Cloud Outage Tracker: Major AWS, Azure, Google Cloud, and Cloudflare Outages Since 2020

Last updated: October 21, 2026. This tracker lists every major AWS, Microsoft Azure, Google Cloud, and Cloudflare outage since January 2020, with each provider’s own stated duration, root cause, and source. It covers 62 major incidents. The median outage lasted 6h 12m, and the most common cause was a configuration or deployment change (20 of 62, 32%), not hardware failure.

We add new incidents as providers publish their post-incident reports. Control Plane publishes this tracker. Every row links to its source, usually the provider’s own post-mortem or status page, so you can verify it.

Key statistics

• 62 major outages since 2020: AWS 14, Azure 20, Google Cloud 12, and Cloudflare 16.
• Median duration: 6h 12m; mean 10h 01m across 55 completed incidents.
• 32% were caused by configuration or deployment changes, and another 23% by software bugs or automation. Together, changes and code caused 55% of major outages.
• 18% were caused by power or cooling failures, including the October 2026 Google Cloud London outage and the May 2026 AWS us-east-1 thermal event.
• 9 of AWS’s 14 major outages hit us-east-1 (Northern Virginia), its oldest and largest region.
• Two AWS regions have been disrupted since March 2026. Drone strikes damaged facilities in me-central-1 (UAE) and me-south-1 (Bahrain), AWS declared both unable to reliably support customer applications, and it has said some data stored only in those regions cannot be restored.
• The longest completed outage in this tracker was Azure East US 2 in January 2025, at 50 hours 13 minutes.

The longest cloud outages in this tracker

The two AWS Middle East regions are listed first because they have not fully recovered. Durations measure the provider’s full event window, from first customer impact to the last service back to normal; where the core fault cleared earlier, that figure is shown in parentheses.

What causes cloud outages

Most major outages start with a change. A new configuration, a deployment, or an automated system acting on bad input accounts for 55% of the incidents here, including the November 2025 Cloudflare Bot Management failure and the October 2025 Azure Front Door outage. Physical events are rarer but last longer: power and cooling failures and the 2026 drone strikes produced several of the longest outages on record.

Outage duration by provider

Counts reflect what each provider documents publicly, not underlying reliability. Microsoft publishes post-incident reviews for more events than other providers, and AWS publishes post-event summaries only for its largest incidents, so these numbers should not be read as a reliability ranking. Cloudflare’s incidents in this tracker are mostly edge and network events, which here recovered faster than regional compute outages.

Full incident log

Newest first. Every row links to its source: the provider’s official post-mortem where one exists, otherwise its status page or a credible news report.

The cost of cloud downtime

Downtime costs vary widely by company size and how much revenue runs through the affected system. These figures come from the publishers’ own reports:

• 57% / 1 in 5: 57% of operators said their most recent major outage cost more than $100,000; for the second consecutive year, 1 in 5 reported costs above $1 million. (Uptime Institute, Annual Outage Analysis 2026 (figures from its 2025 survey), 2026)
• $1.85M per hour: High-impact business outages cost respondents an average of $1.85 million per hour, over $30,000 a minute; the annualized cost of high-impact outages was $73.5M in 2026, down from $76M in 2025. (New Relic, 2026 Observability Forecast, 2026) Respondents skew toward large enterprises.
• $400B a year: Downtime costs Global 2000 companies $400 billion annually, or 9% of profits. (Splunk and Oxford Economics, The Hidden Costs of Downtime, 2024) Covers all downtime, including security incidents (56%), not only cloud outages.
• >$300,000 per hour: One hour of downtime costs more than $300,000 for over 90% of mid-size and large enterprises; 41% put it at $1 million to over $5 million. (ITIC, 2024 Hourly Cost of Downtime, 2024) Self-reported estimates; excludes litigation and penalties.
• $500M to $650M: The October 20, 2025 AWS us-east-1 outage caused an estimated $500 million to $650 million in financial losses for US companies. (Parametrix, 2025) An economic-loss estimate; not comparable with CyberCube’s insured-loss figure below.
• $38M to $581M: Preliminary insured-loss estimate for the October 2025 AWS outage is $38 million to $581 million, with most outcomes expected at the lower end. (CyberCube, 2025) Insurance-loss view, not total economic loss.
• $5.4B: The July 19, 2024 CrowdStrike outage caused an estimated $5.4 billion in direct losses for the US Fortune 500 excluding Microsoft; insured losses $0.54 billion to $1.08 billion. (Parametrix, 2024) A faulty endpoint security update, not a cloud provider outage.

How to estimate your own cost per minute

Cost per minute = (R × A × L) + (S + E) + C

• R: Revenue per minute = annual revenue routed through the affected system ÷ 525,600
• A: Share of revenue affected by the outage (0 to 1)
• L: Share of affected revenue actually lost, not just delayed (0 to 1)
• S: Staff cost per minute = people pulled into the incident × fully loaded hourly cost ÷ 60
• E: Idle-employee cost per minute = employees unable to work × fully loaded hourly cost ÷ 60 × productivity lost (0 to 1)
• C: SLA credits and penalties owed to customers, per minute of breach

Total outage cost = cost per minute × outage minutes + one-time costs (remediation, overtime, customer make-goods, regulatory fines). Use revenue that flows through the affected system, not total company revenue. L matters: some revenue is recovered when customers retry, some is gone. SLA credits are usually a percentage of the monthly bill; convert to a per-minute equivalent or add as a lump sum. Excludes churn, brand damage and stock price impact.

