Meta’s Petal targets a petabit across the Atlantic

Meta has unveiled Petal, a planned 7,000-kilometer subsea cable between France and the United States. The system is designed to carry 1 petabit per second, which Meta says would make it the first ocean-spanning cable to reach that capacity. Service is scheduled to begin in 2029.

One petabit per second equals 1,000 terabits per second, or a theoretical 125 terabytes per second before protocol overhead, operating margins, and reserved capacity. Meta estimates that the aggregate bandwidth could support simultaneous audio streaming for roughly three-quarters of the world’s population.

Subsea cables carry about 99% of intercontinental data traffic. Cloud platforms and AI systems are increasing demand for cross-region transfers of datasets, checkpoints, model weights, replicas, and inference traffic. Petal addresses that demand by placing two optical cores inside each fiber while retaining the standard fiber diameter and established repeater technology.

Capacity moves inside the fiber

Meta’s recent cable systems show the steady increase in fiber count that preceded Petal:

Anjana is designed for 0.5 Pbps across the Atlantic. Petal doubles that target without increasing the number of physical fiber pairs.

Coherent optics and digital signal processing have increased the number of bits carried by each wavelength. Those gains diminish as a channel approaches the Shannon limit, the maximum data rate available for a given bandwidth and noise level. Cable designers can then add more spatial paths through additional fibers or cores.

Meta evaluated three designs for doubling Anjana’s capacity:

• More strands: expand a conventional cable from 24 to 48 fiber pairs.
• More spectrum: carry traffic across both the C and L optical bands on 24 pairs.
• More cores: place two independent optical paths inside each fiber on 24 pairs.

Meta chose the two-core design. Petal’s 48 physical fibers contain 96 optical cores, providing the same number of one-way paths as 48 conventional fiber pairs while using fewer strands.

Two cores fit the standard footprint

Each Petal fiber retains the industry-standard outer diameter of 125 micrometers, roughly the width of a human hair. That compatibility allows suppliers to use established fiber manufacturing and cable assembly equipment.

Sumitomo Electric Industries manufactures the fiber from ultra-pure synthetic silica. Its refractive-index profile confines light within each core, reducing leakage between them. The two cores also carry signals in opposite directions, which helps receivers reject leaked light. Meta reports that the resulting crosstalk is nearly immeasurable.

Keeping both cores isolated over 7,000 kilometers is essential because accumulated crosstalk would otherwise raise the noise floor and reduce the usable data rate. Low optical loss also determines how far signals can travel before amplification.

Repeaters bridge two fiber designs

Optical signals weaken as they travel through glass, so a transatlantic cable typically needs about 100 powered repeaters along its route. These units amplify light directly, avoiding an electrical conversion at every point.

Petal uses NEC single-body repeaters with 96 amplifier channels. Fan-in and fan-out components separate each two-core fiber into individual single-core paths before amplification. A second component recombines the paths after they pass through established single-core erbium-doped fiber amplifiers.

The hybrid design concentrates the newer multi-core technology in the cable trunk while retaining amplifier components with an existing reliability record. It also limits the amount of new hardware that must complete the lengthy qualification process required for equipment expected to operate on the seabed for decades.

Petal is designed to remain within the 18-kilovolt limit of existing subsea power-feeding equipment. The additional capacity therefore avoids a proportional increase in shore-supplied voltage or a new qualification regime for higher-voltage systems. Each landing station will still require project-specific terminal, monitoring, and terrestrial network equipment.

Four partners split the build

• Meta is leading the project and specifying the system architecture and capacity requirements.
• NEC is the turnkey supplier responsible for the cable, repeaters, fan-in and fan-out systems, and marine installation.
• Sumitomo Electric Industries developed the 2C Z-PLUS ULL two-core fiber, whose ultra-low-loss profile supports transoceanic transmission.
• Orange will land the cable on France’s Atlantic coast and connect it to the European terrestrial backbone.

AI demand meets physical limits

Meta has tied its subsea expansion, including Project Waterworth, to the infrastructure required for AI and global cloud services. Transfers between American and European data centers can include training datasets, model checkpoints, production weights, backup replicas, and inference traffic alongside conventional consumer and enterprise data.

Petal’s 1 Pbps figure represents aggregate system capacity distributed across fiber pairs, wavelengths, directions, and services. A single application connection would receive only a fraction of that bandwidth, and usable throughput would account for protocol overhead, protection margins, maintenance reserves, and traffic engineering.

One Petal system is designed to provide the raw capacity of two Anjana-class systems. Operators would still maintain diverse routes and spare capacity so that a cable cut, equipment failure, or maintenance window does not isolate a region.

Where multi-core goes next

Petal’s architecture gives cable builders a practical model for introducing multi-core transmission without replacing every part of the established subsea system. Its design has three broader technical implications:

1. Spatial scaling can supplement wavelength gains. Additional cores create more independent light paths within the same strand while C-band and L-band improvements remain available.
2. The standard fiber diameter can accommodate more capacity. Existing cable machinery and housings remain usable, although splicing, testing, and repair tools must support the two-core structure.
3. Power efficiency will shape further expansion. Petal stays within the current 18-kilovolt class, while future systems will depend on amplifier efficiency and the available shore-fed power budget.

Broader adoption will depend on manufacturing yield, splice loss, fan-out reliability, repair procedures, supplier availability, and the economics observed after deployment. Field data from installation and operation will show whether the architecture can support additional cores and other ocean routes.

Before Petal enters service in 2029, the project must complete cable and repeater qualification, marine surveys, permits, landing construction, route installation, segment splicing, and end-to-end testing. Successful deployment would give the subsea industry its first large-scale operational evidence for two-core transoceanic fiber.