A DNS failure in Amazon’s critical US-EAST-1 region knocked out Snapchat, Fortnite, and major banks, costing billions and reigniting fears over the internet’s fragile reliance on a handful of tech giants.
I. A World Offline: The Digital Domino Effect
On the morning of October 20, 2025, millions of people woke up to a world that had gone digitally silent. Snapchat messages vanished into the ether, Fortnite and Roblox lobbies were empty, and workflows ground to a halt.1 Even the ambient intelligence of the modern home fell quiet, as Amazon’s own Alexa and Ring smart doorbells became unresponsive.3 This was not a localized glitch but the first tremor of a global internet earthquake, originating from a single point in Northern Virginia and radiating outward to touch nearly every corner of the digital economy.
The scale of the chaos was quickly quantified by outage-tracking website Downdetector, which logged over 11 million user reports of service disruptions worldwide.6 The United States was the hardest hit, with over 3 million reports, followed by the United Kingdom with over 1 million and Australia with more than 418,000, demonstrating the truly global reach of the failure.3
No industry was immune to the cascading failure. The outage paralyzed a vast and diverse array of services, revealing the deeply interconnected and often invisible dependencies that underpin modern life.
- Social Media & Communication: Popular platforms including Snapchat, Reddit, and the secure messaging app Signal experienced widespread login failures and severely degraded performance. Signal’s President, Meredith Whittaker, confirmed on the social media platform X that her service was among the casualties.8
- Gaming & Entertainment: The world’s largest gaming and streaming platforms went dark. Epic Games’ Fortnite, Roblox, Pokémon Go, and the PlayStation Network were all knocked offline, alongside streaming services like Amazon’s Prime Video and Crunchyroll.1
- Finance & Commerce: The disruption struck the heart of the digital economy, crippling financial trading apps like Robinhood and Coinbase, and the popular peer-to-peer payment service Venmo.11 Critically, the outage also impacted major high-street banks, including Lloyds, Halifax, and Bank of Scotland in the UK, preventing customers from accessing their accounts.7 Amazon’s own e-commerce behemoth was not spared, suffering from intermittent outages.1
- Enterprise & Productivity: Businesses around the globe saw productivity plummet as essential tools like Slack, Canva, Asana, and Zoom suffered from instability and outages, leaving remote and in-office teams unable to collaborate.4
- Transportation & Critical Infrastructure: The outage extended into critical infrastructure, with major airlines such as Delta and United Airlines reporting disruptions to their websites and mobile apps, leaving passengers unable to check in or view their reservations.12 In the UK, the government’s own tax authority, HM Revenue and Customs (HMRC), saw its website become inaccessible.7
The event was far more than a temporary inconvenience for social media users and gamers. The simultaneous failure of major banks, national airlines, and government tax portals exposed a systemic vulnerability. Services that form the bedrock of modern economies and civil society are now profoundly reliant on the uninterrupted operation of a single commercial technology provider. This reality transforms what might be dismissed as a “tech outage” into a pressing matter of public infrastructure stability and national economic security, a fact underscored by the UK government’s immediate contact with Amazon regarding the incident.7
| Category | Affected Services | Reported Impact |
| Social Media & Communication | Snapchat, Reddit, Signal, WhatsApp | Total outage, login failures, undelivered messages 2 |
| Gaming & Entertainment | Fortnite, Roblox, PlayStation Network, Prime Video | Inaccessible services, server connection failures 1 |
| Finance & Commerce | Coinbase, Robinhood, Venmo, Lloyds Bank, Halifax | Intermittent connectivity, transaction failures 7 |
| Enterprise & Productivity | Slack, Canva, Asana, Zoom | Degraded performance, service disruptions 9 |
| Transportation & Infrastructure | Delta Air Lines, United Airlines, UK HMRC | Website and app inaccessible, check-in failures 12 |
II. Inside the Blackout: Anatomy of a Failure
The epicenter of this digital earthquake was quickly located: Amazon Web Services‘ (AWS) US-EAST-1 region in Northern Virginia.7 This is not just any data center cluster; it is AWS’s oldest, largest, and most critical hub, serving as a foundational nerve center for a vast portion of the global internet. Its history of causing significant outages has made it a well-known Achilles’ heel for the digital world.8
The root cause was not a malicious cyberattack, despite initial speculation on social media, but a subtle and catastrophic internal technical failure.7 According to AWS’s own reports, the problem was a “DNS resolution failure” affecting the “DynamoDB API endpoint”.2
To understand this in simple terms, one can think of the Domain Name System (DNS) as the internet’s phone book.25 It translates human-friendly service names (like dynamodb.us-east-1.amazonaws.com) into the numerical IP addresses that computers use to find each other. On October 20, this phone book failed. When countless applications and services tried to “call” DynamoDB—a fundamental database service that acts as the brain for many applications—they found the number was disconnected.
