Trans-Pacific Express (TPE) Cable System

Based on 38 RIPE Atlas measurements from GeoCables monitoring infrastructure, March–April 2026.

The Trans-Pacific Express, generally referenced in carrier documentation as TPE, is a 17,968-kilometre transpacific submarine cable system connecting Nedonna Beach on the Oregon coast of the United States to four East Asian landings: Qingdao and Chongming (near Shanghai) on the Chinese mainland, Geoje in South Korea, Maruyama on Japan's Pacific coast, and Tanshui in Taiwan. The cable came into service in 2008 and is owned by a seven-member consortium that includes AT&T, China Telecom, China Unicom, Chunghwa Telecom, KT (Korea Telecom), NTT, and Verizon. TPE was, when it was built, the first transpacific cable to land directly on the Chinese mainland, and the consortium structure reflects that historic role: every major East Asian incumbent carrier participates, alongside two American Tier-1 operators that anchor the US-side capacity.

The cable's geographic spread is unusual for a single submarine system. TPE physically connects China, Taiwan, South Korea, Japan, and the United States on the same wet plant — five jurisdictions whose surface-level political relationships have varied between cooperation and acute friction over TPE's eighteen-year operating life. The ownership consortium reflects the same unusual breadth: Chinese state-owned carriers (China Telecom, China Unicom) co-own a cable with the Taiwanese national carrier (Chunghwa Telecom), with both Japanese (NTT) and South Korean (KT) participation, plus the American transit operators. This is a piece of physical infrastructure that quietly continues operating across political contexts that, on the diplomatic surface, would not predict it.

Below the floor: 0.827× of theoretical

The minimum round-trip we observe between Nedonna Beach in Oregon and Tanshui in Taiwan is 145.50 ms. The physics floor for the cable's nominal full 17,968-km length is 175.85 ms; the minimum we observe sits at 0.827× of that floor. Below the floor by about 17%.

This is the same below-floor measurement that SJC exhibits at 0.826× on the Singapore-Japan corridor, that EXA North and South exhibits at 0.716× across the Atlantic, and that ARCOS-1 shows at 0.613× in the Caribbean — the route the packet actually takes is shorter than the cable's nominal advertised length. TPE's 17,968 km wet plant covers a six-landing branched topology, and the great-circle distance between Oregon and Taiwan specifically is roughly 80% of that nominal full length. The 145-ms minimum corresponds to a transit of about 14,800 km, which is approximately the actual fibre length used by the Oregon-Taiwan corridor within TPE's branched system.

For the Oregon-Taiwan path specifically, this 0.827× describes a clean transit through TPE's Pacific main span plus the Taiwanese branching unit at Tanshui. The packet enters the cable at Nedonna Beach, traverses the Pacific via the wet plant, hits the branching unit somewhere in the western Pacific that splits the cable toward Taiwan and toward East Asian mainland landings, takes the Taiwan branch, and lands at Tanshui. The 17% below-floor multiplier is the geometric truth that the cable's full advertised length includes segments to China, Korea, and Japan that the Oregon-Taiwan packet never touches.

Two directions, asymmetric in the modern way

The forward and reverse directions of the Oregon-Taiwan corridor produce strikingly different measurements. Across 38 measurements from Tanshui to Nedonna Beach, the round-trip averages 31.86 ms, with a minimum of 30.48 ms, a maximum of 64.88 ms, and a standard deviation of just 6.24 ms. Taiwan-originated traffic to Oregon clearly travels through a clean, short-latency path. Across 17 measurements in the reverse direction — Oregon to Taiwan — the average is 160.99 ms, with a minimum of 145.50 ms, a maximum of 232.14 ms, and a standard deviation of 31.09 ms. Oregon-to-Taiwan latency is more than five times higher than Taiwan-to-Oregon latency.

