AWS, Azure, Cloudflare: the 2025 outages that exposed the internet’s weak points
In seven weeks in late 2025, faults at Amazon Web Services, Microsoft Azure and Cloudflare knocked apps, banks and websites offline. Each company blamed its own systems, not hackers. A year on, regulators have put the biggest cloud providers under direct oversight.

LONDON On 20 October 2025 a fault inside Amazon Web Services’ Northern Virginia region, known as US-EAST-1, disrupted services from 06:48 UTC until 21:20 UTC 1. Snapchat, Reddit, Fortnite, Amazon’s own shopping site, Prime Video and Alexa were among those affected 2. In Britain, Lloyds Bank, Bank of Scotland, Vodafone, BT and the HMRC tax website were hit, according to Downdetector 2.
Nine days later Microsoft’s Azure Front Door network failed 3. On 18 November Cloudflare, which says 20% of websites worldwide use its services in some form, went down too 4. Each company published a post-incident report. Each company described a fault in its own systems, and Cloudflare said explicitly that it was not caused by an attack 135.
AWS: an empty DNS record
AWS’s post-event summary traced the outage to DynamoDB, its database service 1. “The root cause of this issue was a latent race condition in the DynamoDB DNS management system that resulted in an incorrect empty DNS record for the service’s regional endpoint,” the company wrote 1. Two automated components that update DNS records interacted in what AWS called an unlikely sequence: a delayed one applied an older plan, and a clean-up process then deleted it, removing every IP address for the endpoint 1. Fixing it required manual intervention by operators 1.
DNS was restored by 2:25 AM Pacific time, but knock-on effects lasted into the afternoon 1. A system that manages EC2 servers entered what AWS called “congestive collapse” while recovering, and health checks on its network load balancers failed intermittently; EC2 was fully normal at 1:50 PM Pacific time 1.
AWS said it had switched off the DNS automation worldwide and would fix the race condition and add protections before turning it back on 1. “We apologize for the impact this event caused our customers,” it wrote 1. Reuters, in a report carried by AAP, described the outage as at least the third time in five years that US-EAST-1 had contributed to a major internet outage 2.
The cyber-risk modeller CyberCube put the preliminary insured loss at $38 million to $581 million, expected to fall towards the lower end, and estimated that nearly 70,000 organisations were affected 6.
Microsoft: a configuration that passed the checks
Between 15:41 UTC on 29 October and 00:05 UTC on 30 October, customers and Microsoft services using Azure Front Door, Microsoft’s global content delivery network, suffered timeouts and DNS errors 3. Microsoft 365, Entra ID, Dynamics 365 and parts of the Azure Portal were affected 3. Alaska Airlines said key systems including its website were disrupted, Heathrow Airport’s website had problems and Vodafone was affected 7.
Microsoft’s post-incident review said a sequence of customer configuration changes, made across two different versions of its control software, produced incompatible metadata 3. “These customer configuration changes themselves were valid and non-malicious – however they produced metadata that, when deployed to edge site servers, exposed a latent bug in the data plane,” it wrote 3. Because the crash surfaced about five minutes after deployment, the change passed Microsoft’s safeguards and was saved as the “last known good” configuration used for rollbacks, so engineers had to edit that snapshot by hand 3.
Microsoft said the incident was not directly related to an earlier Front Door outage on 9 October, when engineers had bypassed the protection system during a manual clean-up 3. Its repair list included isolating customers from one another through “micro cell” segmentation, with an estimated completion date of June 2026 3.
How it unfolded
-
Key moment. AWS US-EAST-1 outage, 06:48–21:20 UTC.
-
Key moment. Azure Front Door outage, 15:41 UTC to 00:05 UTC next day.
-
Key moment. Cloudflare outage, 11:20–17:06 UTC; EU names 19 critical ICT providers.
-
Second Cloudflare outage, about 25 minutes.
-
Cloudflare launches “Code Orange: Fail Small”.
-
Cloudflare says the programme is complete.
-
Key moment. UK designation of AWS, Google Cloud, Microsoft and Oracle takes effect.
-
Google Cloud power loss disrupts UK financial firms.
-
Bank of England and FCA give initial response to MPs.
Cloudflare: a file that doubled in size
Cloudflare’s chief executive, Matthew Prince, wrote that significant failures began at 11:20 UTC on 18 November; core traffic was largely flowing by 14:30 and all systems were normal by 17:06 5. X and ChatGPT went down for some users, and so did Downdetector, the site people use to check for outages 4.
According to Prince, a change to permissions on a database cluster caused a query to write duplicate rows into a “feature file” used by Cloudflare’s bot-management system 5. The file doubled in size and exceeded a limit of 200 features in the software that handles traffic, which then failed and returned errors 5. Engineers first suspected a large distributed denial-of-service attack, partly because Cloudflare’s separately hosted status page happened to fail at the same time 5. “The issue was not caused, directly or indirectly, by a cyber attack or malicious activity of any kind,” Prince wrote 5. He called it “Cloudflare’s worst outage since 2019” and added: “An outage like today is unacceptable” 5.
