Cloud Outage Analysis
AWS Outage History - Major Outages, Real Losses & Your Exposure
Updated October 2026 · 10 major incidents documented · Cost sources re-audited October 2026
AWS Outage History, In Brief
AWS has had 10 major documented outages since 2012, and 8 of the 10 struck the us-east-1 (Northern Virginia) region. The most recent was July 24, 2026, an approximately 80-minute failure in the us-west-2 (Oregon) region: networking hardware carrying routing between the region and the Seattle metro failed, so cross-region traffic timed out while intra-region traffic kept working, taking Apple Pay, Reddit, DoorDash, Hulu, and the PlayStation Network offline (some downstream services took hours longer to recover). It came just eight days after July 16, 2026, a roughly 3.5-hour global CloudFront failure in the VPC Origins control plane - a capacity limit in a Frankfurt Availability Zone (euc1-az2) stopped edge processors loading routing config and served worldwide 5xx errors, taking Canvas, Blackboard, Hugging Face, and the UK National Lottery offline. The longest single stretch of AWS downtime remains May 2026, a roughly 28-hour single-availability-zone (use1-az4) failure caused by a data-center cooling/thermal event that knocked EC2 and EBS offline and took Coinbase (about 7 hours), FanDuel, and CME Group down. The largest region-wide event was October 2025, a roughly 15-hour failure triggered by a DynamoDB DNS race condition that cascaded across EC2, Lambda, and STS and affected around 70,000 organizations. Only two of the ten carry a loss estimate from outside AWS: the February 2017 S3 outage, which cyber-risk modelling firm Cyence estimated cost S&P 500 companies about $150M (and US financial-services firms about $160M), and October 2025, for which CyberCube published an insured-loss range of $38M to $581M. For the other eight, including the widely felt December 2021 7-hour us-east-1 event, nobody has published a customer-loss figure, so this page does not invent one. Significant AWS incidents have occurred at a rate of roughly one per year over the recent 2017-2026 window. Full chronology, the revenue-at-risk arithmetic, and SLA credit math below.
AWS Outage & Downtime Statistics
10
Major outages since 2012
~1/yr
Incident frequency (2017-2026)
8 / 10
In us-east-1 (region concentration)
~28h
Longest single event (May 2026)
2 / 10
With any published loss estimate (Cyence 2017, CyberCube 2025)
Frequency reflects the 10 major incidents documented in the chronology below, not every status-page event. Region concentration counts documented major outages: 8 of 10 struck us-east-1; the two exceptions are both from July 2026 - the July 16 CloudFront VPC Origins event, which originated in a Frankfurt AZ (euc1-az2) but degraded customers globally, and the July 24 us-west-2 (Oregon) network event. AWS has had smaller incidents in other regions not listed here.
