Southeast Asia-Japan Cable 2 (SJC2)

Southeast Asia Japan Cable 2 (SJC2) is a submarine cable ready for service in 2025, connecting ten landing stations across eight jurisdictions in East and Southeast Asia: Busan in South Korea, Chikura and Shima in Japan, Fangshan and Tanshui in Taiwan, Chung Hom Kok in Hong Kong, Lingang in mainland China, Quy Nhon in Vietnam, Songkhla in Thailand, and Changi South in Singapore. The cable has a design capacity of 144 Tbps along a trunk of roughly 8,900 km — making it one of the densest intra-Asian systems now in service.

SJC2 is the successor to the original SJC (2013) and follows a similar geographic spine: the Asian mainland coastline from Singapore up through Hong Kong and around to Japan. Unlike its predecessor, SJC2 was designed from the outset for modern coherent transponders and a branching architecture that lets wavelengths be routed flexibly between any two landings. It is a carrier consortium cable — owned by SoftBank, Chunghwa Telecom, Chuan Wei, Donghwa Telecom, Meta, and several other Asian telecoms — providing shared infrastructure rather than a single-sponsor private pipe.

68.5 ms between Singapore and Japan

Our monitor measures SJC2 between Changi South in Singapore and Chikura on Japan's Pacific coast. Over 30 days we collected 35 samples, both directions:

The reverse direction — Chikura to Changi South — is remarkably tight. Six samples across four days with a 7.7 ms spread. Both directions land within 6 ms of each other on their minimums, showing near-symmetric routing between Japan and Singapore along SJC2's trunk.

The forward direction is higher variance because the probe near Changi is using a target that reaches different endpoints at different times. On clean days, minimum RTT drops to 68.5 ms; on busier days, the path lands slightly further into Japan and the latency stretches to 120+ ms.

Below-floor on paper

SJC2's total length is listed at 10,500 km in our dataset (8,900 km per Submarine Networks' catalogue — these differ because the cable's Asian-mainland portion can be counted or excluded depending on which measurements are considered). The theoretical physics floor for a 10,500 km round-trip is 102.6 ms; for 8,900 km it is 87.0 ms.

We measure 68.5 ms minimum on our path. That is below both floors. Not impossible — it simply means the Singapore-to-Japan segment we are measuring does not traverse the cable's full length. A packet from Changi takes the direct route up through SJC2's trunk to Chikura, bypassing the cable's Vietnamese, Thai, and Chinese landings rather than visiting each one.

The great-circle distance from Changi South to Chikura is approximately 5,300 km. At 68.5 ms round-trip, the fibre path is about 7,000 km — roughly 32% longer than the great-circle. That extra fibre reflects the actual cable route, which curves east around the South China Sea and up along the Asian coast rather than taking the straight underwater path (which would require deeper-water cable construction and different repair access).

At 1.32× the great-circle floor, SJC2's Singapore → Japan segment delivers among the tightest latency performances we have documented for any 5,000+ km submarine path. Comparable numbers on intra-regional cables elsewhere: APRICOT's Japan → Indonesia at 1.57× (its great-circle), and this type of performance is increasingly the standard for post-2024 hyperscaler and modern consortium cables.

Ten landings through a politically sensitive route

The landing list is notable for including Hong Kong, mainland China, and Taiwan. This is a set of jurisdictions that some other recent cables have avoided: PLCN, for example, had its Hong Kong segment withdrawn under US regulatory review. SJC2 is different because it does not land in the United States — it is an intra-Asian cable, so it is not subject to the Team Telecom review process that affects US-landing systems.

For Asian traffic, SJC2's Hong Kong and mainland China landings are valuable: Hong Kong remains one of Asia's largest peering hubs, with extensive interconnection to Chinese mainland operators, and direct connectivity between Hong Kong and Singapore or Japan via SJC2 shortens latency for regional cloud services considerably compared to routing through secondary hubs.

The Quy Nhon landing in Vietnam is worth noticing. Vietnam has been investing heavily in cable landings along its coastline, with multiple cables either operational or planned for the Vietnamese coast. SJC2's Quy Nhon landing plugs Vietnam into an intra-Asian backbone with direct reach to Japan and Singapore — useful for a rapidly-growing Vietnamese cloud and data-centre market.

144 Tbps and coherent economics

144 Tbps across roughly 8,900 km of cable represents about 16 Tbps per fibre pair, assuming nine fibre pairs (SJC2's exact pair count has not been publicly confirmed). That is a solid modern spec but not at the extreme ceiling — APRICOT commissioned a year earlier has 12 pairs and 290 Tbps, and 2026-era cables like Medusa push 24 pairs and 480 Tbps. SJC2 was designed with what was state-of-the-art coherent optics at the time of construction (mid-2020s), which is now just below what the latest generation achieves.

Why the moderate spec rather than the maximum? Consortium cables tend to right-size capacity to expected demand rather than maximum technical ceiling. Each co-owner pays proportional to its share of capacity. A consortium of mid-sized Asian telecoms paying for 16 Tbps per pair is a different commercial calculation than a hyperscaler trio (Google, Meta, Amazon) paying for 24 Tbps per pair on their own data-centre traffic. Both designs work; they are optimised for different economic realities.

What our data proves

• SJC2 delivers Singapore → Japan at 68.5 ms minimum, 1.32× the great-circle physics floor. Among the cleanest intra-Asian measurements we have documented.
• Both directions align closely. Forward and reverse minimums land within 6 ms of each other, suggesting symmetric routing on the SJC2 trunk between the two endpoints.
• The measured path is shorter than the full cable length. SJC2's full multi-landing architecture is 8,900 km; the direct Changi-to-Chikura segment uses only part of it (~7,000 km of effective fibre at measured floor).

SJC2 is one of the more densely-landed cables in our dataset and one of the newest. At a year into service, it is delivering consistent performance across its Japan-Singapore segment — the segment that matters most for the broad Asian cloud and content delivery that intra-regional cables are built to serve.

Try it yourself

Live data on the SJC2 cable page. For contrast see APRICOT (2025 intra-Asian hyperscaler), JUPITER (trans-Pacific, dramatic asymmetry), and PLCN (trans-Pacific, Hong Kong segment unlit).