Hypothetical example, for illustration only: A hypothetical SaaS company with $50M annual revenue, all through one production app, during a 180-minute regional cloud outage.

Far below survey averages because those skew to large enterprises; plug in your own revenue, headcount and contract terms.

How teams stay online when a cloud region fails

The incidents above share a pattern: an application running in one region, or depending on services anchored in one region, goes down with it. Applications that run in more than one location at the same time can move traffic to a healthy location without a manual cutover.

• Run active-active, not warm standby. A standby that never serves real traffic is untested until the outage.
• Find single-region dependencies. Identity, DNS, secrets, CI/CD, and observability often live in one region even when the application does not.
• Plan for the data layer. Choose replication per service based on how much inconsistency it can tolerate.
• Use health-checked global routing instead of manual DNS changes, with safeguards so automation cannot remove too much capacity at once.

During the October 2025 AWS us-east-1 outage, 20% of Control Plane customers had us-east-1 in their Global Virtual Clouds. Their workloads were active and running in healthy locations within 10 seconds, with no customer downtime, according to Control Plane internal incident telemetry. The 99.999% availability SLA applies to workloads running at least two replicas in at least two locations. For a comparison of approaches, see the best multi-cloud and multi-region failover solutions.

Methodology

• Scope: major incidents at AWS, Microsoft Azure, Google Cloud, and Cloudflare from January 1, 2020 to October 21, 2026. Major means multi-hour regional impact, global impact, or broad downstream disruption. For Azure, which documents far more incidents, we include the 20 with the broadest impact.
• Sources: AWS Post-Event Summaries and Health Dashboard history, Azure Post Incident Reviews, Google Cloud incident reports, and Cloudflare post-mortems. Where no provider report exists, the row cites a credible news report and is marked as such.
• Durations are computed from the provider’s own start and end times. Where a provider’s summary and timeline disagree, we use the timeline and note the difference.
• Root causes are paraphrased from the provider’s own report.
• Coverage: recent years are better documented because some status histories only keep a limited window, so year-over-year counts understate earlier years.
• Updates: we add incidents after providers publish reports. Corrections are welcome through the Control Plane team.

FAQ

How often does AWS go down?

This tracker records 14 major AWS outages since January 2020, 9 of them in us-east-1. AWS publishes post-event summaries only for its largest incidents, and smaller regional disruptions happen more often.

What was the longest AWS outage?

The longest is ongoing: AWS’s me-central-1 (UAE) and me-south-1 (Bahrain) regions have been disrupted since March 2026 after drone strikes damaged facilities, and AWS has said both cannot reliably support customer applications. Among completed AWS incidents, the May 7-8, 2026 us-east-1 thermal event ran 21 hours 30 minutes to majority recovery, and the November 25, 2020 Kinesis outage lasted 17 hours 8 minutes.

What caused the October 2025 AWS outage?

AWS says a latent race condition in DynamoDB’s automated DNS management left the us-east-1 regional endpoint with an empty DNS record that automation could not repair. The event ran from 11:48 PM PDT on October 19 to 2:20 PM PDT on October 20, 2025, a total of 14 hours 32 minutes.

What caused the October 2025 Azure outage?

Microsoft says customer configuration changes processed across two control-plane versions produced incompatible metadata that crashed Azure Front Door edge servers worldwide on October 29, 2025. The outage lasted 8 hours 24 minutes and affected Front Door, CDN, the Azure portal, Entra ID, and Microsoft 365.

What happened in the Google Cloud London outage in October 2026?

Google Cloud reported a power event in one europe-west2 (London) zone, with elevated errors for some regional services, for 7 hours 52 minutes, according to its status page. Google had not yet published a full incident report.

What causes most cloud outages?

Changes. In this tracker, 32% of major outages came from configuration or deployment changes and 23% from software bugs or automation, while 18% came from power or cooling failures.

Is us-east-1 less reliable than other AWS regions?

9 of the 14 major AWS outages in this tracker were in us-east-1. It is AWS’s oldest and largest region and hosts some global control-plane functions, so its incidents have wider impact, but region size and traffic also explain part of the count.

How much does cloud downtime cost?

Uptime Institute’s 2026 analysis found 57% of operators said their most recent major outage cost more than $100,000, and New Relic’s 2026 survey put high-impact outages at $1.85 million per hour. Parametrix estimated the October 2025 AWS outage cost US companies $500 million to $650 million.

How can I protect my application from a cloud outage?

Run the application active-active in at least two regions or providers, remove single-region dependencies such as identity and DNS, choose a data replication pattern per service, and use health-checked global routing so traffic moves without a manual cutover.