The October 20, 2025 AWS outage was a regional failure with worldwide symptoms—not a total shutdown of AWS or the internet. AWS traced the incident to a race condition in DynamoDB’s automated DNS system in US East (N. Virginia), us-east-1. The resulting endpoint failure disrupted Amazon properties, Prime Video, Fortnite’s login and related online services, Perplexity and many other applications before secondary EC2 and Network Load Balancer problems extended recovery into the afternoon.
Quick answer
- When: The outage began at 11:48 p.m. PDT on October 19, 2025, and affected users on October 20.
- Where: The primary failure was in AWS’s US East (N. Virginia) region,
us-east-1. - Cause: A race condition in DynamoDB’s automated DNS-management system created an empty DNS record for
dynamodb.us-east-1.amazonaws.com. Secondary EC2 and Network Load Balancer recovery failures extended the disruption. - What users saw: Amazon.com, Prime Video, Alexa, Fortnite’s login and related online services, Perplexity, Snapchat, Roblox, Reddit, Signal, financial services and other applications experienced outages or degraded performance.
- Was all of AWS down? No. This was a regional AWS incident that created worldwide symptoms because many services depended on
us-east-1directly or indirectly. - Was it a hack? AWS attributed the event to an internal software failure. There was no public evidence indicating a cyberattack.
- When did it end? AWS said the overall event ended at 2:20 p.m. PDT on October 20, with normal operations reported by Amazon at 3:01 p.m. Some backlogs still needed processing.
This was a resolved historical incident, not a current AWS outage alert. AWS’s final account is available in its post-event summary.
As an Amazon Associate I earn from qualifying purchases.
What happened to Fortnite, Amazon and other services?
The October 20, 2025 AWS outage produced an unusually broad set of consumer-facing symptoms. Amazon.com and Amazon subsidiaries were affected, while contemporary reports listed problems with Prime Video, Alexa, Fortnite, Perplexity, Snapchat, Roblox, Reddit, Signal, Venmo, Robinhood, Duolingo, Canva, Canvas and other services.
Those services did not necessarily fail in exactly the same way. A website might load while login, checkout, matchmaking, playback or background processing failed. An already-established game session might remain active while a new player could not authenticate or join a match. Recovery also varied by service because applications had different dependencies, caches, retry behavior and backlogs.
#1 Best Overall
- DUAL-BAND WIFI 6 ROUTER: Wi-Fi 6(802.11ax) technology achieves faster speeds, greater capacity and reduced network congestion compared to the previous gen. All WiFi routers require a separate modem. Dual-Band WiFi routers do not support the 6 GHz band.
- AX1800: Enjoy smoother and more stable streaming, gaming, downloading with 1.8 Gbps total bandwidth (up to 1200 Mbps on 5 GHz and up to 574 Mbps on 2.4 GHz). Performance varies by conditions, distance to devices, and obstacles such as walls.
- CONNECT MORE DEVICES: Wi-Fi 6 technology communicates more data to more devices simultaneously using revolutionary OFDMA technology
- EXTENSIVE COVERAGE: Achieve the strong, reliable WiFi coverage with Archer AX1800 as it focuses signal strength to your devices far away using Beamforming technology, 4 high-gain antennas and an advanced front-end module (FEM) chipset
- OUR CYBERSECURITY COMMITMENT: TP-Link is a signatory of the U.S. Cybersecurity and Infrastructure Security Agency’s (CISA) Secure-by-Design pledge. This device is designed, built, and maintained, with advanced security as a core requirement.
Amazon later confirmed that Amazon.com, subsidiaries and AWS Support had been impacted. Its statement did not provide a complete consumer-by-consumer breakdown of every Prime Video or Alexa feature that failed, so those effects should be understood alongside contemporary reporting rather than treated as a single identical outage.
