Tannat

Tannat is a 2,000-kilometre submarine cable that connects Argentina to Brazil, with an intermediate landing in Uruguay. Commissioned in 2017 and active since 2018, it was Google's first Latin American submarine cable investment — built in partnership with Antel, Uruguay's state-owned telecommunications operator. Tannat is the working-horse cable that linked Google's Brazilian cloud region to its Argentine network for half a decade, before the longer and entirely Google-owned Firmina superseded it as the backbone of choice.

Two-partner cable, three landings

Tannat was built in partnership with Antel, Uruguay's 100%-state-owned incumbent telecommunications operator, and it remains one of the relatively small number of submarine cables jointly owned by a hyperscaler and a state-owned national operator. The partnership structure is a feature, not an accident. Google wanted direct Argentina-Brazil connectivity but did not want to negotiate individually with the national telcos of both larger countries, and Uruguay — geographically on the route, politically neutral between Argentina and Brazil, and eager for digital infrastructure investment — offered a perfect intermediate partner. In exchange for allowing Tannat to land at Maldonado, Uruguay received capacity rights on the cable and positioned itself as a regional digital hub.

The three-landing layout works well for both parties. Google gets a direct Buenos Aires-São Paulo route via Las Toninas and Santos. Antel gets a cable that serves Uruguay's international traffic without depending on transit agreements with Argentine or Brazilian carriers. The cable body branches at Maldonado to give Uruguay a distinct landing rather than only a relay through the main fibre.

Our measurements

We monitor Tannat between Las Toninas (Argentina) and Santos (Brazil) — the cable's primary 2,000-kilometre south-north trunk. Over 30 days we have 67 clean samples across both directions:

The Las Toninas → Santos measurement is exceptionally clean: 48 samples in an extremely tight range of 24.9 to 25.7 ms. The physics floor for the 2,000-km great-circle path is roughly 20 ms, so the observed minimum of 25 ms represents an approximately 25% overhead above the theoretical minimum — well within the normal range for a real cable route with repeaters, terrestrial tails, and modest routing detours at each landing station. The minimum and maximum are within 0.8 ms of each other, which is the sort of consistency that only a well-utilised, non-congested cable produces.

The Santos → Las Toninas direction runs higher — 43.6 ms minimum, 46.7 ms average. The asymmetry of about 20 ms between directions reflects the fact that the measured path is not identical: outbound traffic from Santos typically exits Brazilian networks through São Paulo peering points that take a slightly different path through the cable's branching unit than inbound traffic arriving at Santos does. At 18.7 ms of symmetry gap, this is a mild asymmetry by submarine-cable standards, and both directions are stable across the 30-day window.

What Tannat means now that Firmina is live

Before 2026, Tannat was Google's only direct Argentina-Brazil cable and carried a substantial share of Google's South American inter-region traffic. With Firmina now operational (14,517 km from the US East Coast to Argentina, entirely Google-owned, 240 Tbps design capacity), the strategic picture has shifted. Firmina can carry a Buenos Aires to Miami or Buenos Aires to Ashburn (Virginia) packet directly on a Google-owned path, without touching Tannat at all. For traffic that enters the region at Santos (Brazilian ingress), Tannat still matters: it connects Brazilian cloud workloads to Argentine users without routing through Miami or Europe.

Tannat thus becomes a Brazilian-ingress cable rather than a transit trunk. Its role is smaller than it was in 2019 — Firmina handles the larger and longer routes — but it remains essential for Google's São Paulo-to-Buenos-Aires traffic and for Antel's Uruguayan international needs. The cable is not obsolete; it is specialised.

Partnership as a business model

Google's approach to Latin American submarine cables has changed between Tannat (2018, two-party with a national carrier) and Firmina (2026, sole Google ownership). The shift reflects how the hyperscaler business model has matured. In 2017, Google was still establishing its submarine-cable operations and benefited from partnering with national carriers for landing rights and local expertise. By 2024, Google had acquired enough landing rights, regulatory relationships, and internal submarine-cable engineering capability to build wholly-owned cables without national-carrier partners. The Firmina design even eliminated the partner-intensive two-end power-feed architecture, making sole ownership more operationally tractable.

Tannat remains the model of the transitional era — hyperscaler capital combined with national-carrier local knowledge and landing rights. Few cables are still being built with that specific structure; newer hyperscaler cables either own their landings outright (Firmina, 2Africa-Meta) or work through tightly-structured consortium vehicles where the hyperscaler coordinates but does not partner with a single state carrier. Tannat sits quietly in the fleet as a reminder of a different era.

Try it yourself

Live latency data on the Tannat cable page. For comparison with newer Google Latin American cables see Firmina (2026, 14,517 km, Google-only) and Monet (2017, Brazil-USA with multiple partners). Our measurements refresh every two hours.

Capacity at 90 Tbps

Tannat's six fibre pairs carry a design capacity of 90 Tbps — 15 Tbps per pair when lit with current coherent transponders. This is modest compared with modern cables like Firmina (16 pairs, 240 Tbps) or Medusa (24 pairs, 480 Tbps), but it remains more than Argentina-Brazil demand actually requires. Google uses two to three pairs for its own traffic; Antel uses one pair for Uruguay's international connectivity; the remaining pairs are available as commercial spare capacity, sold or leased to regional carriers that want a direct southern-cone path. In practice Tannat is rarely more than half utilised, which is why our measurements show such clean, consistent latencies. A cable with plenty of headroom does not develop queueing-induced variance the way a fully-loaded cable does; it just delivers packets at the physical minimum, every time.

This design decision — over-provision capacity on a strategic cable with a long operational life — reflects Google's appetite for reliability over cost optimisation. It is the same reasoning behind Firmina's single-power-feed architecture: the marginal cost of extra capacity and redundancy is low compared with the cost of having to repair a mid-ocean fault on a cable running at capacity.