Fourth-Quarter Outages: A Shift Away From Government Shutdowns
Cloudflare's network, spanning more than 330 cities in over 120 countries and interconnecting with over 13,000 network providers, offers a unique vantage point for observing Internet resilience. This summary covers observed and confirmed disruptions for the fourth quarter of 2024. It is not an exhaustive list; a broader set of detected anomalies is available in the Cloudflare Radar Outage Center.
Unlike the previous quarter, which saw numerous government-directed shutdowns often tied to exam periods, the fourth quarter saw just one such event, which was related to protests. Instead, the quarter's disruptions were driven by terrestrial cable cuts in two African nations, power grid failures including rolling outages after military action, and severe weather including storms and an earthquake. Operational issues also played a role, with maintenance problems at two European providers and a brief, unexplained outage for Verizon customers in several US states.
Simultaneous Fiber Cuts Disrupt African Connectivity
On October 1, local mobile provider MTN Rwanda (AS36890) alerted subscribers via X to a double fiber cut in Tanzania and Uganda. Traffic from Rwanda began to drop sharply after 12:45 local time (10:45 UTC), with a complete outage between 13:15 and 13:30 local time (11:15-11:30 UTC). Recovery was swift, with traffic returning to expected levels by around 19:00 local time (17:00 UTC). MTN posted a follow-up confirming all services were restored.
The African Undersea and Terrestrial Fibre Optic Cables (AfTerFibre) map shows potential backup connections through networks to the west in the Democratic Republic of the Congo (DRC). However, an examination of MTN Rwanda's connectivity suggests its upstream providers or peers may not route traffic via DRC-based networks, leaving no viable fallback path when the cuts occurred.
In a similar event, Airtel Niger (AS37531) posted a thread on X in late November apologizing for a service disruption. The provider cited a simultaneous interruption on the national optical fiber on the Niamey-Dosso and Niamey-Balleyara routes, stating services were "completely interrupted throughout the territory." The outage lasted from 17:30 local time (16:30 UTC) on November 29 until 19:45 local time (18:45 UTC) on November 30.
The timing of Airtel's public acknowledgment is notable, as the thread was posted only after service was restored. It is plausible that the provider lacked backup connectivity to post an update and used more official channels for the initial alert, with the X thread serving as a follow-up once the Internet was available.
Grid Failures Cause Prolonged Outages
Cuba's Repeated Grid Collapses
Cuba's fragile electrical infrastructure led to three nationwide power outages during the quarter. On October 18, the Ministry of Energy and Mines reported the unexpected disconnection of the Antonio Guiteras thermoelectric plant, which caused the entire National Electricity System to go offline at 11 a.m. Internet traffic within the country halved within minutes (15:15 UTC). Connectivity did not return to expected levels until around 23:00 local time on October 21 (03:00 UTC on October 22), meaning the disruption lasted approximately three and a half days.
Several weeks later, on November 6, strong winds from Hurricane Rafael caused another disconnection of the National Electric System. Traffic fell sharply around 14:30 local time (19:30 UTC). After the storm passed, the Uníon Eléctrica posted updates on restoration efforts. Internet traffic appeared to recover to expected levels around 13:00 local time (18:00 UTC) on November 9, though full electrical restoration took longer.
A third nationwide outage occurred on December 4. The Ministry reported that the electrical system (SEN) disconnected at 2:08 a.m. due to the automatic tripping at the Antonio Guiteras plant. This caused a roughly 60% drop in Internet traffic compared to the previous week. Services recovered nearly a day later, around 00:30 local time (05:30 UTC), aligning with a Ministry post announcing all units had been synchronized.
Guadeloupe and Kenya
In the French Caribbean island of Guadeloupe, a labor action on October 25 led to a complete power outage. According to a report in The Guardian, striking workers seized the power station's command room and caused an emergency shutdown of all engines. Data showed traffic dropping nearly 70% at 08:30 local time (12:30 UTC). Restoration took longer than initial estimates suggested. A government press release on October 26 noted that 160,000 users had power restored, while 70,000 remained cut off. It also stated that 76% of Orange subscribers had regained network connections, but 1,800 homes were still without Internet. Traffic did not return to expected levels until around 22:00 local time on October 26 (02:00 UTC on October 27).
