Operations

Vietnam submarine cable incidents and mitigation strategy

BM BizMaC Team Sep 10, 2026 11 min read

August 2026 was a rough month for anyone running services out of Vietnam. Four of the country's eight international submarine cables were hit at the same time, and international traffic slowed to the point where up to 30% of Vietnam's overseas internet capacity was affected. If your team was on a call with a client in Singapore or pushing a container image from a registry in the US, you felt it. This is what actually happens during one of these incidents, what it costs, and how to design around it.

  • Four of eight international submarine cables serving Vietnam were disrupted simultaneously in August 2026.
  • Up to 30% of Vietnam's international internet traffic is affected during a major cable incident.
  • Repairing one subsea cable costs roughly $500,000 to $1 million per incident, and ships are scarce.
  • Redundancy comes from route diversity, not from buying a bigger pipe on the same cable.

What actually breaks during a Vietnam submarine cable incident

A submarine cable incident is a physical cut, shunt fault or power feed failure on one of the fiber-optic cables landing in Vietnam, usually caused by a fishing trawler or a ship dragging an anchor across the cable shelf. Traffic that was routed over that cable is automatically rerouted by carriers onto remaining international links, which is exactly why the whole country feels slower instead of just the customers of one ISP. Nobody loses connectivity outright. Everyone loses headroom.

The symptom profile is consistent across incidents. Domestic traffic inside Vietnam stays fast, because domestic routes do not leave the country. Anything crossing international transit gets congested: packet loss during peak hours, latency spikes, TLS handshakes that suddenly take two seconds, pushes to GitHub timing out. That is the operational reality. It is not a total blackout, it is a degraded international path with unpredictable quality for days or weeks until the repair ship arrives.

海底光缆中断时，国内流量正常，国际流量明显变慢。

When a submarine cable is cut, domestic traffic runs normally while international traffic slows noticeably.

Why repairs take so long (and cost so much)

A subsea fiber repair costs between $500,000 and $1 million per incident, and that is before you count the downtime. Repair ships are a global, thinly distributed resource. A new cable repair vessel runs $60 million to $100 million to build, so there is no rapid scaling of capacity when several cables break in the same region. If the break is in deep water rather than on the continental shelf, the ship has to locate the fault, cut, retrieve, splice and re-lay the section, and then the cable has to be re-tested before it carries production traffic again.

Vietnam's cable roster has also been thinning at the top end. The SEA-ME-WE 3 cable, long a workhorse for the region, was retired in December 2024, so the remaining systems carry the same traffic load with less redundancy than a few years ago. That is a structural problem, not a bad-luck problem: each retirement removes a fallback path.

Practically, this means you should never plan on "the cable will be fixed in a few days". Plan on weeks. Build your incident response around a multi-week degradation, and anything better than that is a bonus.

What a cable outage does to a real business

The damage is uneven, and that is the useful part. If your users, your database and your application servers are all inside Vietnam, and traffic never crosses an international link, a submarine cable incident is a headline you read, not an outage you experience. Domestic bandwidth keeps working. A VPS with a dedicated IPv4 located in Vietnam serving Vietnamese users over domestic routing does not care that four cables are down.

The businesses that get hurt are the ones with a hard dependency on something overseas. Pulling container images, syncing with a SaaS API, running CI/CD against a remote registry, sending mail through an offshore relay, replicating backups to a foreign bucket, video calls with foreign clients. In August 2026 the pattern repeated: teams on a Linux VPS hosted in Vietnam with an application that only served local customers barely noticed, while teams whose build pipelines pulled from abroad lost a day or more. Cross-border e-commerce operations that hold inventory in Vietnam but run their storefront through offshore platforms took the worst of it.

Mitigation strategy: route diversity, not a bigger pipe

There is one principle that separates teams that survive a cable incident from teams that do not: redundancy has to sit on a different route. Upgrading a 100 Mbps link to 1 Gbps on the same carrier and the same cable does nothing. When that cable is cut, the extra capacity is gone with it.

  • Diversify carriers, not just plans. Two uplinks from the same ISP often ride the same international cable. Use a second provider whose traffic leaves Vietnam on a different system.
  • Cache what you fetch across borders. Local mirrors for container registries, a local package cache, local object storage. Every cached byte is a byte that does not need international transit during an incident.
  • Keep latency-sensitive workloads inside Vietnam. Databases, APIs and anything serving Vietnamese users should be reachable over domestic routing.
  • Fail over DNS and origin deliberately. A tested plan to shift a public service to an offshore region, and to shift it back, is worth more than any single redundant link.
  • Have a satellite fallback for emergencies. Not for production throughput, but for control-plane access and critical management traffic.

Satellite and onshore capacity: what changed in 2026

Satellite is no longer purely theoretical as a fallback in Vietnam. Starlink received licenses to operate in the country in February 2026, with a trial period running to the end of 2030 and a cap of 600,000 user terminals. Residential plans sit around VND 1.13 million (roughly $43.4) per month with hardware around $329. That is not a replacement for a datacenter uplink, and it should not be sold to you as one, but it is a genuinely useful emergency path for management access, DNS control and out-of-band recovery when the terrestrial international links are degraded.