Fortnite: mainly an online-services and login disruption
Fortnite was among the games reported as affected during the AWS incident. The most defensible description is that Fortnite users experienced disruption to login and related Epic online services, rather than claiming that every Fortnite match, server and feature failed simultaneously.
Fortnite depends on several separate systems, including account authentication, matchmaking, parties, game-session services, purchases and the Epic Games launcher or store. A failure in an AWS-backed dependency could prevent login or matchmaking even if some game servers and already-running sessions remained available. Contemporary coverage quoted Fortnite or Epic status messaging about login disruption; it did not establish that every Fortnite feature was unavailable for the entire event. See the Epic Games public status page for the type of service-level distinction that matters in incidents like this.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Perplexity: a confirmed AWS-related incident
Perplexity provides one of the clearest examples of confirmed attribution. Its first-party incident history recorded an investigation beginning at 6:54 a.m. UTC on October 20, identification of an AWS-related cause at 7:32 a.m. UTC, “most services operational” by 9:40 a.m. UTC and resolution at 9:55 a.m. UTC.
That timeline also demonstrates why a service can recover before AWS declares the entire event over: Perplexity’s affected path may have recovered while other AWS control-plane and networking problems continued elsewhere.
Was the whole internet or all of AWS down?
No. The event was not a total worldwide AWS shutdown and it was not a general internet failure.
The primary incident centered on the US East (N. Virginia) AWS Region, us-east-1. AWS Regions are separate geographic areas designed to provide isolation from one another, but isolation only helps when a customer has deliberately placed the required application, data, identity path, DNS failover and recovery tooling in another Region. AWS explains the relationship between Regions and Availability Zones in its global infrastructure documentation.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Other AWS Regions were not universally offline. However, an application running outside Virginia could still fail if it depended on a service endpoint, identity system, control-plane operation, global table replica or other component tied to us-east-1. AWS specifically noted effects involving services or features relying on us-east-1 endpoints, including IAM and DynamoDB Global Tables.
The accurate summary is therefore: a regional AWS failure produced global-facing disruption. Calling it “the entire AWS cloud going down worldwide” is broader than the evidence supports, while calling it “just a local Virginia outage” understates the dependency chain.
What caused the AWS outage?
The immediate trigger was a DynamoDB endpoint-resolution failure, but “a DNS problem” is too shallow to describe what happened.
DynamoDB used an automated DNS-management system containing a DNS Planner and multiple independent DNS Enactors. The Planner generated endpoint plans. The Enactors applied those plans to Route 53. The design included a check intended to stop an older plan from overwriting a newer one.
Rank #2
- Dual-band Wi-Fi with 5 GHz speeds up to 867 Mbps and 2.4 GHz speeds up to 300 Mbps, delivering 1200 Mbps of total bandwidth¹. Dual-band routers do not support 6 GHz. Performance varies by conditions, distance to devices, and obstacles such as walls.
- Covers up to 1,000 sq. ft. with four external antennas for stable wireless connections and optimal coverage.
- Supports IGMP Proxy/Snooping, Bridge and Tag VLAN to optimize IPTV streaming
- Access Point Mode - Supports AP Mode to transform your wired connection into wireless network, an ideal wireless router for home
- Advanced Security with WPA3 - The latest Wi-Fi security protocol, WPA3, brings new capabilities to improve cybersecurity in personal networks
According to AWS’s post-event technical explanation, one Enactor became unusually delayed while retrying updates. During that delay, newer plans were generated and applied by another Enactor. When the delayed worker resumed, its original age check was stale. It treated the old plan as acceptable and overwrote a newer plan.
A cleanup process then deleted the old plan. The result was an inconsistent DNS state: the regional DynamoDB endpoint had an empty DNS record and therefore no usable IP addresses. New connections to DynamoDB in us-east-1 failed, and the recovery automation could not restore the state automatically.
In plain English:
A delayed redundant DNS worker applied an old plan after a newer plan had already succeeded. Cleanup then removed the active record, leaving DynamoDB’s regional endpoint without usable addresses.