This single failure triggered a devastating domino effect. Because DynamoDB is a foundational pillar upon which other core AWS services like EC2 (for computing power) and Lambda (for serverless functions) are built, its unavailability caused a cascading internal collapse. The situation was analogous to an office building where the central employee directory is suddenly wiped clean; the workers (services) were all present, but they had no way of finding or communicating with each other, bringing all productive work to a standstill.2
The crisis unfolded over several hours, meticulously documented on AWS’s own Service Health Dashboard:
- 12:11 AM PDT: AWS first acknowledges “increased error rates and latencies” in US-EAST-1.2
- 1:26 AM PDT: The company confirms “significant error rates” specifically for the DynamoDB endpoint.23
- 2:01 AM PDT: A potential root cause is identified—the DNS resolution issue. Crucially, AWS notes that global services are impacted and that customers are unable to create or update support cases.23
- 2:22 AM PDT: Engineers apply initial mitigations and report “early signs of recovery,” but warn of growing backlogs of unprocessed requests.23
- 3:35 AM PDT: AWS declares the “underlying DNS issue has been fully mitigated.” However, this was far from the end of the crisis. The company acknowledged that services like EC2 and Lambda were still working through a massive backlog of queued requests and experiencing elevated errors.15
Full recovery was a slow, painstaking process that stretched on for hours, with many services remaining unstable long after the initial “fix”.3
Perhaps one of the most alarming aspects of the failure was the breakdown of AWS’s own customer support system. At the peak of the crisis, customers were unable to create or update support cases, effectively cutting off their primary channel for communication and assistance.11 This was not merely an inconvenience but a fundamental failure in crisis response. The very system designed to help clients navigate an outage had become a victim of it. This reveals a critical architectural flaw; while AWS had previously re-engineered its public-facing Service Health Dashboard for multi-region resilience after similar incidents, the same resilience was clearly not applied to the backend support ticketing system.30 This left customers isolated and in the dark at the most critical moment, exposing a significant gap in AWS’s own disaster preparedness and risking a serious erosion of customer trust.
III. The High Cost of Downtime: Counting the Billions
The abstract technical failure quickly translated into staggering and concrete financial losses that rippled across the global economy. The cost of a few hours of digital silence is measured in the billions.
According to an analysis by Tenscope, the outage caused major websites to collectively lose an estimated $75 million per hour.14 Amazon’s own operations bore the brunt of this, with its e-commerce and web services divisions hemorrhaging an estimated $72.8 million for every hour they were disrupted.14 Other major digital platforms also faced catastrophic hourly losses:
- Snapchat: ~$612,000 per hour
- Zoom: ~$533,000 per hour
- Roblox: ~$411,000 per hour
- Fortnite: ~$400,000 per hour 14
However, these direct revenue losses represent only the tip of the iceberg. The true economic damage was magnified by a cascade of indirect costs. Mehdi Daoudi, CEO of the internet performance monitoring firm Catchpoint, estimated that the total financial impact could easily reach into the hundreds of billions of dollars when accounting for the collapse in productivity.6 Millions of workers worldwide, reliant on cloud-based tools like Slack, Asana, and Zoom, were left unable to perform their jobs.
The disruption propagated further into the physical world. The impact on airlines threatened to snarl global supply chains, as a significant portion of the world’s air freight is transported in the cargo holds of passenger aircraft.13 The instability of financial trading platforms like Coinbase and Robinhood introduced significant risk into markets, while the failure of payroll services like Xero and Square created the potential for delayed salary payments, a devastating prospect for employees and small businesses.12
To place these figures in context, the 2024 global outage caused by a faulty CrowdStrike software update was estimated to have cost Fortune 500 companies $5.4 billion in direct losses.13 The October 2025 AWS event, given its similar scale and scope, is on track to inflict a comparable level of economic pain.
Even these comprehensive financial models fail to capture the full picture. The most significant, long-term cost may be the erosion of trust. For an e-commerce platform, a failed transaction is not just a single lost sale; it can mean the permanent loss of a customer. For industries like finance and healthcare, any amount of downtime is a serious compliance issue that can trigger regulatory audits, investigations, and hefty fines.18 This incident fundamentally alters the financial calculation for businesses that rely on the cloud. The cost of an outage is no longer a reactive loss to be absorbed but a long-tail risk that demands proactive, and often expensive, investment in robust disaster recovery and multi-cloud strategies to prevent future trust-destroying events.
IV. Analysis: The Internet’s Achilles’ Heel
The October 20 outage was not a random accident but a symptom of a deep, structural fragility within the modern internet: the extreme centralization of its core infrastructure. Experts have long warned of the perils of a digital ecosystem that relies on a small handful of technology giants—primarily AWS, Microsoft Azure, and Google Cloud—to function.7 With AWS alone commanding roughly a third of the global cloud market, its stability has become synonymous with the stability of the internet itself.2
The digital rights group Article 19 framed the event not as a mere technical problem but as a “democratic failure”.6 When a glitch at a single commercial company can silence independent media, disable secure communication tools like Signal, and bring government services to a halt, it exposes a profound vulnerability in the infrastructure that underpins digital society.
