17,184 km · 6 Landing Points · 5 Countries · Ready for Service: 2025
| Length | 17,184 km |
|---|---|
| Status | In Service |
| Ready for Service | 2025 |
| Landing Points | 6 |
| Countries | 5 |
| Location |
|---|
| Agat, Guam |
| Changi North, Singapore |
| Eureka, CA, United States |
| Ngeremlengui, Palau |
| Piti, Guam |
| Tanjung Pakis, Indonesia |
Echo is a 17,184-kilometre trans-Pacific submarine cable jointly owned by Google and Meta, with landings in California, Indonesia, Guam (twice), Palau, and Singapore. It went into service in 2025 and is the first major trans-Pacific cable in over a decade that lands in Palau — a nation of under 20,000 people — as an equal stop alongside Singapore and the United States. Echo is the cable that made a small Pacific island nation a direct participant in the global internet backbone, not just a tail-end customer of someone else's route.
| Specification | Value |
|---|---|
| Length | 17,184 km |
| Ready for service | 2025 |
| Fibre pairs | 12 |
| Landings | Eureka (CA, US), Agat and Piti (Guam), Ngeremlengui (Palau), Tanjung Pakis (Indonesia), Changi North (Singapore) |
| Owners | Google, Meta |
Echo's route is purposeful. Eureka, California, is Google's trans-Pacific landing station — a quiet coastal town hosting capacity that feeds Silicon Valley data centres. Two landings on Guam (Agat and Piti) reflect the US territory's role as the switching hub of the trans-Pacific cable network: most modern cables have redundant Guam terminations because losing one landing to a typhoon or a ship anchor is a recurring operational concern. Tanjung Pakis in Indonesia lands the cable on Java, near Jakarta. Changi North in Singapore gives Echo a presence at the region's largest internet exchange point. And then there is Ngeremlengui, on the island of Babeldaob in Palau — a landing that no purely commercial cable would have chosen.
Palau is a Pacific microstate of some 18,000 people. Its capital, Ngerulmud, has a population of under 300. Before Echo, Palau's international internet connectivity ran through a single older cable (SEA-US, which lands at Babeldaob's Ngiwal station), and internet service on the islands was famously expensive and slow. The Ngeremlengui landing gives Palau a second independent cable — critical resilience, because a single cable fault can sever an island nation's internet entirely — and, more importantly, a direct landing on a hyperscaler-backed backbone.
The commercial logic for a Palau landing is weak. Eighteen thousand customers are not enough to justify the landing station, the civil engineering, and the ongoing maintenance. The logic is strategic and reputational: Google and Meta pay for the Palau landing as part of their Pacific island strategy, which includes diplomacy with Pacific nations and differentiation from Chinese-state-aligned cable operators. For Palau, the benefit is substantial: local broadband prices have room to fall, and the country acquires the ability to host modest international internet services without paying transit to a Guam-based intermediary.
We monitor Echo between Tanjung Pakis (Indonesia) and Eureka (California) — the full trans-Pacific traversal of roughly 15,000 km. Over 30 days we have 82 clean samples across both directions:
| Direction | Samples | Min RTT | Avg RTT | Max RTT | Baseline | Ratio |
|---|---|---|---|---|---|---|
| Tanjung Pakis → Eureka | 53 | 178.2 ms | 183.0 ms | 305.5 ms | 211.1 ms | 1.01 |
| Eureka → Tanjung Pakis | 29 | 125.2 ms | 222.7 ms | 269.9 ms | 252.6 ms | 0.90 |
The Tanjung Pakis → Eureka direction delivers a stable 178 ms minimum with a tight average of 183 ms, essentially at baseline — traffic is taking Echo's direct fibre and the cable is working exactly as designed. Physics floor for a 15,000-km trans-Pacific traversal is roughly 168 ms, so the observed minimum of 178 ms is about 6% above the floor, well within the normal range for real cable architectures with repeaters every 80 km and terrestrial-tail latency on each end.
The Eureka → Tanjung Pakis direction shows more variability. A 125.2-ms minimum is below the physics floor, which points to the fact that our landing-point estimation uses approximate coordinates and the actual fibre route may be shorter than the great-circle distance we assume. The 222.7-ms average with 270-ms maximum suggests that packets in this direction are sometimes routed through alternative cables rather than Echo itself — a common pattern where a hyperscaler's private cable is used on its outbound but traffic returning is handled by whichever cable the destination carrier prefers.
Echo has 12 fibre pairs, fewer than the 16 pairs carried by contemporary hyperscaler cables like Apricot (16 pairs, same route) or by the 24-pair Medusa. This is partly a scheduling artefact — Echo entered the design cycle earlier than Apricot — and partly a reflection of how fast the industry is moving. Twelve pairs fit comfortably inside a single 20-mm cable body, the standard form factor for long-haul submarine cables. The transponders lit on Echo today are 200 or 400 Gbps per wavelength; by the time the cable reaches its nominal 25-year operational life, 800 Gbps and 1.2 Tbps wavelengths will be standard, and the total capacity available on Echo will be several multiples of its initial lit figure.
Google and Meta split Echo as co-owners rather than splitting capacity on a larger consortium cable. Each owns a set of fibre pairs outright. Neither company's public filings disclose exactly how the pairs are divided; the industry convention is that hyperscalers get four to six pairs each for a two-party cable of this size, with remaining capacity sold into the wholesale market. The wholesale capacity on Echo is operated by third-party capacity brokers who sell to non-hyperscaler customers such as regional ISPs, content-delivery networks, and cable-backhaul consortia.
Echo is part of a broader reshuffling of the Pacific cable network. Google alone has Junior, TPU, Apricot (with Meta and others), and Echo in various stages of operational life. Meta has a similar portfolio. Chinese state-aligned operators have their own cables. The net effect is that the trans-Pacific submarine cable market is now crowded, with multiple hyperscaler-private and consortium systems competing for the same data flows. A decade ago, Trans-Pacific capacity was scarce and expensive; today, it is abundant and the bottleneck has moved to the terrestrial side, where fibre into inland US and Asian data centres has to be upgraded to keep up with cable capacity.
For Palau, Echo is quite possibly the most important piece of infrastructure the country will see this decade. For Google and Meta, Echo is one of several redundant paths. The asymmetry of those perspectives is itself the story of hyperscaler-era submarine cables: what is routine for the builders is transformative for the smallest landings.
Live latency data on the Echo cable page. For related cables see Apricot (Google + Meta + partners, 2025) and TPU (Google private, 2026). Our measurements refresh every two hours.
| Status | ✓ Normal |
|---|---|
| RTT | 204.97 ms / base 231.45 ms |
| Last checked | 2026-04-18 20:31 |
Monitored using RIPE Atlas probes. Open monitoring →
| Min | Avg | Max | # | |
|---|---|---|---|---|
| 7 days | 201.8 | 230.1 | 269.9 | 19 |
| 30 days | 125.2 | 220.1 | 269.9 | 21 |
| 60 days | 125.2 | 220.1 | 269.9 | 21 |
Find the actual cable routing distance between any two cities
Open Calculator →