettingMyGlitch) reported 20 minutes ago

  @AWSSupport Yeah its just getting automated responses to everyone saying "Your account will not be re-opened. We regret that we have not been able to address your concerns to your satisfaction." so can you send the mturk team a message to review the error since its more then just me its

• 
  barney (@Southclaws) reported 22 minutes ago

  @jonny_castles @awscloud I am seeing this too, ECR and S3 in eu-west-1, lots of "service unavailable" errors

• 
  Ayush Awasthi (@Ayush_awa) reported 27 minutes ago

  @AWSSupport Its the third day today. The only mail I received from aws was all about “We are getting someone to investigate this” Since then I havent received any update. My only issue is that these things are time sensitive and you cant just take days after days with no real status updates

• 
  Keith Townsend (@CTOAdvisor) reported 27 minutes ago

  I spent over $120 in @awscloud so far this month. Most of it has been on GPU instances and Xeon 6. This is the most I've spent since I've had a long-running VM as a web app server a few years ago.

• 
  Abdulkadir | Cybersecurity (@cyber_razz) reported 30 minutes ago

  ClarityCheck markets itself as a tool to detect catfishing. To verify identities. To keep you safe from liars online. Here's how it works. You upload someone's photo. ClarityCheck runs it through their reverse image search. Finds their dating profile. Their social media. Wherever that face appears. Seems solid. Except ClarityCheck just leaked 9 million photos sitting in an unsecured Amazon S3 bucket. 450 gigabytes of faces. Stored in folders labeled "faces" and "profiles." The photos came from people uploading images of strangers. Dating app screenshots. Private social accounts. Scanned prints. Photos of children. Most of these people never uploaded anything to ClarityCheck. They just got identified by someone else trying to figure out who they were. Then their face got indexed. Stored. And left wide open. For several months. The URL to access it was embedded in ClarityCheck's own website source code. The company's response: this was "temporary storage" and an "ordinary member of the public" would not have found it. An ordinary member of the public with the URL from their own website. Which is not temporary storage. That's just a server. And it wasn't invisible. You'd need 30 seconds and a basic understanding of how websites work. Now ClarityCheck says the data is "secured." A service built to prevent identity fraud just exposed millions of identities. The tool designed to catch people lying about who they are just showed everyone's actual face to anyone listening. The irony isn't subtle. It's a design flaw pretending to be an accident.

• 
  ShenYubao (@ssybb1988) reported 30 minutes ago

  @AWSSupport @awscloud AWS account suspended for additional verification; Production services down for ~24h. All verification docs submitted. Unable to purchase Business Support+ due to suspension. Please expedite review & help restore production. Case ID: 178678971500932

• 
  TeeBeeCTO (@TeeBeeCTO) reported 30 minutes ago

  Setting up a new AWS account @awscloud … and get non stop errors on set up. Looks like their vibe coding dystopia worked… were all going to die because SkyNet will have error pages that end up giving us all death by error page. Amazing work… @JeffBezos !!

• 
  Nicholas Griffin (@ngriffin_uk) reported 32 minutes ago

  @QuinnyPig @awscloud that’s real? it looks like someone who doesnt know how to prompt made a sign in screen and didnt look at it.

• 
  Piragash Velummylum (@piragash) reported 32 minutes ago

  At @awscloud we call this undifferentiated heavy lifting. @Cometml Opik does the hard work, so you can focus on the customer problems.

• 
  Anjishnu Ganguly (@Anjishnu46) reported 38 minutes ago

  Hey @AWSSupport, Our Case 178452910500907 is still not resolved, and we've been left waiting without any meaningful update. Because of this, we're also unable to apply for AWS Activate, which is directly impacting our business operations and slowing down our workflow. This delay is becoming unacceptable. Please look into this urgently and help resolve it as soon as possible.

• 
  Matthew Joughin | 🏗️ Cross Platform Dev Tools (@mrwcjoughin) reported 38 minutes ago

  @jeffdafo @ivanburazin @awscloud Even aspnetcore can run on Linux now - there is no excuse to have (and pay through the nose) for Windows server licenses)

• 
  Adam Peterson (@_adam_here_) reported 41 minutes ago

  is @AWSSupport down?

• 
  Ashutosh Dubey (@_virusbug_) reported 42 minutes ago

  @AWSSupport I’ve already emailed the AWS Receivables team but haven’t received a reply yet, it’s been 3 days now. The support engineer only advised me to contact them about this issue.