This systemic risk is dangerously concentrated in AWS’s US-EAST-1 region. The Northern Virginia data hub has become a notorious single point of failure, having been the epicenter of major global outages in 2017, 2021, and 2023.20 The fact that this is a predictable pattern raises urgent questions about why so many global services continue to build their platforms with a hard dependency on this historically failure-prone location.21
Furthermore, the outage shattered the illusion of simple redundancy. Many companies architect their systems across multiple Availability Zones (AZs)—distinct data centers within the same geographic region—believing this protects them from failure. However, this event demonstrated that such a strategy is insufficient when the failure occurs at a higher, regional level, such as the core DNS and networking layers that all AZs in that region share.18 This forces a much more difficult and expensive conversation about true resilience, which requires architecting systems to fail over across entirely separate geographic regions or even to different cloud providers altogether.18
This exposes a fundamental conflict between the economics of cloud computing and the requirements of true resilience. While AWS and other providers offer the tools for complex multi-region architectures, these solutions are significantly more expensive and operationally complex to implement and maintain.35 The sheer number of services that collapsed on October 20 proves that, for reasons of cost and simplicity, the vast majority of companies have not made this investment. The collective result of these individual economic decisions is a global digital ecosystem that is optimized for cost-efficiency but is, as a consequence, dangerously brittle. This outage may serve as a painful but necessary market correction, forcing the industry to re-evaluate the true cost of systemic risk and treat resilience not as an optional luxury but as a non-negotiable cost of doing business in a connected world.
V. Déjà Vu: A History of Internet Catastrophes
While its impact was immense, the October 2025 AWS outage is the latest in a series of landmark internet failures, each of which broke the digital world in its own unique way. Comparing these events reveals the different layers at which our interconnected world can fail and highlights the evolving nature of systemic risk.
- Facebook BGP Outage (2021): This event was a failure at the internet’s most fundamental routing layer. A configuration error in the Border Gateway Protocol (BGP) essentially erased Facebook from the internet’s global map. The servers were running, but the “roads” leading to them had disappeared, making them completely unreachable.26
- CrowdStrike Software Failure (2024): This was a catastrophic failure at the endpoint software layer. A faulty update from a single cybersecurity vendor bricked millions of Windows machines worldwide. The internet’s roads were open, but the “vehicles”—the computers used by airlines, banks, and hospitals—were disabled.9
- AWS DNS Outage (2025): This incident was a failure of the internal cloud service infrastructure. The roads were open and the vehicles were running, but inside the AWS “factory,” the central “office directory” had been wiped. Services were online but could not find or communicate with each other, leading to a complete breakdown of internal logistics.26
These three catastrophes, while all causing massive global disruption, originated at fundamentally different levels of the technology stack: network routing, endpoint software, and cloud infrastructure dependency. Yet, they all share a common, ominous theme: a single point of failure—a flawed BGP update, a buggy software patch, a misconfigured DNS system in one region—had a disproportionately massive and global impact.
This progression of failures reveals a critical trend: as our technology stacks become more complex and abstracted, the potential points of failure are becoming more subtle, more deeply embedded, and capable of triggering even more devastating cascades. The risk is moving “up the stack.” We are no longer just dependent on the physical network (BGP) or the software on our machines (CrowdStrike), but on the opaque, proprietary inner workings of the cloud platforms themselves. A failure inside a system like AWS’s internal DNS is far more difficult for the outside world to predict, monitor, or defend against than a failure of a more public protocol. This dramatically increases the world’s systemic reliance on the operational excellence of a very small number of hyperscale cloud providers.
VI. Conclusion: Lessons from the Brink
The road to recovery from the October 20 outage was slow and arduous. Even after AWS announced that the root DNS issue had been “fully mitigated,” the system struggled for hours to work through a colossal backlog of queued requests. Lingering errors and degraded performance persisted, demonstrating that in today’s complex, interconnected systems, recovery is not the simple flip of a switch.3
The event has served as a visceral and costly lesson, prompting urgent calls from experts and engineers for a fundamental rethinking of cloud architecture and risk management.15 The key takeaways are clear and unavoidable: deploying critical applications with a dependency on a single cloud region is no longer a viable or responsible strategy. True digital resilience demands investment in multi-region and, for the most critical workloads, multi-cloud architectures, despite their increased cost and complexity. Businesses must now treat their cloud vendors not as infallible utilities but as critical operational dependencies that require rigorous, independent risk mitigation plans.37
Ultimately, the great AWS outage of 2025 should not be viewed as an unpredictable “black swan” event. It was a manifestation of known risks and a predictable consequence of an increasingly centralized digital infrastructure. It stands as a stark, global wake-up call—a powerful catalyst for a new and more urgent conversation about building a more resilient, robust, and dependable internet for the future.