AWS Outage History
| Date | Region | Services | Duration | Impact | At risk / $100M revenue |
|---|---|---|---|---|---|
| Jul 24 2026 | us-west-2 (Oregon) | Regional network path to Seattle metro; Direct Connect, EC2/ECS/ELB, API Gateway, Global Accelerator | ~80 min (long recovery tail) | Cross-region traffic timed out; Apple Pay, Reddit, DoorDash, Hulu, PlayStation Network offline | $15,200 |
| Jul 16 2026 | Global (euc1-az2 origin) | CloudFront (VPC Origins control plane) | ~3.5 hours | Global 5xx; Canvas, Blackboard, Hugging Face, UK National Lottery offline | $40,000 |
| May 2026 | us-east-1 (use1-az4) | EC2, EBS (cooling/thermal event); ELB, EKS, Redshift, MSK | ~28 hours | Single AZ; Coinbase ~7h, FanDuel, CME Group offline | $319,600 |
| Oct 2025 | us-east-1 | DynamoDB DNS, cascaded EC2/Lambda/STS | ~15 hours | ~70K orgs, 17M+ reports | $171,200 |
| Jun 2023 | us-east-1 | Lambda (capacity subsystem); cascaded STS, Console, EKS, Connect, EventBridge | ~3 hours | 104 services degraded; ~2h core Lambda, ~4h full recovery | $34,200 |
| Dec 2021 | us-east-1 | EC2, ECS, Lambda, SNS | ~7 hours | Major global disruption | $79,900 |
| Nov 2020 | us-east-1 | Kinesis (cascading) | ~17 hours | Cascaded to CloudWatch, Cognito, Lambda, EventBridge | $194,100 |
| Aug 2019 | us-east-1 | EC2, EBS (AZ power failure) | ~8 hours | Single-AZ power/generator failure; some EBS data permanently lost | $91,300 |
| Feb 2017 | us-east-1 | S3 | ~4 hours | Significant internet-wide disruption | $45,700 |
| Jun 2012 | us-east-1 | EC2, RDS, EBS | ~8 hours | Netflix, Instagram, Pinterest affected | $91,300 |
Dates, regions, durations, root causes and affected services come from the AWS Health Dashboard and AWS post-event summaries where one exists, plus contemporaneous reporting (The Register, Forbes, ThousandEyes). The last column is ours, and it is arithmetic you can redo: $100,000,000 of annual revenue divided by the 8,760 hours in a year is $11,416 an hour, multiplied by the event's duration. It is a gross upper bound - it assumes the outage took you fully offline for its whole length - and it is not an estimate of what the event cost AWS, its customers in aggregate, or anyone else. Scale it to your own revenue, or use the calculator to add margin, recovery and SLA credits. Updated October 2026.
Published Loss Estimates: Only Two of the Ten
AWS has never published a customer-loss figure for any outage in its history, and its post-event summaries deliberately stop at the technical account. For eight of the ten events below, no company disclosure, regulatory filing or named analyst estimate of customer losses exists at all. Two have one, and these are the only two cost figures on this page that come from outside our own arithmetic.
February 28, 2017 · S3 us-east-1
$150M
Cyber-risk modelling firm Cyence estimated the roughly four-hour outage cost S&P 500 companies about $150M, and US financial-services firms a further $160M. These are estimates of economic impact across those populations, not an audited figure and not a bill AWS paid. Reported by NPR, March 2017.
October 20, 2025 · DynamoDB DNS, us-east-1
$38M - $581M
CyberCube's preliminary global insured-loss range for the event it named "Amazonk", published 23 October 2025. Insured loss is a fraction of economic loss, and CyberCube said the outcome most likely sits at the low end of its own range. Not a total-economic-impact figure.
AWS SLA Credit Rules
| Service | SLA % | Credit (99% to SLA) | Credit (<99%) | Claim Window |
|---|---|---|---|---|
| EC2 | 99.99% | 10% of monthly fee | 30% of monthly fee | 30 days |
| S3 Standard | 99.9% | 10% | 25% | 30 days |
| RDS Multi-AZ | 99.95% | 10% | 25% | 30 days |
| Lambda | 99.95% | 10% | 25% | 30 days |
Credits applied to future invoices. Business or Enterprise support required. Full SLA credit asymmetry analysis.
Frequently Asked
When was the most recent AWS outage?
How many major AWS outages have there been?
How often does AWS have a major outage?
Which AWS region has the most outages?
What was the worst AWS outage?
How much does an AWS outage cost businesses?
Why does us-east-1 have so many outages?
What is the AWS SLA credit for an outage?
Related
AWS July 2026 Deep Dive
CloudFront VPC Origins global 5xx event
AWS May 2026 Deep Dive
use1-az4 thermal event, hour by hour
AWS Oct 2025 Deep Dive
DynamoDB DNS outage, hour by hour
AWS SLA Credit Math
What you actually get reimbursed
Azure Outage History
Compare Azure reliability
GCP Outage History
Compare GCP reliability