Proa

Proa is a 2,891 km submarine cable ready for service in 2026, connecting Japan and Guam with landings at Shima (Japan), Tanguisson Point and Tinian (Mariana Islands). The cable was built by NEC as system supplier, with 16 fibre pairs delivering a design capacity of 25 Tbps. It is a short cable by intercontinental standards — roughly one-seventh the length of a full trans-Pacific trunk — and it is focused specifically on the Japan-Guam corridor that has become central to modern Pacific internet architecture.

Proa is part of a quiet buildout making Guam one of the Pacific's densest cable landing hubs. Alongside JUPITER, APRICOT, Bulikula, SEA-US, and several others, Guam now anchors a meshed Pacific backbone that did not exist a decade ago. Proa's specific role is fast, dedicated Japan-to-Guam capacity — a relatively short hop that supports hyperscaler traffic between Tokyo-region data centres and the growing Guam colocation market.

48 ms forward, 130 ms reverse: an unusual asymmetry

Our monitor measures Proa between Shima in Japan and Tanguisson Point on Guam. Over 30 days we collected 50 samples, and the two directions differ dramatically:

The forward direction (Shima → Tanguisson Point) measures 47.8 ms minimum. Light in fibre has a theoretical round-trip minimum of 28.3 ms for a 2,891 km path. Our measurement is 1.69× the physics floor — reasonable for a short cable in its first year of service where backhaul at each end contributes non-trivial overhead.

The reverse direction (Tanguisson Point → Shima) measures 130 ms at best, with an average of 151 ms and a spike to 312 ms. That is 4.6× the physics floor, or equivalently about 13,300 km of fibre traversal. Not even close to a Proa-only path — the return route is using some combination of other cables to get from Guam back to Japan.

This asymmetry is structural rather than operational. A Shima probe sending a packet toward Guam routes directly onto Proa (the shortest and newest path), but a Guam probe sending a packet toward Japan may have established peering relationships with Pacific hubs that route its traffic via other cables — perhaps via Hawaii, perhaps via a longer alternative. The 4.6× multiplier on the return direction shows the packet is taking a path roughly 4.6 times the Proa fibre length to reach Japan.

Short cables serve short distances

Most of our previous cable measurements have been on systems 8,000+ km in length. Proa is dramatically shorter. This length affects its operational profile in several ways.

Latency sensitivity is lower. A 5% increase in fibre path on a 20,000 km cable translates to ~10 ms of latency — potentially visible to applications. On a 2,891 km cable, the same percentage is only 1.5 ms. Small routing variations are harder to see in latency measurements, which is partly why short cables often appear "smoother" in monitoring datasets.

Repeater count is lower. Submarine fibre repeaters sit roughly every 80 km along the cable body. Proa's 2,891 km route has approximately 35 repeaters per fibre pair; a trans-Pacific cable has 250+. Fewer repeaters mean lower aggregate failure probability per traversal, and simpler total power-feed calculations. Fresh design-life ceiling is essentially the same (25 years) but the operational reliability profile is different.

Capacity per km is higher. Proa's 25 Tbps across 2,891 km is 8.6 Gbps per kilometre of fibre. A comparable metric on JUPITER (14,557 km, 60+ Tbps) is about 4.1 Gbps per km. Short cables have better wavelength-efficiency because shorter fibre paths allow higher-order coherent modulation without degradation — more bits per Hz.

Japan-Guam as a growing market

Shima, on Japan's Mie Prefecture Pacific coast, is one of Japan's most-landed cable stations with fifteen or more systems terminating in the area. Tanguisson Point on Guam is similarly a concentrated cable-landing cluster, with most trans-Pacific systems touching Guam via this location. The Tinian landing is a smaller secondary site in the Northern Marianas — a regional branching point rather than a primary terminus.

For the Japan-Guam corridor specifically, several factors drive capacity growth:

• Hyperscaler data-centre expansion in Guam. The island has become a cost-effective location for US-territory cloud presence that sits closer to Asia than Hawaii or the US mainland.
• Trans-Pacific cable restructuring. As PLCN's Hong Kong segment was withdrawn, Guam's role as an alternate Asia-Pacific hub expanded.
• Regional redundancy. A short cable like Proa provides Guam-Japan connectivity that is physically independent of the longer trans-Pacific trunks, useful for traffic engineering during cable faults.

16 fibre pairs and a modern supplier

Proa's 16 fibre pairs mark it as a modern cable by current design standards — APRICOT (12 pairs), SJC2 (fewer pairs but high per-pair capacity), Medusa (24 pairs). NEC as system supplier is notable because NEC has competed with SubCom and ASN for submarine cable construction contracts for decades, and Proa's award represents a Japanese-built cable for a Japan-anchored route.

The 25 Tbps total design capacity is modest relative to a 24-pair cable but sensible for a 16-pair configuration where each pair carries approximately 1.5-2 Tbps at commissioning, with headroom for future transponder upgrades. The cable was designed for the Japan-Guam traffic volumes expected through 2030, not for headline capacity.

What our data proves

• Proa delivers Shima → Guam at 47.8 ms, 1.69× the physics floor. Reasonable for a short cable in its first year of service.
• Reverse direction is dramatically asymmetric at 4.6× the floor. Traffic from Guam to Japan is using non-Proa cables — a structural routing pattern that may normalize as carriers migrate onto Proa.
• Short-cable economics differ from long-haul. Proa's 2,891 km makes it a focused regional system with different cost, capacity, and reliability tradeoffs than trans-Pacific trunks.

Proa is one of several short, modern, region-focused cables entering service in the mid-2020s. They do not sprint across oceans or connect continents, but they densify specific high-demand corridors. Our 2026 measurements catch Proa in its first year; over the next few years the 4.6× reverse-direction asymmetry will likely compress as traffic engineering adjusts to its availability.

Try it yourself

Live data on the Proa cable page. For context on trans-Pacific cables that Proa complements, see JUPITER, APRICOT, BIFROST, and PC-1. For other Pacific island-focused cables see Bulikula and Tonga Cable.