This was not simply a person deleting one DNS entry, and it was not DynamoDB behaving like a crashed database. The failure involved concurrent automation, stale validation, plan cleanup, an inconsistent state and recovery logic that could not safely repair the result.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Why did one DynamoDB failure affect so many unrelated services?
DynamoDB was a foundational dependency for AWS internal systems as well as customer applications. AWS says services that needed to establish new connections to the regional DynamoDB endpoint began failing immediately. Internal systems used DynamoDB for state management, leases, queues, orchestration and other control-plane functions.
The broad causal chain looked like this:
dynamodb.us-east-1.amazonaws.comstopped resolving correctly.- New DynamoDB connections failed.
- AWS subsystems and customer applications began accumulating failed requests, retries and backlogs.
- EC2 lease-management processes lost progress because they depended on DynamoDB.
- New EC2 instance launches became unavailable, delayed or heavily throttled.
- Newly launched instances experienced delayed network configuration.
- Network Load Balancers saw apparently unhealthy capacity and removed nodes from service.
- Applications experienced connection errors, capacity shortages, delayed jobs or login failures.
This is the central lesson of the incident: cloud services are dependency graphs, not isolated boxes. Multi-AZ application servers can still depend on a regional control-plane service. A product with a global user base can still have a critical path through one Region.
Why did recovery take most of the day?
The DynamoDB DNS record was repaired relatively early. The systems that had been damaged or overloaded while DynamoDB was unavailable took much longer to recover.
DynamoDB and DNS recovery
AWS restored the DynamoDB DNS information at approximately 2:25 a.m. PDT. Cached DNS records expired between roughly 2:25 and 2:40 a.m., after which customers could again resolve the endpoint and establish connections.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsThat did not instantly clear failed requests, expired leases or queued recovery work. Systems had to reconnect, rebuild state and process accumulated demand without triggering another overload.
EC2 launch and network-propagation problems
Existing EC2 instances launched before the incident generally remained healthy. The main EC2 impact involved new instance launches and the network configuration that followed them.
AWS said its DWFM lease checks began failing because DWFM depended on DynamoDB. As leases expired, many capacity units became ineligible for new launches. Recovery attempts then created a large queue, pushing DWFM into what AWS called “congestive collapse.”
Rank #3
- 𝐅𝐮𝐭𝐮𝐫𝐞-𝐏𝐫𝐨𝐨𝐟 𝐘𝐨𝐮𝐫 𝐇𝐨𝐦𝐞 𝐖𝐢𝐭𝐡 𝐖𝐢-𝐅𝐢 𝟕: Powered by Wi-Fi 7 technology, enjoy faster speeds with Multi-Link Operation, increased reliability with Multi-RUs, and more data capacity with 4K-QAM, delivering enhanced performance for all your devices.
- 𝐁𝐄𝟑𝟔𝟎𝟎 𝐃𝐮𝐚𝐥-𝐁𝐚𝐧𝐝 𝐖𝐢-𝐅𝐢 𝟕 𝐑𝐨𝐮𝐭𝐞𝐫: Delivers up to 2882 Mbps (5 GHz), and 688 Mbps (2.4 GHz) speeds for 4K/8K streaming, AR/VR gaming & more. Dual-band routers do not support 6 GHz. Performance varies by conditions, distance, and obstacles like walls.
- 𝐔𝐧𝐥𝐞𝐚𝐬𝐡 𝐌𝐮𝐥𝐭𝐢-𝐆𝐢𝐠 𝐒𝐩𝐞𝐞𝐝𝐬 𝐰𝐢𝐭𝐡 𝐃𝐮𝐚𝐥 𝟐.𝟓 𝐆𝐛𝐩𝐬 𝐏𝐨𝐫𝐭𝐬 𝐚𝐧𝐝 𝟑×𝟏𝐆𝐛𝐩𝐬 𝐋𝐀𝐍 𝐏𝐨𝐫𝐭𝐬: Maximize Gigabitplus internet with one 2.5G WAN/LAN port, one 2.5 Gbps LAN port, plus three additional 1 Gbps LAN ports. Break the 1G barrier for seamless, high-speed connectivity from the internet to multiple LAN devices for enhanced performance.
