Undersea cable cuts and why offshore hosting fails Vietnamese users

If you are hosting a website, application, or gaming server that serves users in Vietnam, the physical location of your server matters more than almost any other infrastructure choice. Most developers outside the region default to Singapore, Hong Kong, or sometimes Japan, proximity seems fine on a map. But Vietnam runs on a handful of submarine cables that snap with alarming regularity. When they do, your offshore server becomes a latency nightmare, and your users feel it immediately.

越南国际链路依赖少数海底光缆，中断时境外主机最先受影响。

Vietnam's international links run on a handful of submarine cables, and offshore hosting feels a cut first.

This article explains exactly how Vietnam's vietnam undersea cable hosting reality works, why offshore hosting is a gamble, and why putting a server physically inside Vietnam, even a modest one, is the only reliable choice for serving this market.

Key takeaways

• Vietnam's international internet capacity runs on 4-5 submarine cables (AAG, APG, AAE-1, IA, SJC2), cuts are seasonal and cause 200-400ms latency spikes.
• Offshore hosting in Singapore drops to 15-20 KB/s throughput during a major cut. Local hosting stays at 5-15ms for domestic users.
• Each cable repair takes 2-6 weeks. A single major cut affects 40-60% of total international bandwidth.
• No international route is "redundant" enough, the traffic simply queues onto the remaining cables, degrading everyone.

Why Vietnamese undersea cables matter to your hosting choice

Vietnam is not connected to the global internet by a fat terrestrial fiber pipe. It is an island of connectivity served entirely by submarine cables landing at points in Da Nang, Vung Tau, and Ho Chi Minh City. The current cable set includes:

• AAG (Asia-America Gateway), the oldest workhorse, connecting Vietnam to the US via Southeast Asia. Frequently damaged by ship anchors and seismic activity.
• APG (Asia Pacific Gateway), high-capacity, connects to Japan, Singapore, and the US.
• AAE-1 (Asia-Africa-Europe-1), the newest major cable, linking Vietnam to Europe via the Middle East.
• IA (Intra-Asia), regional capacity to Singapore, Japan, and Hong Kong.
• SJC2 (Southeast Asia-Japan Cable 2), newer, adds limited redundancy.

These cables are built by consortia of carriers including VNPT (Vietnam Posts and Telecommunications Group), Viettel (the military-run telecom giant), and FPT Telecom (the largest private ISP). Each carrier buys capacity on these cables, they do not own repeaters or landing stations. When a cable snaps, all carriers share the pain equally.

The problem: cable cuts are not rare events

This is not hypothetical. Cable cuts happen every year, often multiple times. A single segment break can take out 40-60% of total international bandwidth to or from Vietnam. During repairs (which take 2 to 6 weeks because a repair ship must sail to the site), all remaining traffic squeezes onto the intact cables. Latency to Singapore jumps from the usual 25ms to 200-400ms. Packet loss climbs above 10%. Throughput for a hosted offshore server can drop below 20 KB/s.

The most recent major incident (Q1 2026) saw the AAG and APG both suffered separate breaks within two weeks. Total external capacity dropped by roughly 50% for almost a month. Users trying to reach a Singapore-hosted site from Vietnam experienced timeouts and partial page loads. Domestic traffic was unaffected, VNPT, Viettel, and FPT kept their internal routes running normally.

This is the core distinction you must understand: traffic inside Vietnam never touches a submarine cable. Domestic peering is fast, stable, and unlimited. The problem only hits when your server is outside the country.

Why Singapore, Hong Kong, or Japan hosting fails

A developer in the US or Europe sees Singapore on a map, 1,500 km from Vietnam, and assumes latency will be 20-30ms. In theory, it is. The major carriers (VNPT, Viettel, FPT) all have direct submarine cable links to Singapore. Under normal circumstances, a server in Singapore delivers 25-35ms round-trip time (RTT) to a user in Ho Chi Minh City. That is fine for most applications.

But the cables that carry that traffic are the same shared assets described above. When a cable breaks, the direct Singapore route degrades or fails entirely. Traffic is rerouted through the remaining cables, often via Japan or the US first, creating a long, congested path. RTT jumps to 250-400ms. TCP window scaling cannot compensate. TLS handshakes add 500ms per connection. Your "fast" web app becomes unusable.

Hong Kong is worse, it has fewer direct cables to Vietnam and relies on backhaul through Singapore or Japan. Japan is better in absolute distance but often routed through multiple cable segments, making it equally vulnerable to chain failures.