Kenya experienced another power-related disruption after similar events in the second and third quarters. A December 18 "Customer Alert" from Kenya Power reported a widespread outage affecting most of the country, except parts of the Western and North Rift regions. Internet traffic dropped by over 70% starting just after midnight local time on December 18 (21:00 UTC on December 17). Kenya Power confirmed all power had been restored by 07:35 local time (04:35 UTC), and Internet traffic had recovered to near expected levels by then.
Infrastructure and Maintenance Failures
The source material also details internet disruptions caused by various other factors including military action, natural disasters, and maintenance issues. These incidents spanned multiple regions and involved different types of network infrastructure failures.
A chart displaying traffic anomalies is available below.
Hurricane Milton and other storms
Hurricane Milton made landfall in Florida as a Category 3 storm at 20:30 local time on October 9 (00:30 UTC on October 10). Flooding, downed trees, power line damage, and infrastructure destruction caused widespread disruption, and state-level Internet traffic reflected the storm’s impact: peak traffic on October 10 was roughly 40% lower than in the preceding days. As power was restored and evacuees returned, traffic rose gradually back toward typical levels.
City-level data showed the pattern of recovery clearly. At 09:00 local time (13:00 UTC) on October 10, more than 70 cities had traffic more than 50% lower than the same time the previous week; by October 14, that count was just over 10.
Cyclone Chido struck the French territory of Mayotte in the Indian Ocean on December 14, damaging power, water, and communications infrastructure as well as homes and public buildings. The storm killed more than three dozen people and injured thousands. Internet traffic dropped sharply around 09:00 local time (06:00 UTC) on December 14, causing a near-complete outage. Traffic growth was extremely slow over the next week, but a diurnal pattern gradually re-emerged, with peak levels increasing through the end of the month. As of the third week of January 2025, traffic continues to recover but remains well below pre-Chido levels.
On December 17, a magnitude 7.3 earthquake struck 24 km WNW of Port-Vila, Vanuatu at 17:46 local time (01:47 UTC). Traffic dropped nearly 90% against the previous week almost immediately, and announced IPv4 address space also fell significantly, indicating damage to core network infrastructure. Recovery was slow, with traffic not returning to expected levels until around 23:00 local time (12:00 UTC) on December 26.
Vanuatu relies on the Interchange Cable Network 1 (ICN1) submarine cable to Fiji for international connectivity. As an editorial on The Maritime Executive noted, “A fire at the cable landing station temporarily interrupted the power supply, disabling internet traffic. The connection was restored 10 days later…” The restoration timeline roughly aligns with traffic returning to normal, suggesting the cable station power loss was a major factor. Starlink service became available on October 7, offering some nominal redundancy; the TAMTAM submarine cable connecting Vanuatu to New Caledonia is expected in 2026 and would add further resilience.
Shutdowns and strikes
Mozambique saw mobile Internet shutdowns beginning October 25, after protests against the re-election of the ruling Frelimo party turned violent. Around 13:00 local time (11:00 UTC), traffic dropped sharply across AS30619 (Telecomunicações de Moçambique), AS37342 (Movitel), and AS37223 (Vodacom). Vodacom and Movitel saw near-complete outages almost immediately; some Telecomunicações de Moçambique traffic remained until just before 02:00 local time (00:00 UTC) on October 26. Connectivity returned the morning of October 26, though some social media platforms and messaging applications stayed unavailable afterward.
Another shutdown hit the same networks on November 3 at around 20:30 local time (18:30 UTC), with connectivity disrupted for nearly 12 hours before recovering around 08:00 (06:00 UTC) on November 4. Further “Internet curfews” followed: November 4-5 and November 6-7 on all three networks, and November 7-8 on Movitel and Vodacom. The country’s Minister of Transport and Communications acknowledged that access was restricted “to avoid the destruction of the country,” adding that providers can interrupt services when they detect misuse, as part of their “civil responsibility” to safeguard “the stability and welfare of the population.”