- 𝐍𝐞𝐱𝐭-𝐆𝐞𝐧 𝟐.𝟎 𝐆𝐇𝐳 𝐐𝐮𝐚𝐝-𝐂𝐨𝐫𝐞 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐨𝐫: Experience power and precision with a state-of-the-art processor that effortlessly manages high throughput. Eliminate lag and enjoy fast connections with minimal latency, even during heavy data transmissions.
- 𝐂𝐨𝐯𝐞𝐫𝐚𝐠𝐞 𝐟𝐨𝐫 𝐄𝐯𝐞𝐫𝐲 𝐂𝐨𝐫𝐧𝐞𝐫 - Covers up to 2,000 sq. ft. for up to 60 devices at a time. 4 internal antennas and beamforming technology focus Wi-Fi signals toward hard-to-reach areas. Seamlessly connect phones, TVs, and gaming consoles.
Engineers throttled incoming work and selectively restarted DWFM hosts. New launches began succeeding again at 5:28 a.m. PDT, but the network-propagation backlog continued. New instance network-propagation times returned to normal at 10:36 a.m. EC2 APIs and new instance launches were fully normal at 1:50 p.m. PDT.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteNetwork Load Balancer health checks
New instances without complete network configuration could not pass NLB health checks. Healthy capacity was consequently removed from service, then re-added as recovery progressed. That produced connection instability for some applications even after the original DynamoDB problem had been addressed.
AWS disabled automatic health-check failover at 9:36 a.m. PDT, restoring available capacity to affected load balancers. NLB recovery was completed by 2:09 p.m. PDT.
The result was a recovery tail with several phases: DynamoDB endpoint restoration, DNS-cache expiration, EC2 lease recovery, instance-network propagation and NLB health-check stabilization.
October 20 AWS outage timeline
All times below are Pacific Daylight Time, as used in AWS’s incident summary.
Recommended Free Tools
| Time | Event |
|---|---|
| October 19, 11:48 p.m. | Customer impact begins with DynamoDB API errors in us-east-1. |
| October 20, 12:11 a.m. | AWS publicly reports increased error rates and latency affecting multiple services. |
| 12:38 a.m. | Engineers identify DynamoDB DNS state as the source. |
| 1:15 a.m. | Temporary mitigations restore some internal connectivity and repair key operational tooling. |
| 2:25 a.m. | DynamoDB DNS information is restored. |
| 2:25–2:40 a.m. | Cached DNS records expire, allowing customers to resolve the endpoint and establish connections again. |
| 4:14 a.m. | Engineers throttle work and begin selective DWFM restarts to recover EC2 lease processing. |
| 5:28 a.m. | New EC2 launches begin succeeding, although throttling and networking backlogs remain. |
| 5:30 a.m.–2:09 p.m. | NLB connection errors affect some load balancers while capacity and health-check state recover. |
| 10:36 a.m. | New EC2 instance network-propagation times return to normal. |
| 1:50 p.m. | EC2 APIs and new instance launches return to normal. |
| 2:20 p.m. | AWS says the overall event has ended. |
| 3:01 p.m. | Amazon says all AWS services have returned to normal operations, with some backlogs still being processed. |
The initial public health notice appeared at 12:11 a.m. PDT on October 20, while AWS’s detailed report places the beginning of customer impact at 11:48 p.m. PDT on October 19. That explains why reports use different start dates.
Which services were affected?
