GlobeNet

Based on 28 RIPE Atlas measurements from GeoCables monitoring infrastructure, March–April 2026.

GlobeNet is, by some distance, the longest submarine cable system we currently monitor: 23,500 kilometres of fibre tracing an elaborate loop across the western Atlantic and Caribbean, with landings in seven cities across five countries. It came into service in 2000 and is still operated today by V.tal — the Brazilian neutral-wholesale operator spun off from Oi in 2021 and, not incidentally, the same owner that built the Malbec cable we've previously written about.

From Tuckerton, New Jersey to Rio de Janeiro, our probes measure a minimum round-trip of 123.03 ms. The physics floor for the full 23,500 km of GlobeNet's laid fibre is 229.99 ms. That puts the observed path at 0.535× of the cable's theoretical floor — meaning the packet we measured is traversing roughly half the fibre that the cable contains.

This is a story about how modern submarine cables handle their own size. When a consortium lays 23,500 km of fibre and invoices that as the system's length, they are describing everything they built. What packets actually use, at any given moment, is always a subset of that — and in GlobeNet's case, a subset you can calculate from the measurement.

The loop that GlobeNet draws

GlobeNet's seven landings don't form a line. They form a meandering loop that touches both the US east coast, the Brazilian coast, the Caribbean rim, and the western Atlantic island of Bermuda.

If you tried to use the whole cable end-to-end — say, from Tuckerton through Boca Raton, across to Bermuda, down to Maiquetía, across to Barranquilla, then all the way south to Fortaleza and Rio — you'd cover something close to GlobeNet's 23,500 km total. No traffic does this. The architecture exists because each of those seven landings was commercially valuable in 2000, and the cable was built to serve all of them.

Our specific measurement — Tuckerton in New Jersey to Rio de Janeiro — lies along one sub-arc of that loop. The great-circle distance between these two cities is approximately 7,800 km. With the typical 10% fibre-over-geodesic overhead, a dedicated straight cable would be around 8,600 km, giving a theoretical floor near 85 ms. But GlobeNet doesn't run straight; its actual fibre path between NJ and Rio, via Florida and possibly via Fortaleza as an intermediate landing, is closer to 12,500 km. That gets us to a 125 ms floor — and our 123 ms measurement is right on it.

Why the cable includes routes no packet uses end-to-end

This isn't wasteful design. It's the commercial logic of multi-landing cable construction. When GlobeNet was planned in the late 1990s, its investors needed to sell capacity to customers in seven different markets: US internet backbone operators, Bermuda's small but strategically located financial-services industry, Venezuelan and Colombian telcos, and Brazil's rapidly growing internet sector. Each of those markets was worth something, and adding landings along the route was much cheaper at construction time than building separate cables to each destination.

The loop also provides resilience. If a fibre cut happens between Rio and Fortaleza, traffic from Rio can potentially re-route via the Caribbean or even via Bermuda to reach US destinations. The full 23,500 km is insurance as much as it is commercial coverage — a fact that's invisible in normal operation but becomes important every time the cable crew has to respond to a fishing-trawl damage incident or an undersea landslide.

V.tal's role and the longevity question

V.tal is interesting as an owner. Before 2021, GlobeNet was operated by Oi, one of Brazil's largest carriers. In 2021, Oi spun off its infrastructure assets — towers, fibre, submarine cable landings — into V.tal, a neutral wholesale operator whose business model is selling capacity on its assets to any downstream retailer rather than competing at the retail level itself. This change is part of a broader pattern in Latin America, where telecoms have been unbundling their infrastructure into separate "network sharing" entities that can serve multiple competitors on equal terms.

V.tal now holds GlobeNet alongside Malbec (Meta co-funded, 2021, Argentina–Brazil) and other assets. The two cables couldn't be more different in vintage or architecture — GlobeNet is a 26-year-old 7-landing loop, Malbec is a 4-year-old hyperscaler-funded short-arc trunk — but they share the same commercial structure: V.tal operates the infrastructure, and customers pay for capacity on commercial terms regardless of who they are.

For GlobeNet specifically, this structure helps explain why a 26-year-old cable is still commercially viable. It doesn't need to earn a retail margin for V.tal's internet-service business — V.tal doesn't have one. It just needs to sell wholesale capacity at a price that covers ongoing maintenance and produces a reasonable return on the infrastructure itself. At 26 years of age, the original construction capex is fully depreciated, and any revenue above maintenance is pure margin.

GlobeNet in the Latin American cable family

The US–Brazil corridor is one of the best-served submarine-cable routes in the world, with at least a dozen systems competing for the traffic. Cables we've previously written about include South American Crossing (2000, Telxius), AMX-1 (2014, América Móvil), Malbec (2021), and TAM-1 (2026). GlobeNet shares its vintage with SAC and AMX-1's predecessors, its ownership model with Malbec, and its physical geometry (multiple landings along a continental coast) with AMX-1 specifically.

What makes GlobeNet distinctive in this family is the Bermuda landing. St Davids is not a major commercial market — the island has a population of about 65,000 — but its position in the western Atlantic makes it a useful waypoint for financial-services traffic that wants to touch jurisdictions outside both the US and Brazil. Bermuda's reinsurance and banking industries account for a disproportionate share of the Atlantic's commercial internet traffic per capita, and GlobeNet's landing there was specifically priced in to the commercial model at construction time.

What we're watching

The 123 ms floor is clean and stable; standard deviation across 28 measurements is 9.16 ms, which suggests the Tuckerton-to-Rio sub-arc has settled routing and consistent BGP preferences. Three things would be interesting to watch over time. First, whether the floor drifts upward — the standard aging signal for a 26-year-old cable. Second, whether V.tal announces capacity-share additions or withdrawals, since neutral-wholesale operators tend to be transparent about infrastructure-level economics in a way that retail carriers aren't. Third, whether a reverse direction ever shows up in our probe rotation; right now we only measure Tuckerton→Rio, and a Rio-origin probe would let us check whether the same sub-arc operates symmetrically or whether traffic in the other direction takes a different path through V.tal's network.