One Cut, 300 Gbps Gone: How the Internet Survived the Philippines-Singapore Cable Break

On the night of October 2, 2026, a glass fiber cable lying on the seabed between the Philippines and Singapore stopped carrying data. Nobody outside a few network operations centers noticed the exact moment it failed. What millions of people noticed was everything built on top of it getting slower: video calls stuttering, overseas websites crawling, downloads dragging. PLDT, one of the country's largest carriers, told customers that the cut had taken 300 gigabits per second of international bandwidth offline.

By October 4, service was fully back to normal: InfiniVAN had restored the C2C segment on October 3, and the major carriers declared nationwide normalization a day later. Roughly 48 hours of degraded internet, then it was over. But those two days are one of the clearest recent demonstrations of how the internet survives the failure of its own foundations, and how fragile those foundations remain. Here is what happened beneath the headlines: how data crosses an ocean, how carriers move it when the shortest path disappears, and how a break in a cable gets found and fixed.

What broke: one segment of a web of glass

The failure was not the whole internet, and it was not even a whole cable. It was a segment of the C2C submarine cable system, the portion running between the Philippines and Singapore. A cable system like C2C is not a single wire but a ring or web of segments linking landing stations in different countries, so a fault takes down only the traffic routed through it. The Philippines-to-Hong Kong route was also affected at first, according to multiple reports, and the Hong Kong link was still under maintenance after the Singapore segment came back, which limited the backup options.

As of publication, the physical cause of the fault has not been confirmed. The Philippines' Department of Information and Communications Technology opened an investigation, and DICT Secretary Henry Aguda said possible causes included fiber cuts, equipment failure, or sabotage, while stressing it was too early to conclude. Most submarine cable faults are eventually attributed to mundane causes like ship anchors dragging across the seabed, fishing trawls, or earthquakes, but the investigation happens on the seafloor and takes time.

Why nobody went dark: the internet plans for failure

The reason 300 Gbps could vanish without taking the country offline is redundancy by design. Major carriers like PLDT, Globe, and Converge buy capacity on several different cable systems, not one. When a segment fails, their traffic engineers shift data onto the surviving paths. Converge rerouted Singapore-bound traffic largely through the Bifrost cable system; Globe diverted traffic through its redundancies on the night of October 2, before the full outage was even public.

Carrier-level decisions about which cable system carries which traffic can be changed in minutes, because routing information between networks is constantly recomputed. The backup routes are longer or more congested than the primary path, which is why users saw slower speeds and higher latency even though the network as a whole stayed up. The internet does not promise the fastest path. It promises any path.

The internet does not prevent cable failures. It plans for them.

What 177 milliseconds tells you

The most interesting evidence of the outage came not from a carrier but from measurement. CableStatus, monitoring the Philippines-to-Singapore route with 28 RIPE Atlas measurement probes, recorded average latency of roughly 177 milliseconds, compared with a recent baseline of about 53 milliseconds, nearly a tripling of the route's round-trip delay.

Latency is the internet's vital sign. When a packet's shortest path crosses a failed segment, the longer replacement adds distance, and distance adds delay, because signals travel at roughly two thirds of the speed of light through fiber. Congestion on the backup path adds more, as packets queue behind the rerouted traffic of an entire country. The jump from 53 to 177 milliseconds is exactly the signature you expect when a huge volume of traffic is squeezed through longer, busier alternate routes.

The measurement had limits, and its authors said so explicitly: 28 probes cannot confirm a physical cable fault or name the affected system. They can only show that the route got much slower. Other monitoring platforms showed varying dips among individual providers without registering a country-level outage. Three different kinds of evidence (a probe sample, an operator's segment status, and carrier service updates) each measured something different, which is why the "restored" announcements ran on three different clocks.

Finding and fixing a break on the ocean floor

Repairing a submarine cable is a slow, expensive maritime operation. InfiniVAN restored the C2C segment at 12:48 p.m. on October 3, which means the fault was isolated to a section that could be bypassed and reactivated at the Philippine landing station. Globe had warned that a full physical repair of subsea fiber could take weeks, the industry norm: operators first locate the break with optical time-domain reflectometry, which times a pulse of light reflected from the fault to narrow the search to a stretch of seabed, then send a repair ship to grapple the cable up, splice in a new section, and lay it back down.

The stakes are high because the world's dependency on these systems is nearly total. More than 95 percent of intercontinental internet traffic travels by submarine cable, according to data widely cited from TeleGeography. The Philippines episode was not unique: in September 2025, cuts to the SMW4 and IMEWE systems near Jeddah, Saudi Arabia, degraded connectivity across South Asia, the Gulf, and parts of Africa, with NetBlocks flagging the disruption and Microsoft warning of increased latency for some Azure traffic in the Middle East. The pattern is consistent. A cut happens, traffic reroutes, latency spikes, repairs take weeks, and the internet absorbs the blow.

Three clocks behind the word "restored"

The Philippines episode also exposed something subtle about how the internet reports its own health. There were three separate restoration stories: InfiniVAN's report that its C2C segment was restored at 12:48 p.m. on October 3; Converge's account of stabilizing service by rerouting; and Globe's October 4 statement that services were fully restored after the landing station reactivated. Each statement was true about a different object: a measured path, a named cable segment, an operator's traffic, the service a customer buys.

The distinction matters. A route can be reachable while latency stays elevated, and a carrier can restore service on an alternate path while the original cable is still being repaired. For businesses running workloads on Singapore-hosted cloud regions, payment systems, and collaboration tools, those are not the same event. The lesson of October 2: "the internet is back" is never one fact. It is a stack of separate recoveries, and the bottom of the stack is a glass thread on the ocean floor.

References

Meyka, Philippines-Singapore Cable Restored After October 2 Outage, October 2026. btw.media, Philippines cable outage: three clocks behind "restored", October 2026. Astig.ph, PH-Singapore undersea cable restored; DICT still investigating cause, October 2026. Communications Today, Philippines telcos divert data traffic after undersea cable outage, October 2026. BusinessMirror, Globe restores services as subsea cable repair finished, October 2026. TeleGeography, submarine cable data. NetBlocks and Microsoft reporting on the September 2025 Red Sea cable cuts.