There is no single AWS-certified public list of every consumer service that experienced an interruption. The table below separates stronger evidence from reports based on user-submitted outage data.
| Service or category | Reported effect | Evidence level |
|---|---|---|
| Amazon.com | Amazon confirmed that Amazon.com and subsidiaries were impacted by the AWS event. | First-party Amazon statement |
| Prime Video and Alexa | Contemporary reporting listed access or service problems during the AWS disruption. | Reputable reporting and outage reports; no complete first-party feature breakdown in Amazon’s summary |
| Fortnite and Epic online services | Login and related online-service disruption was reported; this should not be simplified into every game session failing. | Contemporary reporting and Epic-related status messaging |
| Perplexity | Perplexity’s status history explicitly connected elevated website errors to the AWS operational issue. | First-party service status |
| Snapchat, Roblox, Reddit and Signal | Users reported service disruption during the incident. | AP, Reuters and company or user reports, varying by service |
| Venmo, Robinhood and Coinbase | Reported payment or financial-service disruptions; Reuters said Coinbase and Robinhood attributed problems to AWS. | Reputable reporting and company attribution |
| Duolingo, Canva and Canvas | Reported availability problems, with some institutions separately confirming AWS-related effects in Canvas environments. | Reputable reporting and institutional notices |
| Lyft, banking and government services | Some regional or service-specific effects were reported. | Requires separate confirmation for each service; do not treat the entire category as a confirmed total outage |
Reuters’s affected-service factbox sourced its broad list to Downdetector, which aggregates user reports. That makes it useful for understanding the scale and timing of complaints, but it does not prove that AWS was the sole cause of every report. A careful classification is:
- AWS-confirmed impact.
- Company-confirmed AWS attribution.
- Reputable reporting based on outage data.
- User reports without independent confirmation.
- Unverified or coincidental reports.
For broader contemporary coverage, see the Reuters service factbox, Reuters’ overview, the Associated Press report and contemporary live coverage.
Was data lost?
AWS’s public post-event summary does not report widespread customer data loss. That does not mean every individual service had an identical outcome, but there is no public AWS evidence in the supplied incident account that the outage broadly erased customer data.
It is important to distinguish four different conditions:
Rank #4
- Dual band router upgrades to 1200 Mbps high speed internet (300mbps for 2.4GHz plus 900Mbps for 5GHz), reducing buffering and ideal for 4K stream
- Full Gigabit Ports - Gigabit Router with 4 Gigabit LAN ports, ideal for any internet plan and allow you to directly connect your wired devices
- Boosted Coverage - Four external antennas equipped with Beamforming technology extend and concentrate the Wi-Fi signals
- MU-MIMO technology - (5GHz band) allows high speeds for multiple devices simultaneously
- Access Point Mode - Supports AP Mode to transform your wired connection into wireless network, an ideal wireless router for home
- Availability failure: a request could not be made or a service could not be reached.
- Replication lag: a secondary copy temporarily fell behind.
- Backlog: queued work took time to process after recovery.
- Data loss: information permanently disappeared.
Customers using DynamoDB Global Tables could connect to replicas in other Regions, although replication involving the us-east-1 replica experienced prolonged lag. AWS said all Global Tables replicas had caught up by 2:32 a.m. PDT.
There is no basis for claiming that users generally lost Fortnite inventories, Perplexity conversations, Amazon orders or other application data. If a purchase, upload, account change or order was in progress during the disruption, users should verify its final status before trying it again.
Free tools Windows power users keep installed
One-click scans. No signup required.
Was the AWS outage a cyberattack?
AWS did not attribute the incident to a cyberattack. Its post-event report identifies the DynamoDB DNS-management race condition and the subsequent EC2 and NLB recovery problems as the cause. The Associated Press also reported no indication of a cyber incident at the time.
The precise wording matters: the available public evidence points to an internal AWS software failure, not malicious activity. It is more accurate to say that AWS attributed the outage to an internal software failure than to claim that malicious activity was mathematically impossible.
Why didn’t multiple Availability Zones prevent the outage?
