Global cloud outages highlight just how dependent internet infrastructure is on a small group of providers. Companies should prepare emergency plans to minimise financial losses and protect their businesses. In this article, you’ll also find an overview of bizarre incidents where internet outages were caused by a retiree, a tractor, or even sharks.
The year 2025 went down in history as a time when society was plagued by some of the most severe digital outages ever experienced. For hours, millions of users were unable to access the services they rely on daily, whether due to global cloud outages or malfunctioning game servers.
When Marie Louise tried to log into her Amazon Flex account last October, she realised she had a problem. As a package delivery driver, she needed an internet connection to start work, but was held up by an unexpected outage. It ended up lasting the entire morning. IT technician Peter also faced an unpleasant surprise that morning—most of the software at his company had completely stopped working. This resulted in major delays.
On that day, October 20, a problem at Amazon Web Services (AWS) took down applications, games, web platforms, and banks that depended on AWS cloud servers. The widespread outage, which lasted approximately 15 hours, affected services such as Amazon Alexa, HMRC, Canva, Vodafone, Zoom, Coinbase, Duolingo, Fortnite, Lloyds Bank, Roblox, Slack, Atlassian, and Snapchat. Amazon’s own retail tools were also unavailable.
During the incident, more than 17 million users from over 60 countries reported connection issues. It was one of the largest internet outages recorded by Downdetector, with estimated financial losses ranging from $38 million to $581 million. The outage was caused by a DNS system error that occurred at the company’s largest data centre complex in Northern Virginia.
The world didn’t have to wait long for another massive outage—by November 18, news outlets everywhere were reporting on the global network provider Cloudflare. Cloudflare is an internet infrastructure and cloud services provider that, along with several other companies, forms a key part of the internet’s nervous system. When its services fail, websites can stop working, affecting millions of people and companies—and that is exactly what happened.
Routine permission changes in the Cloudflare database unexpectedly caused a configuration error. Among the sites the public could not access were ChatGPT, X, Spotify, YouTube, and Google. The gambling site Bet365, the League of Legends and Fortnite games, and the accounting and payroll company Sage were also down. Total global economic losses amounted to at least several hundred million dollars, as thousands of revenue-generating websites were offline for more than four hours.
But Cloudflare’s misfortune didn’t end there. A second outage, which occurred on December 5, affected approximately 28% of its traffic and took LinkedIn, Zoom, Canva, Shopify, and other sites offline. This time, the cause was the incorrect deployment of firewall rules. The outage lasted half an hour, and after service was restored, Downdetector recorded more than 4,500 related reports.
In 2025, telecommunications networks in the United Kingdom, Latin America, and the Middle East also suffered significant outages. As a result, users were left without access to mobile service and high-speed internet for hours on end.
Gaming services such as the PlayStation Network also faced prolonged outages, particularly on February 7, when millions of players were unable to access them. This was the second-largest outage on record, lasting more than 24 hours worldwide. February also saw a nationwide power outage in Chile, affecting communications, infrastructure operations, transportation systems, and data centres across the region.
In June, there were widespread outages of many internet services simultaneously. Many platforms reported errors and slowdowns as the widespread outage affected traffic routing in major cloud and CDN services. People were unable to access Gmail, Discord, Google Cloud, Twitch, Spotify, and other platforms.
SpaceX’s Starlink satellite internet service was down for more than two hours on July 24. Millions of users worldwide lost their connection due to a software glitch at key ground stations.
A major power outage in San Francisco knocked out traffic lights, causing Waymo autonomous vehicles to come to a halt at intersections. This incident highlighted that even next-generation systems can face problems during infrastructure outages.
Estimates of the financial losses incurred on October 29 during the eight-hour outage of Azure and Microsoft 365/Xbox services range from $4.8 billion to $16 billion. The incident lasted 8 hours and was caused by a global configuration issue with Azure Front Door/DNS. The outages affected many companies and airlines, with more than 18,000 users reporting issues with Azure and nearly 20,000 with Microsoft 365 at the peak of the outage.
Just before the end of the year, we experienced yet another incident with massive impact. On December 25, gaming giant Steam suffered a several-hour outage that affected thousands of players worldwide. It affected not only games such as Fortnite, Rocket League, ARC Raiders, PSN, and Steam, but also related platforms, including the Epic Games Store and PlayStation Network. Unfortunately, it occurred during the Christmas holidays, when gaming activity typically spikes. “It couldn’t have come at a worse time, as the annual winter sale is currently underway and many of us were likely planning to buy games,” said Rhiannon Bevan. AWS claimed that its systems were functioning normally, so the incident was attributed to an unrelated online event.