The 31.86-ms Taiwan-to-Oregon number is too low to be consistent with a transpacific submarine traversal — at the speed of light in fibre, an 11,000-kilometre point-to-point Pacific path would measure around 105 ms minimum, and even a 5,000-kilometre Pacific shortcut would be 50 ms or more. The implication is that Taiwan-side measurements toward our specific Nedonna Beach test endpoint are not actually traversing TPE's transpacific wet plant at all — they are most likely terminating at a much closer endpoint that resolves to the same Oregon address space, possibly a CDN edge node or anycast presence inside Taiwan or Asia. JUNO and other Pacific cables show similar asymmetric patterns when test endpoints have anycast deployments.

The Oregon-to-Taiwan direction at 145-160 ms is the meaningful TPE measurement. It shows the cable doing its actual job: carrying US-origin packets across the Pacific to a Taiwanese destination via the TPE wet plant. That number is consistent with the cable's geometry and is, by current Pacific-corridor standards, a normal latency budget for an eighteen-year-old cable still carrying significant commercial traffic.

Trans-Pacific Express in the 2026 Pacific corridor

TPE entered service in 2008, the same year as Matrix Cable System in Southeast Asia and a few years after ARCOS-1 and EXA North and South from 2001. The cable is now eighteen years into its service life, and like other vintage transpacific systems it has been progressively upgraded through terminal-optic refreshes rather than wet-plant replacement. AJC from the same vintage cohort is operating on a similar trajectory; both cables are senior elements of their respective Pacific routes and continue to carry meaningful regional traffic alongside newer high-capacity systems.

The TPE consortium is structurally interesting in 2026 because of the political evolution that has occurred on this corridor since the cable was built. Several more recent transpacific cables have explicitly avoided Chinese mainland landings under regulatory pressure from the US Federal Communications Commission, which has progressively tightened approval requirements for cables touching Chinese-controlled jurisdictions. JUNO, the 2025-vintage Google-NTT transpacific cable, lands only in Japan and the United States; Honomoana, the Google 2026 cable, takes the long Pacific route via French Polynesia rather than going through any East Asian mainland landing. TPE, which was built before the regulatory environment hardened, retains its Chinese landings and continues to operate them. The cable is therefore one of the few remaining transpacific systems that physically connects Chinese mainland fibre directly to the US Pacific coast through a single owned-by-consortium wet plant.

Maruyama, Geoje, and the Pacific cable cluster

TPE's Maruyama landing places the cable into the same Chiba-coast cluster as AJC, FASTER, and several other transpacific cables. The terrestrial backhaul from Maruyama into Tokyo data centres is the same diversified network that all cables in this cluster share. The Geoje landing on the South Korean island of Geoje connects TPE to the Korean carrier ecosystem in Busan and onward to the Seoul metropolitan area; KT's participation in the consortium means South Korean traffic on TPE has guaranteed access to Korean terrestrial backhaul. Tanshui on Taiwan's northwest coast is the standard Taiwanese cable landing area, hosting multiple international cables alongside TPE.

The two Chinese landings — Qingdao on the Shandong Peninsula and Chongming in the Yangtze estuary near Shanghai — give TPE direct access to two of China's largest commercial regions. Qingdao's coastal industrial economy and Shanghai's financial and commercial dominance make these landings strategically valuable for traffic between the Chinese mainland and the US Pacific coast, and the dual landing provides redundancy against single-point cable cuts in the East China Sea.

What we will keep watching

TPE is in its eighteenth year of service and continues to operate as one of the few transpacific cables with both Chinese mainland and Taiwanese landings on the same wet plant. The Oregon-Taiwan measurement we currently have at 145 ms is consistent with the cable's design parameters and with what an aging but well-maintained transpacific consortium cable should produce. The Taiwan-side asymmetry in our specific measurement endpoint reflects modern CDN and anycast deployment patterns rather than anything specific to TPE's cable geometry; we will look to add additional measurement coverage to other TPE landings as our probe network expands its East Asian footprint.

What we measure on TPE — 145 ms minimum across 17,968 km of transpacific wet plant, with a 0.827× chord behaviour for the Oregon-Taiwan corridor — is what an eighteen-year-old multi-landing trans-Pacific consortium cable looks like in 2026. The cable continues to serve the same five East Asian and US-coast economies it was built for, and its continued multi-jurisdictional operation across politically contested space is one of the structural facts about the modern Pacific submarine corridor that newer cables cannot replicate.