On the morning of October 20, 2025, millions of users found that Snapchat, Fortnite, Airbnb, and Reddit were frozen. Additionally, voice assistants remained silent, while screens displayed blank pages. The root cause was an AWS DNS outage at Amazon Web Services (AWS), originating from the AWS US-EAST-1 region in Virginia, which disrupted DNS resolution and caused DynamoDB errors when accessing the API. The first signal appeared at 09:11 in Paris, with a gradual recovery by late morning: our dependence on the cloud is laid bare. At 09:11 in Paris, the unavailability of AWS services became apparent. The recovery would be gradual, and our dependence on the cloud, glaring.
What We Know at This Time
On Monday, October 20, 2025, a failure at Amazon Web Services (AWS), Amazon’s cloud subsidiary, caused high error rates and latencies in the AWS US-EAST-1 region located in Virginia. Some of the most widely used platforms in the world faltered, from Snapchat to Fortnite and Roblox, from Airbnb to Canva, from Reddit to Prime Video, not to mention Alexa, Perplexity, Signal, Zoom, Tinder, or Ring. The difficulties were felt in Europe, the Americas, and Africa, causing connection problems. In France, this resulted in a flood of error messages and connection impossibilities. Real-time monitoring converged on the AWS Health Dashboard (AWS Status (Health) dashboard), while Downdetector (AWS outage) aggregated waves of reports.
First signals at 09:11 in Paris; notable returns between 11:27 and 11:35; at 13:06, AWS indicated a major improvement. However, local slowdowns persisted depending on the areas and services.
Where the Outage Originated
The indications published by AWS and reported by the specialized press converge on a Route 53 outage (DNS resolution) that would have hindered access to the DynamoDB API, generating DynamoDB errors. The failure, located in AWS US-EAST-1, spread in a cascading effect to other components, triggering spikes in errors and latencies on dependent services. In an ecosystem where managed components — databases, queues, load balancers, API gateways — intertwine, a node that derails and the orchestra stalls. Nothing at this stage indicates a malicious act. On the contrary, everything points to the presence of a network bug. Its propagation is spectacular because it affects millions of users.
The Concrete Repercussions for Users
The outage took the form of a suddenly jammed daily life. Unavailable applications (AWS outage), and silent voice assistants, are symptoms. Additionally, Alexa’s alarms become erratic. Login pages display only blank pages. On the side of encrypted messaging, Signal acknowledged disruptions. Video conferencing services like Zoom also experienced hiccups. Gaming platforms saw their queues lengthen, and some sessions abruptly ended. For creators and freelancers, dependent on online tools, the forced intermission reminded them of this fragility. Indeed, their activity has become natively connected.
Companies and Institutions Facing the Grain of Sand
In companies, the incident summoned all IT teams. SLAs were put under pressure, support channels saturated, and security and production management closely followed the provider’s updates. In the UK, institutions and major players like Lloyds Bank, Vodafone, BT, or the HMRC tax administration experienced AWS service disruptions, alongside signals noted by Downdetector (AWS outage). The operational losses are difficult to quantify immediately. However, the abrupt halt of a payment, a reservation, or a customer service quickly generates measurable costs. Additionally, it affects the company’s image.
The Precedent That No Longer Surprises
This outage once again highlights the extreme concentration of the cloud market, dominated by AWS, Microsoft, and Google. It underscores our hyper-dependence on technical layers invisible to the general public. However, each familiar application represents the emerging tip of these technologies. The modern, distributed, and elastic architecture promises resilience. However, this promise is kept only if one anticipates black scenarios. Additionally, one must accept the cost of true redundancy. A single region, however vast, remains a point of fragility. A single provider, however powerful, remains a systemic risk.
Precise Timeline of a Chaotic Morning
Over the hours, markers piled up. 09:11: the first anomalies are reported from consoles and metrics. 11:27 then 11:35: essential services begin to respond correctly again, sometimes with latencies higher than normal. 13:06: the majority of operations are announced as restored. However, vigilance is maintained on certain subsystems likely to accumulate backlogs and internal queues. On a global scale, the impact moved in waves. Indeed, the hourly salvos of connections relentlessly tested the remaining capacity.