The worst outages in Europe mirrored global trends. The PlayStation Network topped the list with 1.7 million reports, followed by Snapchat with 989,559. British operator Vodafone experienced major outages of high-speed internet and mobile services in October, resulting in 833,211 reports. WhatsApp, Spotify, and the Dutch company Odido also made the list of the most serious incidents in the region.
Although these incidents were unpleasant for many, they highlighted a much larger, overlooked problem. Given that a significant part of the global economy depends on the internet, cybersecurity experts warn that the internet’s infrastructure has become overly reliant on a few giants.
“The less diversity you have in any ecosystem, the more vulnerable you are—and there is no diversity at the top of the internet supply chain,” says Casey Oppenheim, CEO of Disconnect, a cybersecurity company. “Pick any key area of the internet, and you’ll find only a very short list of companies that control it.”
The issues at Cloudflare occurred less than a month after outages at other cloud service providers, Amazon’s AWS and Microsoft’s Azure. Together with Google Cloud, these three companies make up roughly two-thirds of the infrastructure on which the digital world relies. According to Alp Tolker, CEO of the internet watchdog Netblocks, the internet is not designed to function this way. Although these companies have decentralised infrastructure, a single weak link is enough to take their services offline.
Furthermore, companies didn’t have to use Amazon’s cloud services directly to encounter problems. They were also affected by the outage when they relied on tools that depend on AWS, such as certain identity services, communication platforms, CI/CD pipelines, monitoring tools, or billing systems.
In the event of an AWS outage, it’s also worth highlighting the risks of a “build fast, fix later” approach. Companies that stored their data on Amazon’s cloud often failed to test what backup options they had in the event of a single data centre outage.
Today, therefore, we can state with certainty that the degree of dependence on third parties is higher than most teams realise. The question, then, is not whether the cloud is powerful—it is. Rather, we should be concerned with its resilience.
Cloud and infrastructure outages are a useful reminder that system failures are inevitable, so cyber resilience should be a priority. Assume your primary services will fail and plan for continuity. We’ve prepared a summary of strategies worth including in your contingency plan.
Start by mapping your business-critical SaaS tools to their cloud providers and regions. In your business continuity and disaster recovery planning, treat these as Tier-1 external dependencies.
Your security situation is inextricably linked to your weakest supplier. Serious incidents in 2025 demonstrated that supply chain vulnerabilities are the dominant threat vector. Therefore, actively monitor the security status of every third-party system connected to your network. Start by conducting a formal security assessment of your suppliers based on objective criteria and audits. For every software purchase, request a Software Bill of Materials so you can assess the risk if a vulnerability is found in any component.
Also, isolate third-party systems. A vendor’s service connection should never have unrestricted access to your critical data. Monitor for anomalies and ensure you can immediately revoke the vendor’s access if a security breach is reported.
Especially for your most critical systems. There are alternatives for operating key business infrastructure—your company can design backup systems that activate when something goes wrong, or you can develop your own internal cloud infrastructure. Dr. Aybars Tuncdogan, a professor of digital innovation and information security at King’s College London, suggests: “We need to diversify our technological infrastructure. If we don’t rethink our architecture, we should expect further outages of this magnitude, whether due to malfunctions or targeted attacks.”
Benjamin Schilz, CEO of the communications technology company Wire, also recommends that companies increase the diversity of their suppliers: “True resilience isn’t just about preparing for redundancy; it’s about maintaining control over one’s own data. This means ensuring that organisations can continue to operate securely and independently, without being tied to a single provider or platform.”
Your architecture should include configurations for switching to a backup system across multiple regions, with tested runbooks. In the event of a primary region outage, operations must automatically switch to a backup region. If you continue relying on providers that operate in only one region, add fallback measures such as read-only modes, graceful degradation, or manual workflows so your team can keep working despite outages. Also, ensure offline or replicated access to your most critical sheets and documentation within SaaS tools such as Jira or Confluence.
Track independent telemetry data from platforms such as ThousandEyes, Kentik, or Cloudflare Radar, and integrate it into your alerting systems and dashboards. This will help you determine whether a problem is specific to your environment or affects the entire internet.
According to Professor Feng Li, Vice Dean for Research and Innovation at Bayes Business School, UCL, outages like the October AWS outage impacted a wide range of platforms—from consumer apps to financial and public services—which suggests that many organisations still underestimate the level of concentration risk in their digital infrastructure.
