Australia-Japan Cable (AJC)

Based on 36 RIPE Atlas measurements from GeoCables monitoring infrastructure, March–April 2026.

The Australia-Japan Cable, generally referred to in carrier documentation as AJC, is a 12,700-kilometre transpacific submarine cable connecting Australia and Japan via a relay landing in Guam. It came into service in 2001 — twenty-four years ago — and remains in active commercial operation under a consortium of five major carriers: AT&T, NTT, Softbank, Telstra, and Verizon. The cable was originally equipped at 80 Gbit/s out of a 640-Gbit/s design capacity, then upgraded in 2008 to 240 Gbit/s of equipped capacity against a raised design ceiling of 1,000 Gbit/s. Further capacity has been added since. AJC lands at six points across three jurisdictions: Oxford Falls and Paddington on the Sydney coast in New South Wales; Tanguisson Point and Tumon Bay on Guam; and Maruyama together with Shima on Japan's Pacific coast.

The Guam relay is the structural fact that defines AJC. Unlike point-to-point transpacific cables such as JUNO, which connect a US mainland landing directly to Japan in a single fibre run, AJC is built as two segments meeting in the middle of the Pacific. The southern segment connects the Sydney landings to Guam; the northern segment connects Guam to the Japanese landings. Either segment can be used independently, and traffic between Australia and Japan can in principle stop at Guam for regeneration, switching, or off-ramping into the Pacific island networks that Guam serves. This three-region topology was the standard pattern for transpacific cables of AJC's generation, and it is one of the reasons the cable has remained operationally relevant after a quarter-century: the Guam landings provide value as an intermediate destination, not just as a relay.

Below the system floor: 0.935× of theoretical

The minimum round-trip we observe between Paddington in Sydney and Maruyama in Mie Prefecture is 116.16 ms. The physics floor for the cable's full 12,700-km length — the smallest possible round-trip if a photon traversed the entire cable end to end at the speed of light in fibre — is 124.29 ms. We measure 0.935× of that floor. Below the floor.

That number does not violate physics. It tells us, as the same kind of below-floor measurement on EXA North and South and ARCOS-1 tells us, that the route the packet actually takes is shorter than the cable's nominal advertised length. AJC's nominal 12,700 km includes both the southern segment (Sydney to Guam) and the northern segment (Guam to Japan). The Sydney-to-Maruyama corridor uses both, but it does so along a route that is slightly shorter than the cable's plotted geographic path — partly because the great-circle distance between Sydney and Maruyama via Guam is itself shorter than the laid fibre, and partly because at this latitude the fibre route deviates around bathymetric obstacles that the great-circle line does not. The 0.935× multiplier is the natural consequence of measuring against a system-wide length that overstates the route length for any specific origin-destination pair. ARCOS at 0.613× describes the same phenomenon at the structural extreme; AJC at 0.935× describes a milder version of it on a simpler two-segment cable.

Two directions, two completely different paths

The forward direction — Paddington to Maruyama — gives a striking measurement profile across 26 samples. The minimum round-trip is 116.16 ms, the average is 144.75 ms, and the maximum reaches 263.41 ms. The standard deviation is 29.66 ms. Those numbers describe a path that is doing well in the best case and badly in the worst case, with most observations scattered across a 60-millisecond corridor between the two extremes. The traceroute trace in this direction is short — only three hops median, suggesting the Australian carrier we measure from delivers packets directly into the cable terminal with very little intervening transit.

The reverse direction — Maruyama to Paddington — looks like a different cable. Across 11 samples, the round-trip averages 156.29 ms, with a minimum of 154.31 ms, a maximum of 157.74 ms, and a standard deviation of 1.38 ms. The traceroute median in this direction is 21 hops — far more transit AS-level handoffs than in the forward direction, but a path that is consistent from one measurement to the next. The two directions therefore differ both in absolute latency (an average of 38 ms higher in the reverse direction) and in how they are routed (three hops out of Sydney versus 21 hops out of Japan).

This is asymmetry by routing decision rather than by cable behaviour. Submarine fibre is symmetric: light travels at the same speed in both directions, and the wet plant introduces no direction-dependent latency at this scale. The asymmetry is introduced in BGP — by the policy choices each side's carrier makes about how to reach the other coast. The Australian originator we measure from clearly sends Japan-bound traffic straight onto AJC's southern segment, picks up the Guam relay, and delivers via AJC's northern segment, with no detour. The Japanese originator does not commit traffic to AJC the same way: the high hop count and the consistent 38-ms additional latency suggest the return path is not retracing AJC at all, but is instead routing through a different transpacific cable entirely — most likely an alternate path via the major Japanese peering points and one of the newer US-Japan cables, before crossing the Pacific via a route that does not pass through Guam.

This kind of split-path behaviour is normal on multi-segment relay cables. Matrix Cable System shows a structurally similar story between Jakarta and Singapore, where one direction commits to the Indonesian carrier's owned cable and the other direction is load-balanced across alternatives. AJC sits in the same family pattern, but on a much longer route and with more decades of accumulated carrier history shaping the routing decisions.

Twenty-four years of transpacific continuity

AJC entered service in the same year as ARCOS-1 in the Caribbean and EXA North and South across the Atlantic. All three cables are still in commercial operation, which makes 2001 a notable vintage of submarine infrastructure: cables built that year are now the senior elements of their respective oceans, often carrying traffic alongside cables a quarter-century younger. AJC's continued operation is itself the headline. The 2008 capacity upgrade from 80 Gbit/s to 240 Gbit/s of equipped bandwidth was an early sign that the consortium intended to keep the cable in service well beyond its original commercial assumptions, and the fact that the system reads 0.935× of its physics floor on the Sydney-Maruyama corridor today suggests the wet plant remains in the kind of working order that the original engineering would have predicted.

The choice of Maruyama and Shima as Japanese landings places AJC into the same coastal cluster that hosts most of Japan's transpacific cables. Japan's seventy-plus submarine cable landing stations are concentrated along three coastal regions, and the Mie Prefecture stretch is the principal Pacific-facing one. The same backhaul that takes JUNO's 2025-vintage traffic into Tokyo also takes AJC's 2001-vintage traffic into Tokyo; the difference is that AJC has been doing it for almost a generation longer.

What 0.935× means for the corridor

For users and operators, the practical effect of AJC's slightly-below-floor measurement is that the cable continues to deliver near-optimal Sydney-Japan latency twenty-four years after entering service. The Sydney-direction minimum of 116.16 ms is competitive with what the newer Pacific cables produce on the same corridor, even though those cables benefit from improved optical equipment, denser fibre counts, and shorter routes than AJC's Guam-relay topology offers. Gondwana-1 between New Caledonia and Sydney sits at 1.081× of its single-trunk floor — a much shorter cable, much closer to its theoretical limit. JUNO sits at 1.010× of its single-trunk transpacific floor — newer, longer, and almost exactly at the physical limit. AJC sits at 0.935× of a multi-segment floor, which describes a different geometry but a comparable level of operational discipline.

What we measure on AJC — 116 ms minimum to Maruyama via the Guam relay, with directional asymmetry that says more about routing choice than about cable health — is what a long-serving multi-segment transpacific cable looks like in 2026 when its consortium has kept upgrading it. The 24-year-old wet plant continues to carry the corridor's traffic, the Sydney-direction route hugs the geometric shortcut its three-segment topology offers, and the Japan-direction route uses other cables to make the return trip. We will continue to track both.