Availability Zones protect against failures affecting one zone within a Region. They do not automatically protect an application from a failure in a regional service or control plane used by every zone.
An application can have redundant EC2 instances, databases and load balancers across several Availability Zones and still fail if all of those components need a regional DynamoDB endpoint, IAM operation, DNS path or capacity-management process in us-east-1.
Multi-Region architecture can provide stronger fault isolation, but only if the failover path is genuinely independent. AWS’s Well-Architected guidance distinguishes multi-AZ from multi-Region resilience and notes that workloads requiring protection beyond one Region need deliberate cross-Region design. See AWS’s guidance on fault isolation across Availability Zones and Regions.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Does multi-Region automatically solve this problem?
No. Running application servers in two Regions is not enough. A resilient design must also account for:
- Database replication, conflict handling and write authority.
- Identity and authentication endpoints.
- DNS and traffic failover.
- Secrets, certificates and encryption keys.
- Container images and machine images.
- Queues, event processing and scheduled jobs.
- Configuration stores and deployment tooling.
- Available capacity in the failover Region.
- Cross-Region dependencies that quietly lead back to the impaired Region.
- Data-residency, compliance and cost requirements.
A standby Region that cannot authenticate users, retrieve secrets, receive traffic or access the authoritative database is not a complete failover solution. AWS’s multi-Region dependency guidance recommends analyzing AWS, internal, third-party, network, certificate, key, secret and parameter dependencies. Its multi-Region architecture guidance covers the additional design and operating trade-offs.
What should consumers do during a similar outage?
- Check the affected service’s official status page.
- Check the AWS Health Dashboard when several unrelated services fail at once.
- Try an unrelated website or service to distinguish a local network problem from a provider outage.
- Avoid repeatedly logging in if the provider has acknowledged a backend incident.
- Do not repeatedly submit a payment, purchase or account change until you know whether the first attempt succeeded.
- After recovery, verify orders, payments, uploads, game purchases, account changes and delayed messages.
- Contact support if the problem continues after the service reports recovery.
Reinstalling an app, clearing its cache or changing Wi-Fi cannot repair a failed AWS-side dependency. Those steps are useful for local problems, but not for a confirmed provider outage.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallRecovery can also be uneven. One person may regain access while another still sees errors because of DNS caching, regional traffic, rate limits, queued work or service-specific recovery. A website returning to “operational” does not guarantee that every delayed job has already completed.
Best Value
- Free Omada Essentials Cloud Management: Free cloud management with no additional fees, everything is managed in the cloud without the need for hardware or software controllers. Simply launch the Omada app, scan the S/N code on the package, and you're ready to deliver
- Ultra-Fast True Wi-Fi 6 Speeds: Designed with the latest wireless Wi-Fi 6 technology featuring 1024-QAM, HE60 and Long OFDM Symbol, the EAP650 boosts dual-band Wi-Fi speeds up to 2976 Mbps
- Ultra-Slim Design: Compact design ensures simple installation while saving space. The elegant appearance makes EAP650 blend into any modern office, hotel, classroom, or cafe
- Integrated into Omada SDN: Omada Software Defined Networking (SDN) platform integrates network devices including access points, switches and gateways with multiple control options offered - Omada Hardware controller, Software Controller or Cloud-based controller. Standalone mode also supported
- Cloud Access Omada Compatibility: Remote Cloud access and Omada app enables centralized cloud management of the whole network from different sites, all controlled from a single interface anywhere, anytime
What businesses should learn from the incident
The outage does not prove that cloud hosting is inherently unreliable, nor does it mean every company must immediately duplicate everything across multiple clouds. The narrower and more useful lesson is that centralized dependencies and non-independent recovery paths can turn a regional control-plane failure into a broad customer outage.
Application operators should:
- Map direct and indirect dependencies, including control-plane, identity, deployment and observability services.
- Test the application when a regional DynamoDB endpoint cannot resolve, not only when a database returns errors.