At the same time, he points out that this problem affects not only companies but also users: “Too many of them still view the cloud as the only reliable source, rather than an environment of shared responsibility.” According to him, governments and researchers will have to address what happens when critical digital infrastructure becomes so concentrated that a single regional failure can have global consequences.
One of the first major outages occurred in 1997, when a failure at Network Solutions Inc.—a company that issues domain names for websites—disrupted service. The outage affected approximately one million websites, which at the time accounted for a large portion of the internet.
Even the thousands of kilometres of cables laid across the ocean floor face dangers that have caused outages. They bear the bite marks of sharks, barracudas, and other fish. Such gnawing has caused problems with telephone and telegraph cables since at least 1964.
In Armenia, a Georgian retiree was all it took to cause an outage in 2011. At that time, the internet connection for 2.9 million Armenians depended on a single fibre-optic cable that ran through Georgia. The cable was accidentally cut by a 75-year-old woman digging with a spade in search of copper near the village of Ksani. “I have no idea what the internet is,” she later told reporters. In 2017, the entire country of Zimbabwe was left without internet access for half a day when a tractor severed the cable in South Africa.
When the internet went down in the Matanuska-Susitna community in Alaska in 2018, a malware attack was to blame. In one government office, employees had to dust off their typewriters to get their work done. It took ten weeks for most systems in the area to come back online.
In 2022, a quarter of Canada’s internet and telephone services were taken offline due to an outage at Rogers Communications, one of the country’s largest telecommunications providers. Emergency services were unable to receive phone calls, hospitals cancelled appointments, and businesses nationwide could not process debit card transactions.
On Friday, July 19, 2024, approximately 8.5 million computers crashed, and Microsoft’s dreaded “blue screen of death” appeared worldwide. The cause was a failed software update from CrowdStrike, a cybersecurity giant. According to IT experts, this served as a good reminder that updates should never be released on a Friday. “It will happen again, and from a technical standpoint, the fix was relatively simple for CrowdStrike. Next time, we may not be so lucky,” says Ritesh Kotak, a cybersecurity and technology analyst.
Shutting down the internet is also a common method of government censorship. According to the organisation Access Now, there have been more than 1,500 internet outages ordered by governments, militaries, and police forces since 2016. However, the view that internet access is a human right is gaining traction. “Think about everything the internet gives you access to: employment, healthcare, education, communication, business, and simply understanding the world around you. We’ve found that internet outages actually hinder humanitarian aid and prevent the documentation of crimes being committed,” says Zach Rosson, a data analyst at Access Now.
The CrowdStrike outage also overshadowed an incident that began the same day. Bangladesh faced a near-total internet blackout after the government shut down the internet in response to violent clashes between protesting students and police. According to Access Now, this is a widespread tactic used in at least 83 countries, including India, Iran, Russia, Algeria, Senegal, Tanzania, Cameroon, and Venezuela.
In this context, it is worth mentioning a report compiled by the Top10VPN platform. It compares the financial losses incurred by countries due to government-imposed internet shutdowns in 2025. The internet outage in Russia tops the list and accounted for more than half of the total global economic impact recorded last year—specifically, $11.9 billion (USD).
The Russian outage is considered “unprecedented in terms of both scale and technical sophistication.” It combines nationwide shutdowns, selective service restrictions, and targeted interference with protocols. Although the official justification is the threat of drones, “these measures reflect a broader shift toward infrastructure-level censorship.” The goal is reportedly to suppress access to information, weaken economic activity, and make circumvention more difficult.
Also drawing attention is the nationwide internet shutdown in Iran, which began on January 8 and is still ongoing at the time of this article’s publication. The nonprofit organisation NetBlocks, which monitors internet censorship worldwide, claims that this move “was implemented by the authorities in an effort to suppress widespread protests while also covering up reports of the regime’s brutality.” Simon Migliano, who compiled the Top10VPN report, points out that the cost of this approach is skyrocketing: “Every hour that the internet is down costs the Iranian economy $1.56 million.”
If you, too, are thinking about enhancing the protection of sensitive data in your company, you can find all the useful information about freelance IT specialists here.
However, if you’re a freelancer looking to protect companies from sophisticated hacker attacks, check out the top-notch projects in our portfolio or sign up to receive tailored IT project offers.
Titans that have
joined us
Clients that have
joined us
Succcessfully supplied
man-days
The supported funding benefits from European Union support under the Equity Facility for Growth established under Regulation (EU) No 1287/2013 of the European Parliament and of the Council establishing a Programme for the Competitiveness of Enterprises and small and medium-sized enterprises (COSME) (2014-2020)
