RETN CEO warns fake subsea network resilience threatens global AI traffic

The reliance on concentrated subsea corridors creates a single point of failure for the global streaming and AI ecosystem. As AI clusters migrate toward regions with cheaper electricity rather than proximity to users, the demand for long-haul, high-capacity backbone connectivity increases. If providers like Exa Infrastructure and Tata Communications continue to prioritize cheaper, established routes over geographic diversity, a single maritime accident could disrupt latency-sensitive inference traffic globally. This shift forces a transition from historical data processing to real-time human activity monitoring, making physical route independence a strategic necessity. Watch for whether hyperscalers begin funding terrestrial alternatives to bypass traditional maritime chokepoints.

RETN's warning about concentrated subsea corridors arrives as multiple operators race to add capacity on the same routes. In March 2026, Ericsson's networks chief Per Narvinger noted at MWC that AI workloads will shift traffic patterns through the mid-cycle 5G evolution, with uplink demand growing as AI agents move from text-based queries to real-time video and sensor data. That same uplink surge is what drives long-haul backbone requirements, making the physical diversity of subsea routes a prerequisite for the AI infrastructure buildout that operators and content providers depend on.

The business case for route diversification is sharpening as hyperscalers and carriers commit billions to new cable systems. Ericsson's own Mobility Report from June 2025 quantified the shift: gen AI traffic already carries 26% uplink versus the traditional 10% ratio in mobile networks, and the report projects that immersive AI experiences using augmented reality will flood uplinks with video and sensor data. For backbone operators like RETN, Exa Infrastructure, and Tata Communications, this means the cost of a single cable cut on a shared corridor now carries exponentially higher financial exposure than it did when traffic was predominantly downlink video streaming. The International Cable Protection Committee has long flagged anchor strikes and fishing as the leading causes of subsea faults, but the economic stakes per incident have multiplied as AI inference traffic grows.

On the technology side, 400G coherent optics is becoming the baseline for new subsea and terrestrial backbone deployments, enabling operators to squeeze more capacity from existing fiber pairs without laying new cable. Ericsson's agentic AI research, published in July 2025, describes a multi-agent architecture that processes over 60,000 KPIs to identify 20 distinct classes of network issues and reduces analysis time by 80%, a framework directly applicable to monitoring subsea link health and triggering rerouting decisions before a fault cascades. Meanwhile, enterprise-focused vendors are also embedding agentic AI into network management; Cradlepoint announced in 2025 that it is integrating agentic AI into NetCloud, making it the first enterprise 5G vendor to do so, signaling that is spreading from carrier core to edge and enterprise layers. For RETN and its peers, the combination of higher-capacity optics and AI-driven fault detection offers a partial mitigation, but it does not eliminate the geographic concentration risk that O'Sullivan identifies as the industry's structural weakness.