Metas Petal Cable Hits 1 Pbps Multi-Core Fiber Goes Mainstream in Subsea Networks

The Subsea Bottleneck Nobody Talks About

Approximately 99% of intercontinental data traffic — every message, phone call, and video stream between continents — travels through glass strands lying on the ocean floor. Since the 1980s, erbium-doped fiber amplifiers (EDFA) kicked off decades of capacity growth, and coherent optics in the 2010s delivered another 10x. But the Shannon Limit eventually caught up, and the industry pivoted to spatial division multiplexing (SDM): cramming more fiber pairs into a single cable.

Meta's progression tells the story: Marea (8 fiber pairs) → Amitié (16 pairs) → Anjana (24 pairs, 0.5 Pbps). Now comes Petal — the first transoceanic cable to hit 1 Pbps using 2-core fiber technology. For a broader look at how infrastructure trends shape the developer ecosystem, check out our analysis of sovereign AI datasets and culturally-grounded models.

Why 2-Core Fiber Matters

Instead of building two separate 0.5 Pbps cables, Petal deploys a 24 fiber-pair system with 2-core fiber, effectively yielding 48 fiber pairs. This halves material usage, carbon footprint, and physical infrastructure — a critical consideration when each repeater must survive 25 years on the ocean floor.

Inside the 2-Core Fiber Ecosystem

Fiber Manufacturing: Low Loss + Low Crosstalk

Two engineering challenges dominate 2-core fiber design:

1. Low attenuation while keeping the outer diameter at 125 μm
2. Minimizing crosstalk between the two cores to preserve optical performance

Sumitomo Electric solved the first with ultra-pure synthetic silica in the preform. The second was achieved by engineering high refractive indexes in the cores against lower indexes in the surrounding medium, combined with counter-propagating signals.

# Simplified crosstalk estimation for counter-propagating 2-core fiber
# Simplified crosstalk estimation for counter-propagating 2-core fiber
import numpy as np

def crosstalk_db(coupling_coeff, length_km, propagation_constant_diff):
    """
    coupling_coeff: power coupling coefficient (1/m)
    length_km: fiber length in km
    propagation_constant_diff: delta-beta between cores (rad/m)
    """
    L = length_km * 1000  # convert to meters
    # Counter-propagating signals average out phase matching
    return 10 * np.log10(coupling_coeff * L / (1 + propagation_constant_diff**2))

# Example: ultra-low crosstalk target for petabit transmission
xt = crosstalk_db(coupling_coeff=1e-8, length_km=7000, propagation_constant_diff=500)
print(f"Estimated crosstalk: {xt:.2f} dB")  # Target: < -30 dB

Repeater Design: 96 Cores in a Single Body

A 7,000 km cable needs ~100 repeaters. Petal's repeater uses a Fan-In/Fan-Out (FIFO) interface to split each 2-core fiber into two single-core fibers for amplification, then recombine. This preserves the efficiency of proven single-core EDFA amplification while doubling total capacity.

The Shannon Limit Workaround

Rather than pushing more bits per Hz (blocked by Shannon), Petal exploits spatial parallelism. This is the same conceptual shift that drove multi-chip GPU architectures — when you can't go faster per lane, add lanes.

Limitations and Watch Points

• Multi-core fiber is not plug-and-play. The FIFO interface introduces insertion loss and alignment complexity that single-core systems avoid entirely.
• Ecosystem lock-in risk. NEC and Sumitomo are the primary suppliers for this generation. Broader vendor diversification will take years.
• Repair difficulty. Splicing 2-core fiber mid-ocean is significantly harder than single-core. Field repair vessels will need retooling.
• Timeline pressure. Petal enters service in 2029 — a long horizon in which alternative technologies (e.g., hollow-core fiber, higher-order SDM) may shift the economics.
• Security surface. Higher aggregate capacity per cable means a single physical fault affects more traffic. For context on how infrastructure-level failures cascade, see our React Server Components CVE-2025-55182 guide.

What to Learn Next

1. Spatial Division Multiplexing (SDM) fundamentals — the next decade of optical networking
2. Multi-core fiber splicing techniques and emerging ITU-T standards
3. Submarine cable economics — why hyperscalers are now the primary builders
4. Counter-propagating transmission for crosstalk suppression in densely packed fibers

The Bottom Line

Petal isn't just a capacity record — it's a proof of concept that multi-core fiber works at scale. By moving 2-core fiber from laboratory curiosity to deployed infrastructure, Meta is signaling to the entire industry that spatial parallelism is the viable path past the Shannon Limit.

For developers and infrastructure engineers, the practical takeaway is this: the physical layer underpinning your cloud services is being fundamentally rearchitected. Latency, throughput, and even cost models for transatlantic workloads may shift by 2030. The engineers who understand SDM today will be the ones designing the networks of tomorrow.

Source: Meta Engineering — Petal: Petabit Transoceanic Subsea Cable