AI Infrastructure

The AI infrastructure build super cycle in data centers is catalyzing an optical network super cycle. Scale-Across, fiber construction, high bandwidth service innovation, and the "Nvidia Effect" are all moving the optical market in a high-growth direction.

AI as the great network catalyst

“Scale-across and AI infrastructure build are catalyzing a new wave of optical network construction.”

Ian Redpath, Chief Analyst, Optical Networks

Carriers have or are rapidly becoming highly attuned to the AI bandwidth opportunity; this momentum will continue in 2026. There are two subsegments within the AI infrastructure build: the ultra-high-capacity data center core subsegment and the access into AI resources subsegment.

Many carriers will play where they have strength, based on their starting point assets. The industry is extremely focused on the AI data center campus buildout roadmaps, which are tightly tied to the availability of power. The industry is undergoing a resurgence of terrestrial and subsea physical fiber buildouts. The AI buildout will continue to occur where the AI-cloud providers have strength, which will lead to “home court” regions seeing the early construction. The buildout will, in time, reach a global footprint.

This research is taken from:

AI-driven growth: Cloud providers lead optical components market

"Cloud service provider data centers dominate data communications (datacom) optical components (OC) revenue. In 2024, total OC spend for cloud service provider data centers was $6.7bn, over 60% of the global datacom market and a year-over-year market revenue increase of 81%, bolstered by the buildout of AI networks.”

Esther Cheh, Research Manager

GenAI is driving cloud service providers toward faster data rates and equipment refresh cycles. The market remains split between North America and China, where most data centers operate. Success will depend on delivering 400G+ transceivers, with linear drive technology showing promise for 800G+ speeds while meeting power targets.

Cloud service providers are evaluating co-packaged optics (CPO) for AI networks and optical circuit switches for spine networks, with rollouts starting in 2026. Coherent technology is expanding into campus/metro networks under 10km, while coherent-lite solutions may emerge in data centers in the future.

This research is taken from:

Optical Components Intelligence Service

AI-driven broadband networks

“The use of AI requires a data-centric architecture where data is analyzed to produce benefits for the operator.”

Stephen Wilson, Senior Principal Analyst

As broadband access networks continue to evolve AI will take on a growing importance across multiple areas. AI can enhance network maintenance and some operators, such as large cable operators in the USA, have been active in developing their own models for some years. The benefits include better proactive maintenance to avoid network downtime as well as improved mean time to repair. As the focus on the broadband customer experience increases, improved reliability can translate not just into lower opex but also into reduced churn and improved subscriber acquisition. AI will also assist operators in capacity planning and this could involve better understanding the need to upgrade FTTP rollouts from the Gigabit Passive Optical Network (GPON) standard to 10G PON and beyond. AI can help by using telemetry data to understand the actual traffic patterns generated by subscriber behavior and predict SLA conformance, for instance, in terms of the probability of meeting a regulator-imposed speed test requirement. This is likely to become more important over time as traffic continues to grow, and more regulators seek to ensure that subscribers receive the speeds they are promised in operator advertising. In order to enjoy the full benefits of AI operators will need to move to a data-centric architecture where high quality data is accessible to multiple parties across the organization.

This research is taken from:

Broadband Access Intelligence Service