Citi: Nvidia Roadmap to Catapult CPO Market to ¥185 Billion

CPO Component Market to Exceed ¥185 Billion by 2028

Citigroup projects the market for four core Co-Packaged Optics (CPO) components will expand to over ¥185 billion (approximately $25.5 billion) by 2028 as AI infrastructure scales. The forecast breaks down the opportunity into four key segments: Fiber Array Units (FAU)/connectors, External Laser Source Pluggable (ELSFP) modules, fiber interconnect modules, and fiber trays. This surge is predicated on data centers shifting from traditional pluggable optics to more power-efficient, high-bandwidth CPO technology to handle increasingly complex AI workloads.

The most dramatic growth is expected in 2027, which Citi identifies as a key inflection point. The FAU/connector market is forecast to surge over 3,400% that year, with the ELSFP market growing over 2,600% and fiber trays expanding over 3,200%. Fiber interconnect modules are set to become the largest segment, projected to reach ¥89.6 billion by 2028, accounting for nearly half the total market as demand shifts from scale-out network expansion to large-scale, rack-to-rack scale-up interconnects.

Nvidia's Roadmap Drives 140-Fold Jump in CPO Switch Demand

The explosive market growth is directly tethered to Nvidia's CPO adoption roadmap. Citi's model relies on the phased rollout of CPO technology within Nvidia's next-generation architectures, starting with the Vera Rubin Ultra platform and culminating in the fully CPO-integrated Feynman Kyber 1152 systems. Nvidia's influence is expected to coordinate the entire supply chain, accelerating mainstream adoption.

This technology shift will drive an unprecedented expansion in CPO switch demand. Citi forecasts demand will leap from 5,000 units in 2026 to 209,000 in 2027, before reaching 691,000 units in 2028—a nearly 140-fold increase in under three years. This projection is modeled on the deployment of approximately 7 million Nvidia Rubin Ultra GPUs, highlighting the direct link between next-generation AI processors and the underlying optical network architecture.

Supply Chain Braces for Bottlenecks as Demand Intensifies

The transition to CPO will not be seamless, with significant supply chain constraints emerging as a primary risk. Nvidia's planned three-stage CPO migration, starting with a hybrid approach in the second half of 2026, implicitly acknowledges that the supply chain is not yet prepared for a full-scale transition. This gradual rollout is designed to allow component manufacturers time to ramp up capacity for critical parts like laser sources and optical engines.

Key suppliers like Lumentum are already experiencing this demand firsthand, with a reported OCS backlog surpassing $400 million and CPO procurement commitments extending into 2027. However, the projected annual growth rate for components like indium phosphide optics suggests that supply-side constraints will likely persist. While these bottlenecks pose a risk to deployment timelines, they also support strong pricing power for well-positioned suppliers. Investors, however, note the primary risk remains whether global AI investment can sustain its current intensity to justify this infrastructure buildout.