The same logic applies to any offshore location: every byte crosses a submarine cable. Every cable is a single point of failure. No amount of BGP tuning on your side fixes the physical break 200 km offshore.

What "local hosting" actually means for performance

Hosting your service inside Vietnam on a VPS with an IPv4 address in Vietnam changes the equation entirely. Domestic traffic stays within the country's fiber backbone, operated by VNPT, Viettel, and FPT. It does not leave Vietnam's borders. So:

• Latency: 5-15ms from anywhere in Vietnam, regardless of cable status.
• Packet loss: near-zero on domestic backbone links.
• Throughput: limited only by your VPS plan and the ISP peering, typically hundreds of Mbps for domestic transfers.
• No single cable dependency: your server works the same during a cable cut as during normal operation.

For international users, and you, the developer, the server still connects out to the global internet via the same cables. That means you may see slower connections from your own location during a cut. But your Vietnamese users see full speed. For a Vietnamese audience, that is the only metric that matters.

Addressing the "redundancy" counterargument

Experienced ops engineers will say: "Use multiple offshore locations. Route through Singapore and Japan with BGP failover. Use a CDN. Implement fallback DNS." These are all valid for global redundancy but they miss the core point for the Vietnam case.

BGP failover between Singapore and Japan means both paths use submarine cables. If the South China Sea segment is the broken one, both paths are affected. A CDN node in Singapore still cannot reach its origin faster than the cable allows. DNS failover takes minutes and does not help existing TCP connections.

The only genuine redundancy for Vietnamese users is a server physically inside Vietnam. That is the single path that is not submarine-dependent. Everything else is a shared risk.

When offshore hosting is acceptable

This is not a universal mandate to move every workload to Vietnam. There are valid cases for offshore:

• Your audience is global with Vietnamese users as a minority, the compromise may be acceptable.
• You need heavy compute or storage that a local VPS cannot provide (unlikely for most workloads, 16GB RAM, NVMe, and 8 vCPU are available).
• You run an API that must be near a specific offshore cloud service (AWS S3 us-east-1, for example). You would then accept the cable risk.

But if your primary market is Vietnam, or Vietnamese users make up a significant portion of your traffic, a local server is not optional, it is the only way to guarantee performance during the recurring cable cuts that define Vietnam's international connectivity.

What to look for in a Vietnam hosting provider

If you decide to move to a VPS in Vietnam, check these specifics before committing:

• Domestic bandwidth, is the 1 Gbps port shared or dedicated? Do they throttle domestic traffic? Most providers do not.
• IPv4 in Vietnam, ensure the IP address is geolocated in Vietnam. Some resellers hand out IPs from other countries.
• Full root access, you need to configure the OS yourself (or choose a pre-installed image). Avoid "managed" panels that lock you out of SSH.
• KVM virtualization, no OpenVZ or shared-kernel containers. KVM gives you real kernel control and performance isolation.
• Data center location, ideally Ho Chi Minh City (HCMC) or Hanoi. HCMC is better for southern users, Hanoi for northern. Tier 3 data centers (Viettel IDC, VNPT IDC) are the standard.

A provider like thueVPS (BizMaC) offers exactly this: NVMe SSD VPS with a dedicated Vietnam IPv4, KVM virtualization, full root access, and monthly billing. You install whatever OS you need, Ubuntu 24.04, Debian 12, AlmaLinux 9, or Windows Server, and configure it as you would any cloud server. The critical difference is that the data path to your Vietnamese users never touches a submarine cable.

Practical verification: test before you migrate

Before moving a production workload, test the difference yourself:

• Ping test: from a Vietnam-based tool (ping.pe, or run from a friend's Vietnam IP), measure RTT to your offshore server and to your candidate local VPS. Expect ~30ms vs ~5ms.
• MTR during a cable incident: run MTR to both servers during a known cable cut. The offshore path will show packet loss at the cable segment. The local path will show none.
• Load test: simulate 100 concurrent users from Vietnam (use a load testing tool). The offshore server will fail or degrade under load during repairs. The local server will not.

If you cannot do a load test, at minimum run a traceroute from a Vietnam origin. If your traffic leaves the country to transit through another Asian hub, you have a single point of failure.

FAQ

Related articles

• Vietnam VPS vs Singapore VPS for serving Vietnamese users
• Why a Vietnam IPv4 matters when your users are in Vietnam
• Does Decree 53 require my company to host data in Vietnam
• Migrate a website to a new VPS with zero downtime