Build1 publisher3 min readPublished
Meta reaches a petabit at 7,000 km by doubling the cores inside a 125-micron fiber
Petal keeps Anjana's 24 fiber pairs and puts two cores in every fiber, so throughput per pair stays near 20.8 Tbps and each of roughly a hundred repeaters has to amplify 96 cores in one body.
The Engineer · Build desk
What happened
- Meta's Petal cable will run about 7,000 km between France and the United States at 1 Pbps, which the company says is double what the most advanced subsea cables carry at that distance.
- Each repeater is a single body holding 96 amps, with a Fan-In/Fan-Out interface splitting every 2-core fiber into two single-core fibers for amplification and recombining them afterwards.
- Meta is building Petal with NEC and Sumitomo Electric Industries, with Orange supporting the landing in France, for service in 2029.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- capability Double the transatlantic capacity now fits behind one cable and one set of landings, so the next petabit of transatlantic supply arrives without a second wet plant to build, permit and repair.
- constraint Capacity growth is coming from core count, so a 2030s capacity model should be sized against repeater bodies and amplifier counts.
- cost Ninety-six amps in one body across about a hundred repeaters is roughly 9,600 amplifiers to power and cool, and Meta's claim is that power rises less than proportionally, not that it stays flat.
- decision Anyone specifying a post-2029 cable now has a third option to price against going to 48 fiber pairs or opening the L band, and it comes with a FIFO repeater in the wet plant.
Meta specifies Petal at 1 Pbps, or 1,000 Tbps [5], across what it describes as a 24 fiber-pair system built on 2-core fiber and equivalent to 48 fiber pairs [12]. That is about 20.8 Tbps per equivalent pair [1]. Anjana, Meta's 24-pair transatlantic system, was the first at 0.5 Pbps [10], which is also about 20.8 Tbps per pair [2]. Throughput per pair stays put [3].
Meta also renders the figure as roughly the capacity for 75% of the world's population to stream music at the same time, with an asterisk [6].
Meta's own history explains why the gain has to be spatial. Coherent transmission and dispersion-uncompensated cable designs bought 10x and more per fiber through the 2010s until the Shannon Limit pushed back [8], and the industry answered by putting more fibers in the cable [9]: Marea's eight pairs, Amitie's 16, Anjana's 24 [10]. Petal holds the pair count and adds cores.
Three candidates were on the table, according to the post: go to 48 fiber pairs, open the L-band for 24-pair C+L transmission as on the PLCN cable, or adopt 2-core fiber [11]. Meta took the third [12].
Outer diameter is the constraint that forces the hard engineering. A 2-core fiber here has to hold 125 micron width and low attenuation at the same time, and the cores must not talk to each other [14]. Meta says the first comes from ultra-pure synthetic silica in the preform, and the second from a high refractive index in each core against a lower index in the surrounding medium, plus counter-propagating the optical signals, which leaves crosstalk "nearly immeasurable" [15]. Counter-propagation is also a direction assignment: the two cores of a fiber have to run in opposite directions.
The repeater is where the doubling turns into hardware. A 7,000 km cable typically needs about a hundred repeaters [16], roughly one every 70 km [5]. Petal's is a single body with 96 amps [17], which is 24 pairs times two fibers times two cores [4]. Amplification is still single-core: a Fan-In/Fan-Out interface splits each 2-core fiber into two single-core fibers inside the repeater, amplifies them, then recombines them into 2-core fiber [17]. Over the full route that is on the order of 9,600 amplifiers [6].
Meta's phrasing is that multi-core fiber doubles capacity per fiber "without a proportional increase in power or physical infrastructure" [3]. Proportional. Ninety-six amplifiers in one body draw more than 48, and Meta did not publish a power budget for the wet plant. The saving it does quantify in kind is the cable count: two 0.5 Pbps systems become one petabit system, which Meta says cuts materials, resources and carbon footprint [13].
For the 1 Pbps number to transfer to another route, the route needs this fiber, this repeater, and roughly the same 7,000 km span [1]. Petal is expected in service in 2029, built with NEC and Sumitomo Electric Industries, with Orange supporting the French landing [2][4].
What to watch
- Whether Meta publishes a wet-plant power figure for the 96-amp single-body repeater, which is the only way to test the power claim.
- Whether NEC and Sumitomo Electric Industries sell 2-core fiber and FIFO repeaters to other cable buyers before Petal enters service in 2029.
- Whether a later system combines 2-core fiber with C+L band transmission, stacking the two options Meta weighed separately.