Omdia: Optical Network Market Enters AI Infrastructure Supercycle, Hits 25-Year High

Ian Redpath, Research Director for Optical Networks at Omdia, said at a forum held during CIOE 2026 that the global optical network market has entered a supercycle driven by AI infrastructure buildout, reaching a 25-year high in market size as the growth engine shifts from bandwidth upgrades to AI compute interconnection.

According to Omdia estimates, the annualized global optical network market reached $16.9 billion in the second quarter of 2026, up 13.7% year over year, while the single quarter grew 26.7% compared with the same period in 2024—both setting 25-year records. Redpath described 2025 as the start of recovery and 2026 as a period of forecast growth based on order backlogs and AI support.

Redpath highlighted two types of new players in the AI infrastructure market. The first are NeoClouds, operators of large centralized AI data centers, many with cryptocurrency backgrounds, who have GPU high-power infrastructure experience and early partnerships with Nvidia, and are filling capacity gaps left by hyperscalers. The second is Nvidia itself, which is vertically integrating the ecosystem through minority investments spanning CoreWeave, Crusoe, Anthropic, OpenAI, and optical component makers such as Corning, Coherent, Lumentum, and Marvell.

Redpath noted that Nvidia's distributed training experiments are driving the rise of the Scale Across concept, which interconnects AI xPU resource pools across multiple data center sites to maximize training efficiency while minimizing latency, power, and cost. In a trial between Chicago and Ashburn, about 1,000 km apart with 1,536 GPUs at each site, round-trip latency was measured at 21 milliseconds with scaling efficiency above 96%, and Nvidia has set a target of scaling to 500,000 GPUs.

Multiple technology innovation tracks are underway, including multi-layer/multi-vendor integration, multi-rail line systems, commercial optical circuit switching, liquid cooling for pluggable modules, and the arrival of 2nm coherent DSPs. Redpath also said operators can capture opportunities in the "AI middle mile" and connecting enterprise data lakes to AI training centers, but stressed that Omdia's forecasts assume normal geopolitical conditions and the realization of AI promises.