Europe, Meta and Japan Race to Claim the First Petabit-Per-Second Submarine Cable

Bandwidth needs have exploded. Artificial intelligence training clusters now span continents. Hyperscalers move vast datasets in real time. The thin glass fibers that carry 95 percent of intercontinental traffic suddenly look inadequate.

Three players want to change that before 2030. European governments back a 24-fiber-pair system across the North Sea. Meta pours billions into its globe-spanning Project Waterworth. Japanese engineers test multicore fiber that could push single cables into the petabit range. The winner will set the standard for the next decade of AI infrastructure.

Current subsea cables deliver hundreds of terabits per second. That once seemed enormous. Yet AI models grow larger every quarter. Inference workloads spread across regions. The pressure builds. Operators now speak openly about petabit-class systems. One petabit equals one million gigabits. Enough to carry three years of 4K video in a single second, according to Meta engineers.

The European effort centers on IOEMA-1. This 1,600-kilometer cable will link the Netherlands, Germany, Denmark, Norway and the United Kingdom. It carries EU strategic project status under the Connecting Europe Facility. Target readiness sits in the first quarter of 2029. The design calls for 24 fiber pairs. That configuration points directly at petabit-scale capacity once paired with advanced optical equipment.

APTelecom serves as key advisor. Andrew Parsons, involved with the project, said partnering with the firm “brings additional deep expertise.” Sean Bergin of APTelecom added that “IOEMA-1 is a strategically important project that addresses the growing demand for resilient, high-capacity connectivity across Northern Europe.” Both statements appear in a recent TechRadar report.

But Europe does not move alone. Greece has begun construction on ARTEMIS. Grid Telecom, a subsidiary of the Independent Power Transmission Operator, leads the effort. The cable runs roughly 280 kilometers from Crete to mainland Greece, including terrestrial segments. It features at least 24 fiber pairs. Design capacity starts at 720 terabits per second. Engineers see clear potential to exceed one petabit per second total.

Subsea Cables noted the system “is poised to become the first petabit-class subsea cable in Greece and the Mediterranean, with a potential total capacity exceeding 1 petabit per second.” The project adds new landing stations in Chania and Attica. It integrates with existing routes to create multiple independent paths. Regional connectivity gains matter. So does the signal it sends about Mediterranean ambitions.

Meta’s Massive Bet Reshapes the Competition

Across the Atlantic and beyond, Meta has taken the most ambitious step yet. In February 2025 the company announced Project Waterworth. The cable will stretch more than 50,000 kilometers. It will touch five continents. That makes it the longest subsea cable project on record. Depths reach 7,000 meters in places. Enhanced burial techniques protect against anchors and natural hazards.

The system relies on 24 fiber pairs using the highest-capacity technology available today. Meta describes it as a multi-billion-dollar, multi-year investment aimed at strengthening the world’s digital highways. The explicit goal centers on AI. “Project Waterworth will … drive AI innovation around the world,” the company stated in its engineering blog. Santosh J., Head of Global Infrastructure at Meta, reinforced the point: the project helps ensure “the benefits of AI and other emerging technologies are available to everyone, regardless of where they live or work.”

Meta already understands the economic stakes. Earlier cables such as Marea have added roughly $18 billion each year to Europe’s GDP since 2019, according to analysis cited by the company. New systems planned for the continent could contribute $65 billion annually by 2027. Those figures come from Meta’s own engineering announcement.

Yet Meta does not chase petabit capacity in isolation. Its submarine cable systems engineer Philippe Perrier has spoken publicly about architectures for multi-petabit systems. In podcast comments referenced across industry discussions, Meta engineers target a doubling of current transatlantic performance. Half a petabit today becomes one petabit tomorrow. The push comes from internal demands. Llama models and multimodal AI require massive data movement. Private ownership gives Meta control over routes, latency and upgrades.

Japan brings different strengths. The country operates more than 20 international cable landing stations. NEC ranks among the world’s top suppliers of submarine systems. Its work with Google on the Topaz cable delivered 240 terabits per second across 16 fiber pairs. That project connected Japan and Canada. NEC also supplies equipment for many of Meta’s builds.

Japanese researchers now test 12-core multicore fiber over 7,280 kilometers. Early results show hundreds of terabits with coupled-core techniques and MIMO processing. The CSIS report on Japan’s cable strategy highlights how government and industry coordinate. A $440 million fund launched under former Prime Minister Fumio Kishida supports new stations and rural data centers. METI and MIC treat secure connectivity as a national priority. Partnerships with Europe include a 2022 digital partnership council that reaffirmed commitments to resilient infrastructure.

The technological path to petabit scale runs through spatial division multiplexing, C+L band amplification and, increasingly, multicore fiber. Traditional single-core designs hit physical limits. Adding fiber pairs raises costs and repeater power demands. Multicore approaches pack multiple light paths inside one glass strand. Challenges remain. Signal crosstalk, amplifier complexity and marine installation all demand solutions. Still, the trials suggest feasibility within the decade.

Market forecasts reflect the frenzy. The submarine cable sector could nearly double in value by the mid-2030s. Hyperscalers already account for more than two-thirds of internet traffic. They prefer ownership over traditional consortia. That shift accelerates deployment but concentrates risk. Geopolitical tensions add another layer. Cable cuts in the Red Sea and elsewhere have exposed vulnerabilities. Nations now speak of friendshoring critical routes.

Europe’s IOEMA-1 carries explicit strategic classification. It reduces reliance on single chokepoints. Greece’s ARTEMIS does the same for the eastern Mediterranean. Meta’s Waterworth opens new oceanic corridors. Japan’s focus on domestic manufacturing and repair vessels aims to cut dependence on foreign ships. Each approach carries trade-offs. Speed versus scale. Public coordination versus private capital. Open systems versus controlled routes.

Timelines matter. Subsea projects take five to seven years from concept to service. That lag explains the current urgency. AI demand does not wait. Clusters in the U.S., Europe and Asia need low-latency links today. By the time these petabit cables light up, the models they serve will have grown again. The first mover gains pricing power, route preference and data on real-world performance.

Technical hurdles persist. Powering repeaters across thousands of kilometers at higher capacities requires new designs. Cable laying ships remain scarce. Only a handful of factories produce the specialized fiber. Repair vessels matter as much as new builds. Japan works to expand its fleet. European governments explore similar options.

But the race continues. IOEMA-1 targets 2029. ARTEMIS construction has started. Waterworth advances in stages across multiple oceans. Japanese multicore trials gather data that could feed future systems. No single winner will emerge. The industry moves toward a mix of ultra-high-capacity trunks and diverse regional links. Petabit systems will sit at the core.

One fact remains clear. The glass at the bottom of the sea now determines how fast AI can think across borders. The players who lay the right fibers first will shape that capability for years ahead.