In Syria, an Internet disruption on November 9 was possibly linked to an Israeli airstrike near Aleppo and Idlib earlier that morning. Traffic fell by about 80% at around 04:00 local time (01:00 UTC), with announced IP address space dropping significantly as well. Service returned roughly four hours later, around 08:00 local time (05:00 UTC). City-level analysis showed a similar disruption in Aleppo.
Russian missile strikes on November 17 targeted electrical power infrastructure in Ukraine, causing rolling outages across multiple regions and the usual knock-on effect on Internet traffic. Between 07:30 local time (05:30 UTC) on November 17 and 02:00 local time (00:00 UTC) on November 23, traffic was lower than the prior week in Odessa, Zaporizhzhia, Mykolaiv, and Sumy. Odessa traffic dropped by around 50% on November 17, while the other regions saw drops of over 20% the following day. Odessa largely recovered by November 21; the other regions took several more days.
Further strikes came on November 28, with additional Russian airstrikes again hitting electrical infrastructure. Ukrainian officials imposed emergency power outages starting around 07:00 local time (05:00 UTC), and traffic fell by as much as 65% versus the previous week in Kherson Oblast, Mykolaiv, Ternopil Oblast, Rivna, and Lviv. Traffic stayed lower for several days but generally recovered by December 1.
Maintenance and repair
Swiss provider Salt Mobile (AS15796) was taken fully offline early on December 3 by reported maintenance. The image on the provider’s homepage explained the situation.
Observed traffic sat at or near zero for nearly three hours, between 01:25 and 04:20 local time (00:25 - 03:20 UTC).
Greenland’s Tusass A/S (AS8818, formerly TeleGreenland) explained a complete outage on December 10 via an update on its site. The outage, lasting from 02:30 to 05:15 local time (04:30 - 07:15 UTC), happened because “preventive maintenance was to be done on the connections in Canada between 02:00 and 06:00 last night, but with a combined fault on our connection to Denmark we lost nationwide connectivity. Fortunately, the fault on the connection to Denmark occurred on land, and therefore easy to repair.” During the outage, traffic dropped to zero, no IPv6 address space was announced, and announced IPv4 address space fell by 94%.
According to TeleGeography's Submarine Cable Map, the Greenland Connect cable system links Greenland to Newfoundland, Canada. The fault affecting Denmark may have occurred on the Greenland-to-Iceland segment of Greenland Connect; the Iceland-to-Denmark leg runs over the DANICE submarine cable.
Verizon Fios: Short Outage, No Explanation
In the pre-dawn hours of November 12, a significant portion of Verizon's Fios Internet service went dark. A report on the Outages mailing list flagged the start of a multi-state disruption at 12:28am EST, affecting customers in Virginia, Washington DC, Maryland, New Jersey, and parts of eastern Pennsylvania.
The impact was visible at the network level. Traffic on AS701, the autonomous system Verizon uses for Fios, fell by roughly 30% starting around 00:30 Eastern (05:30 UTC). In specific states—Pennsylvania, Delaware, Maryland, and Washington DC—the drop was more pronounced, with traffic down between 50% and 70%.
The issue was short-lived. A follow-up post to the mailing list indicated service was fully restored by 3:23am EST (08:23 UTC). Several hours later, Verizon Support acknowledged the incident on X, stating, "A network issue early this morning disrupted service for some Verizon Fios customers in the Northeast for a short period of time. As soon as the issue was identified, our engineering teams quickly restored the service." The company did not, however, offer any detail on the root cause.
Resilience Beyond the Headlines
The quarter's more dramatic failures—storms, infrastructure damage, power loss—make the case for redundancy. But quiet successes matter too. Two separate cable cuts in the Baltic Sea on November 17 and 18 caused minimal disruption thanks to resilient internet connectivity. Whether those cuts were accidental or deliberate, they keep the security of subsea cable infrastructure a live concern.
Terrestrial cables and other critical internet infrastructure warrant equal attention. Storms, earthquakes, military action, and power outages will keep testing those systems; preparation is the only way to speed recovery when they fail.
