Taiwan Strait Express-1 (TSE-1): latency up to 3.2× normal, 25 Jul 2026
What we measured
| Cable | Taiwan Strait Express-1 (TSE-1) |
| Segment | Tanshui, TW → Fuzhou, CN |
| Detected | 2026-07-25 05:01 UTC |
| Resolved | 2026-08-21 11:20 UTC |
| Duration | 27d 6h |
| Observed latency | 183.4 ms |
| Baseline | 57.9 ms · 3.17× |
| Increase | +194% |
| Severity | warning |
Who is affected
Countries where Taiwan Strait Express-1 (TSE-1) comes ashore, how much of their submarine diversity this cable represents, and what our measurements into them show.
| Country | Dependence on this cable | Other submarine cables | What the measurements show |
|---|---|---|---|
| China resilience index: 100 · resilient |
1 of 24 cables (4% of submarine diversity) | APCN-2, Asia Africa Europe-1 (AAE-1), Asia Direct Cable (ADC) +20 more | median 159 ms today vs 160 ms over 30 days (99 routes) - unchanged |
| Taiwan resilience index: 100 · resilient |
1 of 21 cables (5% of submarine diversity) | APCN-2, Apricot, Asia Pacific Gateway (APG) +17 more | median 9 ms today vs 147 ms over 30 days (129 routes) - faster |
Dependence is structural (share of the country's submarine cables), not a traffic share: real traffic distribution across cables is not public.
Path analysis
Traffic changed route while latency was elevated. A reroute of this kind usually follows a fault somewhere on the original path.
Cross-probe verdict. Too few independent vantage points were watching this endpoint to reach a cross-probe verdict.
Timeline
- Alert 2026072501: warning - ratio 2.94x
- Alert #157 improving: 1/3 consecutive OK
- Alert #157 improving: 1/3 consecutive OK
- Alert #157 improving: 1/3 consecutive OK
- Alert #157 improving: 2/3 consecutive OK
- Alert #157 improving: 1/3 consecutive OK
- baseline_shift: Alert #157 Fuzhou closed as BASELINE SHIFT after 7+ days: frozen 57.9ms -> new baseline 177.0ms (current 183.4ms). Route changed, not a cable incident.
Detection method
Latency is sampled continuously from our probe network. An alert opens when round-trip time to a monitored endpoint holds above its rolling baseline, and closes automatically once it returns. Full method is described on the methodology page.