On 5 December Cloudflare failed again, for about 25 minutes, affecting customers that accounted for about 28% of the HTTP traffic it serves 8. The company said a change to how it parses request bodies, made while protecting customers against a vulnerability in React Server Components, triggered the fault 8. “Any outage of our systems is unacceptable, and we know we have let the Internet down again following the incident on November 18,” it wrote 8.
A few points of failure
Commentators pointed to how much of the internet depends on a few providers 24. Alp Toker, director of the connectivity monitor NetBlocks, said Cloudflare had become “one of the internet’s largest single points of failure” 4. “Companies are often forced to heavily rely on the likes of Cloudflare, Microsoft, and Amazon for hosting their websites and services, as there aren’t many other options,” said Jake Moore of the security firm ESET 4.
In London, the House of Commons Treasury Committee wrote to the Treasury after the AWS outage, asking why Amazon had not been designated a critical third party to the financial sector and whether key UK infrastructure was hosted abroad 9. On 18 November 2025 the EU’s three financial supervisory authorities designated 19 critical ICT providers under the Digital Operational Resilience Act, putting them under direct EU oversight; the list includes the European arms of AWS, Google Cloud and Microsoft 1011.
Where things stand
Cloudflare launched a programme called “Code Orange: Fail Small” in December 2025, saying “We failed our users and the Internet as a whole in these past two incidents” 12. On 1 May 2026 it said it had “completed the work that would have avoided” both outages, including staged roll-outs of configuration changes with automatic rollback 13.
In the UK, regulations made on 8 July 2026 designated AWS, Google Cloud, Microsoft and Oracle as critical third parties with effect from 13 July, bringing them under the oversight of the Bank of England, the Prudential Regulation Authority and the Financial Conduct Authority 1415. “AWS supports the objectives of the UK regulators to ensure a robust financial system,” the company said 16.
The regime has already been tested. On 20 October 2026 a power loss at a Google Cloud data centre in its europe-west2-b zone disrupted service for about seven hours, affecting banking and payments services at UK financial firms 17. Google attributed it to an underground cable fault and a back-up generator that failed to receive its start signal 17. In a letter dated 4 November, the Bank of England governor, Andrew Bailey, and the FCA chief executive, Nikhil Rathi, said regulators first learned of the disruption from affected firms rather than from Google, and that a fuller response would follow by 20 January 2027 17.
Why it matters
Within seven weeks in 2025, internal faults at three different providers disrupted services worldwide, and none of the three involved an attack 135. Regulators in the EU and UK have responded by placing major cloud providers under financial-sector oversight 1014. As Bailey and Rathi put it, the delivery of important services “may depend on a small number of critical external providers” 17.
Sources and references
- Official documentAmazon Web Services, “Summary of the Amazon DynamoDB Service Disruption in the Northern Virginia (US-EAST-1) Region”. Link
- ArticleAAP (Reuters reporting), “Amazon cloud struggles as outage disrupts websites”. Link
- Official documentMicrosoft Azure, “Post Incident Review (PIR) – Azure Front Door – Connectivity issues across multiple regions (YKYN-BWZ)”. Link
- ArticleThe Star (Kenya), BBC News, “Cloudflare outage takes down X and ChatGPT”, 18 November 2025. Link
- Official documentCloudflare, “Cloudflare outage on November 18, 2025”, 18 November 2025. Link
- ArticleInsurance Journal, “CyberCube: Insured Loss Estimate From AWS Outage Likely About $40M”, 27 October 2025. Link
- ArticleInsurance Journal (Reuters), “Microsoft Azure’s Services Restored After Global Outage”, 30 October 2025. Link
- Official documentCloudflare, “Cloudflare outage on December 5, 2025”, 5 December 2025. Link
- ArticleComputing, “AWS outage exposes critical cloud vulnerabilities and raises sovereignty concerns”. Link
- ReportESMA, “European Supervisory Authorities designate critical ICT third-party providers”, 18 November 2025. Link
- ReportEBA, EIOPA, ESMA, “List of designated critical ICT third-party service providers”, 18 November 2025. Link
- Official documentCloudflare, “Code Orange: Fail Small — our resilience plan following recent incidents”, 19 December 2025. Link
- Official documentCloudflare, “Code Orange: Fail Small is complete”, 1 May 2026. Link
- Official documentlegislation.gov.uk, “The Critical Third Parties (Designation) Regulations 2026 (SI 2026/777)”, 8 July 2026. Link
- ArticleThe Stack, “UK cloud hyperscalers are critical third parties subject to financial services regulation”, 10 July 2026. Link
- Official documentAmazon Web Services, “AWS designated as a critical third party to the UK financial sector”, 10 July 2026. Link
- Official documentBank of England / FCA, “Letter from the Governor to Dame Meg Hillier, 4 November 2026”, 4 November 2026. Link
Sources and references are provided for transparency and further reading. Allegations and claims are attributed to their respective sources. Triple Frontiers distinguishes reported allegations from independently established facts.
The Frontier Brief
The week’s essential international reporting and new investigations, every Friday.



