TL;DR · 30-second read
The Short Version
Meta, the owner of Facebook and Instagram, plans a new internet cable along the ocean floor from the United States to France. It should open in 2029.
It is billed as the first ocean-crossing cable able to carry about as much data as a million fast home connections running flat out.
The trick is a glass strand with two light channels inside, which doubles the room in the cable. But the boosters that keep signals strong along the way still handle each channel separately. As artificial intelligence raises data demands, those boosters matter.
Data Center Dynamics reported on September 23, 2026 that Meta plans Petal, a 7,000km (4,350-mile) subsea cable connecting the US to France and described as the first transoceanic subsea cable to reach petabit capacity. It is being built with NEC and Sumitomo Electric Industries and is expected to enter service in 2029. Telco Orange will manage the landing in France, the point where the cable comes ashore.
Executive Summary
Petal uses two-core fiber, in which a single glass strand carries two separate light-guiding channels. Its 24-fiber-pair system therefore offers the equivalent of 48 fiber pairs. A fiber pair is the basic unit of a subsea cable: one fiber for each direction of traffic.
The more telling detail sits inside the repeaters, the amplifiers placed along a cable to keep its light signals strong. Petal does not amplify two-core fiber directly. Within each repeater, a Fan-In/Fan-Out (FIFO) interface separates each two-core fiber into two single-core fibers, which are amplified and then recombined. Density improves along the cable, but amplification still happens path by path.
This matters because intercontinental bandwidth sits underneath AI and cloud services that run across regions. Petal shows one way to raise capacity on a transoceanic route, and it shows which part of the system does not get simpler when that capacity doubles.
Why Doubling the Fiber Does Not Shrink the Repeater Job
The arithmetic explains the design. Twenty-four fiber pairs means 48 physical fibers. With two cores in each, that is 96 light paths, or the 48 pair-equivalents in the published specification. Because each two-core fiber becomes two single-core fibers inside the repeater, the cable’s 96 paths are handled as 96 single-core paths at every amplification point along the 7,000km route.
So the density gain belongs to the cable span, not to the amplifier site. The cable carries twice the paths per fiber, while each repeater takes on the added FIFO step and still has to amplify every core. That has practical consequences. Subsea repeaters draw their electricity from shore through the cable itself, sit on the ocean floor where repairs are slow and costly, and must fit their components into a sealed housing. Every additional interface is another part that has to perform for the system’s working life, and optical connections of this kind generally add some signal loss.
This is the point for people who specify, supply and finance these systems: a route’s capacity headline does not settle where the engineering burden falls. Petal does not prove that repeaters are the limit on all future cables, since it is one design from one project. It does show that on the first system announced at this scale, moving to multicore fiber did not remove single-core amplification. Repeater suppliers, power-feed designers and cable buyers comparing two-core proposals should look at that layer as closely as at the fiber count.
A Petabit Milestone With a 2029 Delivery Date
One petabit per second equals 1,000 terabits, or a million gigabits. If Petal’s capacity is exactly one petabit, simple division across 48 pair-equivalents gives roughly 21 terabits per second for each one. That is our arithmetic rather than a company figure, but it suggests how the headline is reached: by multiplying paths as well as pushing data through each one.
NEC frames the project in strong terms. “Achieving petabit-per-second capacity across a transoceanic submarine system represents a major technological milestone in the history of global telecommunications,” said Eduardo Mateo, chief strategy officer of NEC’s Submarine Network Division, who also described it as an “achievement” reflecting the company’s sustained R&D investment.
The milestone is best read as a design target for now. Petal is planned, with service expected in 2029, so its status as first depends on it being delivered on schedule and on no rival system getting there sooner. Its petabit figure also needs definition. Design capacity and capacity actually switched on at launch are often different numbers on subsea systems, and real-world performance on a live 7,000km route will be the test of the claim.
Long-Haul Bandwidth as a Planned Asset
Meta has invested in more than 20 subsea cable projects, including Project Waterworth, the world’s longest subsea cable. Petal extends a pattern of a large platform company directly commissioning intercontinental capacity rather than relying only on what carriers offer. Google landing its Nuvem cable in Portugal is another example of a cloud giant active on Atlantic-facing routes, though two companies are a small sample rather than proof of a market-wide shift.
The AI link is an inference. Petal’s published specifications do not allocate capacity to AI or to any other workload, and a company of Meta’s size moves large volumes of ordinary traffic as well. The case for treating it as AI-era infrastructure rests on timing and lead time: a cable announced in 2026 for 2029 service commits to demand roughly three years out, during a period when compute buildout is driving infrastructure planning. Whatever the eventual traffic mix, that lead time means capacity decisions have to precede the demand they serve.
The likely commercial beneficiaries are the partners with defined roles: NEC and Sumitomo Electric as builders, and Orange as manager of the French landing. The effect on wholesale carriers is less clear. Direct investment by a large customer could reduce what that customer buys elsewhere, but the scale of any such effect depends on how Petal’s capacity is owned and whether any of it reaches the open market.
Background
Subsea cables carry most data traveling between continents. They were historically associated with telecom carriers and consortia, but large internet and cloud companies have become major investors in their own right. Meta’s more than 20 projects, including Project Waterworth, place it among the most active of those companies.
Cable capacity can be raised by pushing more data through each fiber path or by adding more paths. Multicore fiber takes the second approach by placing multiple light channels in one strand. Petal applies that idea through two-core fiber while retaining single-core amplification in its repeaters. NEC, which has a dedicated Submarine Network Division, and Sumitomo Electric are building the system, and Orange will manage its French landing. Source: Meta plans Petal, the first transoceanic subsea cable at petabit capacity, Data Center Dynamics’ report on the planned US-France cable, its partners and its two-core fiber design.Sources