The Ocean Floor Is Already Wired for AI

Ciena and Meta hit 800 Gb/s across 16,608 km on the Bifrost cable, cutting power use in half and upgrading AI bandwidth capacity

2 Apr 2026

On March 18, Ciena and Meta set a new global benchmark for long-distance data transmission, pushing 800 Gb/s across 16,608 kilometers of transpacific submarine cable in a single unbroken run. The test used Meta's Bifrost system, which spans Singapore to the US West Coast, and Ciena's WaveLogic 6 Extreme coherent optics.

The trial also delivered 18 Tb/s of total fiber pair capacity while cutting power consumption per bit in half compared to the previous modem generation. That efficiency gain matters enormously at subsea landing stations, where rack space and electrical capacity are perpetually scarce. The entire terminal equipment fit within just 10 rack units, a compact footprint as operators face mounting pressure to scale without gut-renovating their infrastructure.

The result lands at a tense moment for transpacific connectivity. AI training runs, cloud synchronization, and real-time inference are pushing existing routes toward capacity ceilings, while new ocean cable projects typically demand years of planning and hundreds of millions in capital. Unlocking 800G performance on a cable that already exists rewrites the math on capacity expansion.

For Meta, the win serves double duty. Reducing power draw across its privately owned subsea network helps the company manage the energy footprint of its fast-growing AI operations while advancing its net-zero targets. For the broader industry, the trial confirms that coherent optics upgrades can deliver next-generation throughput on virtually any transpacific route currently in service.

The global submarine cable market is projected to climb from roughly $16.4 billion in 2026 to $26.4 billion by 2031, driven by hyperscaler investment and AI demand. Terminal upgrades like this one are emerging as among the fastest and most cost-effective ways to meet that pressure. The infrastructure the AI era needs, it turns out, may already be resting on the ocean floor.