Against the backdrop of continuously exploding demand for AI computing power, the optical communications industry is entering a new phase of growth. As the global OFC 2026 conference approaches, multiple institutions are reassessing the long-term demand for optical network infrastructure.
According to analysis, Bank of America Securities points out in a recent report that AI infrastructure is driving optical network demand into a new "super-cycle." The core driver is not just computing power expansion but the evolution of data center architecture from "Scale-Up" to "Scale-Across," a shift that is significantly increasing the need for optical interconnects between data centers.
Bank of America forecasts that the global optical transmission market will maintain growth exceeding 10% in 2026-2027. Within this market, 800G coherent optical modules and ZR/ZR+ pluggable optical modules are expected to be the fastest-growing segments.
Optical networks are emerging as the next bottleneck. Over the past two years, AI infrastructure investment has primarily focused on GPUs and computing servers. However, as computing scale rapidly expands, network bandwidth is gradually becoming a new performance constraint.
The report identifies three primary drivers for the current surge in optical communication demand:
First, increased bandwidth requirements in traditional data centers. Even in non-AI data center scenarios, cloud providers continue to enhance network capacity, with companies like Microsoft significantly increasing procurement for front-end network equipment.
Second, a new wave of hyperscale data center construction. Bank of America estimates that cloud providers will add several gigawatts of data center capacity between 2026 and 2028, with the equipment investment for each gigawatt of capacity being approximately $55 billion.
Third, the demand for inter-data center connectivity driven by the Scale-Across architecture. As AI training scales up, a single data center can no longer house all the necessary computing power. Computing clusters are now being deployed across multiple data centers, leading to a surge in demand for high-bandwidth optical interconnects.
This signifies a shift in AI infrastructure from being solely "compute-driven" to a model of "compute and network synergy."
The transition to Scale-Across represents a key turning point in AI cluster architecture. Traditional AI clusters primarily relied on a Scale-Up model, continuously adding more GPUs within a single data center.
However, as model sizes surpass trillions of parameters, this architecture faces two major limitations: power and cooling bottlenecks within data centers, and increased network topology complexity. Consequently, major cloud providers are shifting towards a Scale-Across architecture—distributing AI computing resources across multiple data centers connected by high-speed optical networks to form a unified computing pool. Under this architecture, AI training tasks can be flexibly scheduled across different data centers, network bandwidth requirements are significantly elevated, and Data Center Interconnect becomes core infrastructure.
Bank of America predicts the Scale-Across AI networking market will grow eleven-fold in 2026 to $808 million, becoming one of the fastest-growing areas in AI network infrastructure, a trend directly fueling the optical module demand explosion.
With the surge in optical interconnect demand, optical module technology is also rapidly advancing. The current mainstream technology paths include copper cabling for very short distances inside server racks, IMDD optical modules for intra-data center networks, and coherent optical modules for long-haul connections between data centers. The most critical demand from AI infrastructure falls into the third category—coherent optical interconnects. While 400G ZR modules have been the mainstream solution in recent years, the market is rapidly transitioning to 800G.
Data from Dell'Oro shows that while 400G ZR held a 93% market share in 2025, its share is expected to drop to 61% in 2026, with the 800G share leaping to 35%. More importantly, the 800G market size is projected to grow nearly tenfold in 2026. Bank of America anticipates an 83% compound annual growth rate for 800G ZR/ZR+ optical module revenue from 2025 to 2030, positioning it as the most significant growth engine for the optical communications industry.
Another major shift in optical networking for the AI era is the widespread adoption of the pluggable optical module architecture. Traditional telecom networks typically procure integrated optical system equipment, but cloud providers have chosen a different path: separating routers/switches from optical modules, inserting modules directly into equipment in ZR/ZR+ form, leading to a more modular network architecture. The advantages of this model include lower cost, reduced power consumption, and greater network flexibility. As a result, pluggable optical modules have become the mainstream solution for hyperscale data center interconnect. 2025 marked a key inflection point for coherent pluggable modules, with the ZR module market growing 36% and the overall optical transmission market growing 18%.
In this optical network upgrade cycle, Bank of America identifies two companies as likely primary beneficiaries.
The first is Ciena, a leader in cloud optical networking with approximately 30% market share in the optical transmission market. Its core strengths include its WaveLogic series DSP technology, leading position with cloud customers, and RLS optical line systems. Bank of America expects Ciena to expand its share in the 800G optical module market, noting that its WaveLogic 6 Nano chip, using a 3nm DSP process, holds a clear advantage in power efficiency. Additionally, Ciena's development of the DCOM data center management system for Meta is projected to generate over $100 million in revenue in 2026. Bank of America rates Ciena a "Buy" with a $355 price target.
The second is Cisco, which holds a first-mover advantage in the 800G cycle. While Cisco's market share in optical transmission systems is only about 6%, it holds a stronger competitive position in the pluggable optical module segment. In the 400G ZR module market, Marvell holds approximately 52% share, Cisco about 35%, and Ciena about 13%. In the emerging 800G market, Cisco's current share exceeds 50%, compared to Ciena's 29%. Bank of America believes that Cisco, leveraging its router and switch ecosystem, maintains an important position in AI data center network upgrades. Bank of America rates Cisco a "Buy" with a $95 price target.
Following the focus on AI computing power, networking is becoming the next super-cycle. Over the past two years, the AI industry chain's focus has concentrated on GPUs and computing servers. However, as model sizes continue to expand, with computing density and data traffic exploding simultaneously, network infrastructure is becoming the new critical bottleneck. In Bank of America's view, this signifies that the AI investment cycle is transitioning from a "computing power cycle" to a "dual-wheel cycle of computing power plus networking." In this process, Scale-Across architecture, Data Center Interconnect, and 800G/1.6T optical modules will collectively propel the optical communications industry into a new long-term growth phase. OFC 2026 may well mark the starting point of this optical module super-cycle.
Comments