More importantly, the onshore side is being rebuilt. Vietnam's target is at least 10 new undersea cable routes by 2030, bringing the total to no fewer than 15 systems with a minimum combined capacity of 350 Tbps. The region as a whole is committing around USD 11 billion between 2026 and 2035 to subsea cable expansion, and ASEAN guidelines published in 2026 push international best practice on cable protection, landing-station security and repair response. Cybersecurity at cable landing stations and in network management systems has moved up the agenda too, because a landing station is now understood as critical infrastructure.

None of that improves your connectivity this quarter. What improves your connectivity this quarter is knowing which of your own traffic crosses a border, and having a plan for when it degrades.

Building the fallback on your own infrastructure

The most controllable part of the whole problem is what runs on your servers. If your web front end, your application and your database all live on infrastructure inside Vietnam, and the only thing that needs to cross the border is an outbound API call, then a cable incident moves you from "everything is broken" to "some deploys are slow this week". The blast radius shrinks because the surface is small.

That is the argument for keeping the latency-sensitive core local. A Vietnam VPS with NVMe storage and a Vietnam IPv4 gives Vietnamese users a short path and removes the international hop from the request that matters. Pair it with local caching for the outbound dependencies and you have converted a country-wide incident into a manageable nuisance. If you are running heavier workloads, a dedicated server or colocation gives you the same domestic positioning with more headroom, and more room to host a mirror of what you would otherwise fetch across a degraded link.

Two smaller habits pay off repeatedly. First, keep your own mirror of anything versioned that your builds depend on: container base images, package repositories, SDK bundles. Second, make sure monitoring measures international latency as a first-class signal, not just domestic uptime, so you know the difference between a slow app and a cut cable within minutes rather than hours.

Troubleshooting when you suspect a cable incident

Before you spend an hour debugging your application, prove that the problem is outside your network. The commands below run on any Linux VPS and take under a minute.

# Compare latency to a domestic target vs an international one
ping -c 5 vnpt.vn
ping -c 5 8.8.8.8

# Trace where the path leaves the country and where it stalls
traceroute -n 1.1.1.1

# Check packet loss to an overseas service over 30 seconds
mtr --report --report-cycles 30 github.com

Interpretation matters more than the tools. If domestic latency and loss are clean but the international trace shows loss starting at the first hop outside Vietnam, your provider is congested and it is probably an international link incident. If loss starts inside your own VPS or at your local gateway, the problem is yours. Also check whether the degradation is peak-hour only. Cable congestion is typically worst during Vietnamese business hours and improves overnight, which is a strong signal that it is transit, not your server.

For a sustained incident, cache aggressively, defer non-urgent cross-border jobs to off-peak hours, and fail over the specific services that cannot wait. Do not restart the whole stack hoping it clears. It will not.

FAQ

How often do Vietnam submarine cable incidents happen?

Vietnam's international connectivity is disrupted frequently enough that it should be treated as a recurring operational event rather than a rare one. In August 2026, four of the country's eight international undersea cables were down at the same time, affecting international access to overseas services.

Does a submarine cable break take my Vietnamese website offline?

Usually not. If your server, users and data are all inside Vietnam and traffic stays on domestic routing, a cable incident does not take you offline. You get hurt when your application depends on something overseas: registries, APIs, offshore mail relays or remote backup targets.

How much does it cost to repair a submarine cable?

Subsea fiber repair runs roughly $500,000 to $1 million per incident. A new cable repair ship costs upwards of $60 million to $100 million, which is why repair capacity cannot scale quickly when several cables fail in one region, and why outages stretch into weeks.

Is Starlink a real fallback for a business in Vietnam?

It is a useful emergency path, not a production replacement. Starlink was licensed in Vietnam in February 2026 with a trial period until the end of 2030 and a cap of 600,000 terminals. Residential plans are around VND 1.13 million per month. Use it for management access and critical control traffic, not for bulk production bandwidth.

What is the fastest way to reduce exposure to a cable incident?

Move the latency-sensitive parts of your stack inside Vietnam so they use domestic routing, then cache every cross-border dependency you can. Route diversity across different carriers matters more than raw capacity, because a bigger link on the same cable disappears with the cable.

Will the new cable routes fix this by 2030?

They will reduce the probability and severity of incidents. Vietnam's plan is at least 10 new undersea routes by 2030 for a total of no fewer than 15 systems and 350 Tbps combined capacity, and Southeast Asia as a whole is committing about USD 11 billion between 2026 and 2035. That improves resilience for the region; it does not remove your need for a local fallback plan.

Related articles

越南海底光缆中断与应对策略

2026年8月，越南八条国际海底光缆中有四条同时故障，最多影响约30%的国际互联网流量，而修复一条光缆的费用在50万到100万美元之间。关键结论是：国内流量通常正常，只有跨境流量明显变慢，因此把面向越南用户的业务放在本地机房能大幅缩小影响范围。企业应做到多运营商路由冗余、在本地缓存容器镜像和软件包、把延迟敏感的服务留在越南境内，并用 Starlink 等卫星链路作为紧急管理通道，而不是生产带宽。越南计划到2030年新增至少10条海底光缆线路。

Note: This guide is for general reference. Every system and infrastructure has its own specifics, so test each step in a safe environment and consult a qualified engineer before applying it in production.