How the Giants Repair on the Fly
Behind the scenes, AWS teams first stabilized the DNS layer, then cleared the backlogs in queues and caches, before gradually restoring the entry points of managed services. Clients, in turn, triggered BCP/DRP plans calibrated for these contingencies: traffic switching, failover via Route 53, limiting non-critical features, activating degraded modes, suspending certain writes until a safe path is found. In startups as well as in very large companies, incident communication played a central role. Internally, it served to coordinate actions, while externally, it informed without saying too much.
Resilience Paths for Tomorrow
Resilience is not decreed, it is architected. A pragmatic multi-cloud can protect stateless services by replicating them across multiple providers, provided one invests in portable identity management, synchronized secrets, and real network neutrality. A multi-region strategy, in active-active or active-passive, is essential for vital workloads, with automated failover regularly tested via Route 53 at AWS or its equivalents. RTO/RPO objectives must be assumed, measured, even disputed, then trained through chaos engineering exercises. On the client side, circuit breakers prevent relentless requests when the downstream collapses. Queues and events — SQS or Kafka — allow absorbing shocks and smoothing peaks. Offline modes and local caches support essential functions when the network goes silent. Finally, observability unifying logs, metrics, and traces across multiple providers offers a less myopic view. This is particularly useful during stormy nights.
For Readers: Where to Get Information and What to Do
To know the state of affairs, the AWS Health dashboard publishes incidents by service and region; its documentation reminds the meaning of alerts. Official updates also pass through @awscloud and the AWS Status (Health) dashboard, while the field view remains fueled by Downdetector (AWS outage) and its mapping of reports in France.
In case of DNS-related errors on the user side, one can clear the machine’s cache. Then, it is recommended to restart the box. Next, it is possible to test a public resolver like 8.8.8.8 or 1.1.1.1. Finally, clear the browser cache as well as the cookies of the targeted site. In companies, immediate reflexes focus on internal resolvers, firewall rules, a possible split-horizon DNS, and the health of proxies or an SD-WAN.
What This Says About Our Digital Present
It would be wrong to see in today’s incident just a bad moment to pass. This outage illustrates the fragility of a world that has outsourced its memory to private clouds. Additionally, computing power is also outsourced there. However, this leads to sometimes thoughtless dependence. It questions contractual clauses, penalties related to SLAs, exclusions of liability, but also data governance and architectures. It forces arbitration between the simplicity of all-integrated and the complexity of multifaceted redundancy. It finally reminds that sobriety is not the enemy of performance. Indeed, frugal services tolerant to intermittence rely on cache. Moreover, graceful degradation allows them to better absorb the unexpected. This is more effective than cathedrals of micro-services where each action depends on another.
The Face of AWS and the Weight of the Cloud
At the head of Amazon since 2021, Andy Jassy long led AWS. Indeed, he was the architect of its rise. Under his leadership, the cloud division became the group’s cash cow and its innovation lab, from serverless functions to managed databases, to on-demand machine learning. This success has anchored AWS at the heart of an economy where on-demand computing has become the norm. This centrality explains the scope of incidents: when a backbone falters, a sensitive part of global traffic is slowed. Regulators and CIOs see this as an issue of sovereignty and technological dependence. Meanwhile, architects question the right balance between specialization and portability.
The concentration of the market fuels a recurring debate on exit costs and the lock-in induced by differentiating services. The adoption of open standards can loosen this grip. However, the generous use of containers and common interfaces often leads to costs. Moreover, there are rarely compromises on performance. Conversely, the promise of tight integration can justify attachment to a single provider. Additionally, a rich ecosystem can also justify it, but this mainly concerns the short term. Today’s incident forces these arbitrations to be reconsidered.
What to Remember: Resilience, Prudence, Diversity
It only takes one zone down, in this case, AWS US-EAST-1, for the world to become aware of the dark side of its digital comfort. The incident of October 20, 2025, where we saw from Snapchat to Reddit, from Airbnb to Prime Video, stumble in the same momentum, offers a simple lesson: resilience is not an option. It is an intention that is conceived, financed, and tested. The following hours will tell if the root cause — a DNS resolution outage degrading access to DynamoDB — is confirmed. They will not erase the essential: to remain fluid, our digital present requires changing scale, not in raw power, but in prudence and diversity. For real-time updates, refer to the AWS Health dashboard.