• 
  Vinit Upadhyay (@vinitcodes) reported 43 minutes ago

  @AWSSupport @kirodotdev @awscloud I’ve already raised a case explaining the entire issue. Case ID: 178664512400070. Unfortunately, I still haven’t received any genuine help, which is really disappointing. I’m not the only one facing this issue-many others are experiencing the same problem. Please look into this.

• 
  jd42 (@jdonovan42) reported 44 minutes ago

  @DarioCpx Pls recalculate considering the massive clouds the AI revs sit on. Each $1 in AI attaches $1.5-2.5X attached cloud biz and your down to 30-45% of revs. Then as open wgt models gain apply the 100% revenue retention vs. 60% on frontier and look at EBITDA % from frontier is 10-15% max. But nice try ;) No doubt Anthropic and OpenAI both stimulate demand for Amazon AWS. It creates 2x the cloud biz than it does the direct AI biz. So why not count the full picture of things vs just the #'s that fit one narrative.

• 
  Amanda (@AMyrick1989) reported 47 minutes ago

  @krassenstein @Tesla Be careful. All these at the same time makes me think they are hacked, I’m fairly certain Amazon AWS was hacked that day when everything when down, given my Grok we hacked and no one ever gave explanation. I’m no expert but ya know, all these signs are pointing to this. Also, the Obamas helped produce a movie where someone hacked our satellites and alluded to it being the Middle East and all the teslas went haywire and self drive themselves to pile up on all the freeways. I wish I remembered the name however I made note of this terrifying movie given Obama clearly knows things we don’t. Just sayin, these incidents aren’t scattered. I would not be driving that ***** if I were you.

• 
  Mr Sage (@GreatSage_0x) reported 50 minutes ago

  Custos makes verdicts on AI agent transactions. Problem: it had no memory. Every decision started from zero. Meet Anamnesis — I gave it a memory layer using CockroachDB. Built for the @CockroachDB x @awscloud Hackathon

• 
  Onkar Deshpande (@i_am_onkar) reported 53 minutes ago

  @bilal_akh @awscloud what were potential issues? account compromised? keys exposed? last time this happened for me was when some keys got leaked and someone emptied my startup credits -this was 8years ago tho

• 
  Joy | 🔥 (@JoyKingdomx) reported 54 minutes ago

  As we all know, cloud outages are not a new topic here. But what is changed is how many products now depend on the same infrastructure. A few years ago, an AWS outage mostly affected AWS customers. Where today one infrastructure issue can ripple across AI apps, APIs, developer tools, streaming platforms and payment services. That is the downside of concentration. We have seen @awscloud outages impact millions. We have also seen @OpenAI outages pause production apps because entire inference pipelines depended on a single provider. It is not the outage that matters - it is the dependency behind it. That is the reason why decentralized compute makes sense. Instead of relying on a single cloud or region, @ionet distributes compute across a global network of GPUs in 138+ countries. No infrastructure is immune to failure, the goal is to ensure one failure does not become everyone's failure.

• 
  Dan Lynch (@danlynch) reported 59 minutes ago

  @QuinnyPig @awscloud I need to sign in with LEGO 😎

• 
  IDrabs (@drabs_i) reported 1 hour ago

  @PlayStation @awscloud Wtf you playing at fix the outage.

• 
  Ashutosh Dubey (@_virusbug_) reported 1 hour ago

  @AWSSupport I’ve already emailed the AWS Receivables team but haven’t received a reply yet, it’s been 3 days now. The support engineer only advised me to contact them about this issue. @AWSSupport

• 
  Gopal Goti (@gopal_goti) reported 1 hour ago

  @AWSSupport @AWSSupport Thank you for the update, but 'high volume' cannot justify a support case sitting completely 'Unassigned' for over 4 days for a basic account verification issue. This is not a complex technical issue. it is a backend system error on the AWS verification form.

• 
  M (@mdw864) reported 1 hour ago

  @awscloud can you PLEASE answer me? You’ve just turned off my Business because you will not help me. I need you to stop charging me. Do not tell me to login because I do not know the email address that was used to open the account. All I know is my bank account is being charged

• 
  🦋Kieran🦋 💻 (@sir_divs_alot) reported 1 hour ago

  @AWSSupport Thanks for your response and I've been keeping track for any updates very closely but the problem is these cases are either unassigned and if they are, no one is following up. I'm at the point of just giving up entirely cuz it's pointless waiting endlessly.

• 
  Brad (@beeradmoore) reported 1 hour ago

  @QuinnyPig @awscloud I made a thing that posts in Slack when a new AZ goes live. It has pipped up twice in two days. I thought that was odd, it’s sat silent for so long and now twice in two days! Turns out it’s broken and will just repeat until I tell it that yes I know about eu-west-2d