Mariana-Guam Cable

Based on 224 RIPE Atlas measurements from GeoCables monitoring infrastructure, March–April 2026.

The Mariana-Guam Cable is 268 kilometres long, has four landings, and first entered service in 1997. It has one owner — PTI Pacifica, a small Pacific-focused operator with no global ambitions — and its entire commercial purpose is to provide connectivity between three islands of the Commonwealth of the Northern Mariana Islands (Saipan, Rota, Tinian) and the much larger US territory of Guam.

In 2026, it is 29 years old. Typical submarine cable design life is 25 years.

It is, nevertheless, the only submarine cable system directly connecting the Northern Mariana Islands to anything. Our measurements show it still working — and working with a degree of transit stability that would flatter a cable half its age.

What the cable actually is

MGC's four landings trace a 268 km arc along the Mariana chain: Tanguisson Point on the northwest coast of Guam, then Rota, Tinian, and Saipan as the cable moves north through the Northern Marianas. The distance from Saipan (the largest city in the CNMI) to Tanguisson Point is roughly 200 km. From Rota to Guam, barely 100 km. These are extremely short hops by submarine-cable standards — the physics floor for an end-to-end Saipan-to-Guam round-trip is just 2.62 ms.

Guam, as the landing at Tanguisson Point, is the upstream end. Guam hosts landings for Apricot, PROA, SEA-US, HANTRU-1, and several other regional systems — it is one of the most cable-dense small territories in the Pacific. Everything the Northern Mariana Islands reach beyond their own waters first travels the 268 km of MGC fibre to Guam, and then from Guam out to the world.

The local direction

From the 46 measurements in the Saipan → Tanguisson Point direction, the fastest round-trip is 8.35 ms. At fibre speed (200,000 km/s), that corresponds to roughly 835 km of optical path. The direct Saipan-Guam hop is only about 200 km, so the bulk of that time is transit-network overhead: packets entering PTI Pacifica's router at Saipan, traversing the cable, arriving at Guam, then being passed to an onward peer that can return the traffic. The 8.35 ms floor tells us MGC itself is functioning as expected — a 29-year-old cable still delivering its design-level performance for the traffic it was built to carry.

Average RTT in that direction is 96.79 ms, max 110 ms. The 11× spread between min and average is a signature of occasional re-routing — the fastest measurements use the direct path, while the rest pick up additional latency somewhere in Guam's transit mesh. None of this suggests a fault in the cable itself.

The long-haul directions

The remaining 178 measurements come from probes in Minsk, Tbilisi, Jerusalem, Sevastopol, and Almaty, all targeting the same Tanguisson Point endpoint on Guam. These measurements don't directly traverse MGC — they reach Guam through Pacific backbones (Unity/EAC-Pacific, SEA-US, and others) and then use Guam's terrestrial transit before finally stepping onto MGC to reach the CNMI. But Tanguisson Point itself is an MGC landing, so each of these measurements terminates there — and together they paint a picture of how stable the Pacific-to-CNMI journey actually is in 2026.

The Almaty→Tanguisson route holds its RTT within 0.67 ms across 36 measurements — a total observed spread of 3.5 ms on a 349 ms baseline. That is an extraordinary level of routing stability. Jerusalem is similar, at 1.01 ms standard deviation. Tbilisi comes in at 6.5 ms, Sevastopol at 3.6 ms. Only Minsk shows looser variance (15.9 ms), reflecting a more contested transit path with occasional re-routing.

These tight variances tell us something useful: the transit providers carrying these paths have converged on stable BGP preferences through the Pacific. A packet leaving Almaty travels through an expected sequence of transit operators, reaches the Pacific at a consistent entry point, crosses one of the major trans-Pacific systems, arrives in Guam at a specific peering point, and hands off to MGC in exactly the same way every time. For a cable whose oldest segment was lit in 1997, the global transit infrastructure feeding it behaves with remarkable predictability.

Past design life

Submarine cables are typically designed for 25 years of operation. The repeaters — the in-line amplifiers that boost the optical signal every 60–100 km of cable — have pump laser components that degrade over time; the cable itself, buried or laid on the seabed, faces abrasion, fishing-trawl damage, and thermal cycling; and the electronics at each landing station age through multiple generations of networking equipment. Past year 25, everything is running on borrowed time.

Many older cables have been quietly retired in exactly this window. Others — like Pacific Crossing 1 (1999, US-Japan), APCN-2 (2001, Asian ring), and FLAG North Asia Loop (2001) — are still running well past their design life because the traffic they carry is both economically valuable and operationally irreplaceable. MGC falls firmly into this category. There is no alternative submarine route between the CNMI and Guam. Losing the cable would mean satellite-only connectivity for 55,500 people across three islands. Keeping it alive is a PTI Pacifica responsibility that doesn't have a substitute provider waiting.

Why the CNMI doesn't have a newer cable

Several major 2020s Pacific cables — Bulikula, Apricot, Jupiter — have landed at Guam, but none has extended north into the CNMI. The commercial math is straightforward: 55,500 people across three islands is not enough demand to justify a new 200-km branch from a hyperscaler-funded cable. MGC remains commercially viable precisely because it was built in the 1997 cost environment, when small-scale Pacific cables were still buildable on the economics of regional telecoms. Today that calculation only works for replacement, not for a second operator to enter the CNMI market.

When MGC eventually does need replacement, the most plausible path is a PTI-Pacifica-led refresh of the same route with modern equipment — a new cable on the same 268 km of seabed, reusing the same landing stations and customer relationships.

What we're watching

Two things. First, whether the 8.35 ms Saipan→Tanguisson floor drifts upward over time. A gradual RTT increase on a cable this old would be the earliest signal of repeater degradation — the in-line amplifiers getting less efficient, packets needing more retiming. Second, whether the long-haul tight-variance pattern holds. Almaty's 0.67 ms standard deviation is not a fluke measurement artefact; it reflects a specific set of BGP decisions by upstream operators, and changes in those decisions would show up first as loosened variance. Either shift would tell us something concrete about the Pacific's transit infrastructure that couldn't be read off any static dataset.