- Keep recovery procedures from depending on the same impaired service whenever possible.
- Use bounded retries, exponential backoff and circuit breakers to prevent retry storms.
- Preserve cached or static functionality so read-only or degraded traffic can continue.
- Maintain enough capacity in a secondary Region to meet the required recovery objective.
- Test DNS failover, not merely server failover.
- Test stale queues, expired leases, delayed instance networking and health-check flapping.
- Monitor backlog age and queue depth, not only request success rates.
- Document manual recovery procedures for automation that can enter an inconsistent state.
- Rehearse failover and failback under production-like load.
AWS said its own follow-up work included disabling DynamoDB DNS Planner and Enactor automation worldwide pending fixes and additional protections, adding Network Load Balancer velocity controls, expanding EC2 recovery testing and improving throttling based on queue size. The incident therefore exposed both an initial automation race and weaknesses in the systems that had to recover from it.
The most important correction to the “AWS crashed the internet” headline
The headline captures the user experience but compresses several different technical facts into one phrase. The event was:
- Real and broad enough to affect users around the world.
- Centered on one AWS Region rather than every AWS Region.
- Triggered by a DynamoDB DNS-automation race condition.
- Extended by EC2 lease and launch recovery problems.
- Compounded by NLB health checks removing apparently unhealthy capacity.
- Uneven across services, users and features.
- Not publicly attributed to a cyberattack.
That combination explains why Amazon, Prime Video, Fortnite, Perplexity and unrelated third-party services could show failures at the same time without the whole internet being offline.
Frequently Asked Questions
Is AWS still down?
No. The October 20, 2025 incident was resolved. AWS reported the overall event ended at 2:20 p.m. PDT, while Amazon said normal operations had returned by 3:01 p.m., with some backlogs still being processed.
Was the whole internet down?
No. The primary failure occurred in AWS’s US East (N. Virginia) Region, or us-east-1. Other services and Regions could still be affected when they depended on us-east-1 endpoints or control-plane systems.
Why did Fortnite stop working?
Fortnite users experienced disruption to login and related Epic online services. That does not prove that every Fortnite game session, server or feature failed at the same time.
Recommended Free Tools
Why was Perplexity affected?
Perplexity’s own incident history explicitly linked its website errors to the AWS operational issue and recorded recovery later that morning in UTC.
Did Amazon.com and Prime Video use different infrastructure?
Amazon.com and Prime Video are separate products, but services can share regional infrastructure, identity paths, APIs or other dependencies. Amazon confirmed impact to Amazon.com and subsidiaries; contemporary reports also listed Prime Video problems.
Was the AWS outage a hack?
AWS attributed the outage to an internal software race condition in DynamoDB’s DNS-management automation. No public evidence indicated a cyberattack.
Was data lost?
AWS’s public post-event summary does not report widespread customer data loss. It does describe availability failures, replication lag and backlogs, which are not the same as permanent data loss.
Why didn’t another AWS Availability Zone save the service?
Multiple Availability Zones protect against many zone-level failures, but they do not automatically isolate an application from a regional control-plane dependency such as a regional DynamoDB endpoint or IAM path.
How can businesses protect themselves from a similar outage?
Map regional and control-plane dependencies, maintain a genuinely independent failover Region, test DNS and identity failover, use bounded retries and circuit breakers, monitor backlogs, and rehearse recovery from expired leases and health-check failures.
The Bottom Line
The October 20, 2025 AWS outage was a regional us-east-1 failure with worldwide consequences. A race in DynamoDB’s automated DNS system removed usable endpoint information, and the resulting EC2 and Network Load Balancer recovery problems prolonged the disruption. Fortnite users mainly encountered login or related Epic online-service problems, while Perplexity explicitly confirmed an AWS-related incident. AWS reported normal operations by the afternoon and attributed the event to an internal software failure, not a cyberattack.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
