Workshop: Which Fiber Architecture Will Power AI-driven Networks and Cross the Commercial Chasm?

As AI workloads scale globally, the physical layer faces unprecedented capacity, latency, and density demands. Although researchers keep pushing the performance limits of single-mode waveguides, operators prioritize protecting infrastructure investments and service stability. For this reason, they prudently favor maximizing the capability of mature single-mode fiber systems via enhanced packaging density, advanced amplification, and system-level optimization, before committing to the large-scale introduction of radically new fiber types.

This workshop examines when this approach reaches its limits and the inflection points required for new optical architectures, such as hollow-core, multi-core, or few-mode/multi-mode fibers, and ultra-low-loss silica, to cross the commercial chasm and transition from academic demonstrations to operational necessities.

Moving beyond the reductive cost-per-bit debate, the discussion centers on critical drivers of commercialization, including:

• Identifying application-level performance gaps that inherently mandate new physics
• Evaluating manufacturability at scale
• Addressing the “ecosystem tax” of integration and standardization
• Adapting to shifting system-level KPIs

The workshop places a specific focus on access and edge networks. In these environments, access-length spans, constrained fiber density, low-latency fronthaul, edge AI interconnects, time-sensitive networking, industrial PON, and premium last-mile services may expose requirements that cannot be easily addressed by conventional single-mode fiber.

The workshop will also consider practical deployment challenges, including coexistence with existing fiber infrastructure, splicing, connectorization, and field repair.