Taiwan's Ministry of Digital Affairs (MODA) announced on September 18, 2026 that the Taiwan-Matsu No. 4 subsea cable has entered operation. Chunghwa Telecom built the cable with government subsidy. It links the main island to Matsu's Dongyin, Nangan and Xiju, and MODA says it lifts total Taiwan-Matsu capacity to about 1.9 Tbps. It is a useful test case for how a government can pay for resilient infrastructure without taking control of the network.
The case for state money is strong
Start with the strongest argument for public subsidy. Matsu is a thin market of roughly 14,000 residents, and a commercial operator will not earn a return on a cable that mostly exists to be a spare. Resilience is a public good that private balance sheets underprice. If the state does not fund it, nobody does.
The record supports that argument. In February 2023, two cables serving Matsu were cut within days of each other. PBS NewsHour reported that the first cut came on February 2 and the second on February 8, and that residents went roughly a month with degraded internet. Microwave backup gave slow text and dropped calls, and video was unwatchable. The same report noted 27 cuts to Taiwan's cables in the preceding five years.
What MODA built, and what it is cautious about
The outages kept coming. According to the Taipei Times, No. 3 failed on January 15, 2025 and No. 2 followed on January 22. MODA attributed both to natural deterioration rather than sabotage. A MODA release then reported No. 2 completely broken on February 16, 2025.
What stands out is that MODA's response was not a single cable. In that release it committed to what it called the principle of "backup and backup". The plan covered asynchronous satellite stations at key government sites, subsidies to expand microwave capacity, and construction of Taiwan-Matsu No. 4. The September launch is the delivery of the last item. A June 2026 MODA release says each of the four Matsu townships will have at least two microwave systems able to connect to the outside world.
The engineering choices as reported by SubTel Forum and MODA's announcement address the main causes of cable faults. The cable is double-armored and has a target burial depth of 2.5 metres, which protects against anchors and fishing gear. An automatic cable warning system that combines AI recognition with AIS ship-tracking data works with the Coast Guard to flag vessels behaving dangerously near a route.
Why this is the right model
The regulation debate around subsea cables often swings between two bad options. One is a state-owned monopoly. The other is a laissez-faire approach that assumes the market will fund redundancy nobody can monetize. Taiwan's approach sits in between, and the design has four virtues.
- Private operation, public purpose. Chunghwa Telecom builds and runs the cable. The subsidy pays for the resilience premium, not for control of traffic.
- Layers, not a single point. Cables, microwave and satellite fail for different reasons. A storm, a dragged anchor and a jammed link rarely hit all three at once.
- Prevention plus recovery. Burial, armoring and the automatic warning system reduce the odds of a break. Microwave and satellite limit the damage when one occurs.
- Measurable adequacy. In the February 2025 release, MODA specified microwave backup of 12.6 Gbps against Matsu's peak daily traffic of 9.5 Gbps. Setting a target above observed demand is the kind of evidence-based standard that regulators should publish.
The contrast with 2023 is the point. Back then, backup meant a microwave link too slow for ordinary use. The 2025 benchmark of backup capacity exceeding peak demand is a different standard of service.
Where the policy should be pressed
Proportionate does not mean unexamined. Three questions deserve scrutiny.
First, the warning system combines AI classification with vessel tracking. Tracking commercial ships near cable routes is a legitimate safety use. But the rules on data retention, on who can see the alerts, and on how false positives are handled should be public. An alert that leads to the boarding or detention of a vessel is a coercive act, and it should rest on documented thresholds and not on model output alone.
Second, attribution should stay disciplined. In 2023, Taiwan's regulator blamed two Chinese vessels, but there was no direct evidence of deliberate sabotage. In 2025, MODA cited natural deterioration. That restraint is a strength. Policy built on proven causes, such as aging cables, shipping traffic and fishing gear, is easier to defend than policy built on suspicion. It also keeps the focus on engineering fixes that work whatever the cause.
Third, the subsidy should be benchmarked. Public reporting on cost per unit of resilience, repair times and uptime would let taxpayers and other governments judge whether the model is worth copying. The source materials for this article do not give a verified cost for No. 4, so that gap is itself worth closing.
The bottom line
A fourth cable does not make Matsu invulnerable, and MODA does not claim it does. Its "redundancy upon redundancy" framing is the honest one: assume some layer will fail, and make sure another carries the load. For policymakers elsewhere weighing how to protect island and remote communities, the Taiwanese template is worth studying. It uses public money where markets will not, private operators where they are competent, and published capacity targets where accountability is needed. The remaining task is to make the monitoring layer as transparent as the cable layer is robust.