Optical interconnects are emerging as the most critical bottleneck and opportunity within AI infrastructure. As the OFC 2026 exhibition prepares to open this month in Los Angeles, the global optical communications supply chain will gather to showcase the latest advancements in next-generation connectivity technologies. Driven by the intense bandwidth demands of AI data centers, 800G optical modules are transitioning from pilot projects to mainstream adoption, while 1.6T products are entering the volume production ramp-up phase, signaling a structural transformation in optical interconnect technology roadmaps.
According to analysis from SemiVision Research, the rapid expansion of generative AI model sizes is shifting the core data center bottleneck from transistor performance to interconnect bandwidth and latency. The concept often referred to in the industry highlights that while computing power can be continuously enhanced through process node advancement and 3D packaging, the rate of improvement for chip-to-chip I/O speeds lags behind, creating a significant "I/O wall." This structural challenge is accelerating the migration of optical components closer to the compute and switch chips to overcome limitations in power consumption, signal loss, and transmission distance.
The OFC technical conference is scheduled for March 15-19, with the exhibition running from March 17-19 at the Los Angeles Convention Center. Key topics will include technology paths for 1.6T and 3.2T optical modules, co-packaged optics (CPO) and novel optical I/O architectures, silicon photonics heterogeneous integration, and the ongoing debate between pluggable and CPO solutions.
**Copper Reaches Physical Limits, Optics Move In-Package**
The exponential growth in AI cluster scale is positioning optical interconnect technology at the very core of infrastructure architecture. SemiVision Research notes that traditional data centers typically use copper cables for short-reach connections within a rack, while pluggable optical modules handle longer inter-rack links. However, as SerDes speeds increase to 200 Gb/s per lane, the physical properties of copper are becoming a critical bottleneck.
A report indicates that at these speeds, traditional passive copper cables can no longer reliably span even a single server rack, facing distance limitations even within a rack. This physical constraint is driving the continued migration of optical components directly alongside compute and switch chips, aiming to establish a new balance between power, loss, and reach.
**800G Goes Mainstream, 1.6T Nears Volume Production**
From a product evolution perspective, the optical transceiver market is showing a clear upgrade cadence. 800G products entered a period of strong growth in 2025 and are accelerating their penetration into the mainstream market. Meanwhile, 1.6T products began their production ramp-up in the second half of this year.
At the product level, Accelink has publicly demonstrated and provided samples of its 1.6T OSFP224 DR8 transceiver tailored for AI data center applications. SemiVision Research analysis suggests that as AI clusters rapidly expand to encompass hundreds of thousands of GPUs, optical interconnects have become one of the most critical bottlenecks in the AI infrastructure stack, as well as one of the most investable sectors.
**CPO vs. Pluggable Debate Takes Center Stage**
The debate over technology roadmaps, specifically co-packaged optics (CPO) versus pluggable solutions, will be a central topic at OFC 2026. CPO, which integrates the optical engine and switch chip on a common substrate, can significantly reduce power consumption and signal loss, positioning it as a long-term evolution path for hyperscale data centers. Pluggable solutions, however, maintain dominance in the near term due to their flexibility and serviceability.
SemiVision Research indicates that industry leaders are scheduled to speak at OFC on manufacturability challenges, an issue directly impacting the timeline for CPO's transition from technical validation to large-scale commercial deployment. Concurrently, the evolution path from pluggable to linear-drive pluggable optics (LPO) and onward to CPO, along with the resulting supply chain restructuring, will also be a key focus.
**Silicon Photonics and Laser Tech as Key Supply Chain Variables**
At the foundational technology level, silicon photonics heterogeneous integration—encompassing the fusion of silicon photonics platforms with materials like thin-film lithium niobate (TFLN) and III-V compounds—is highlighted as a major area of focus for OFC 2026. SemiVision Research points out that supply constraints for laser technologies have become a key factor limiting the scale-up of optical interconnect solutions.
Furthermore, the competition between VCSEL and MicroLED technologies for ultra-short-reach optical interconnects, as well as optical solutions designed to replace copper cables within AI systems, will receive significant attention. Notably, Optical I/O (OIO), a technology enabling native optical links at the package level, is emerging as a promising new direction capable of supporting disaggregated deployments for AI systems.
**Outlook: A Strategic Window for Optical Interconnects**
OFC 2026 convenes at a pivotal moment as demand for optical interconnects in AI data centers accelerates. The simultaneous progression of multiple technology curves—from 800G to 1.6T product iterations, the architectural shift from pluggable to CPO, and the optical replacement of copper for short-reach links—collectively forms the technological foundation of the current optical interconnect revolution.
SemiVision Research concludes that as AI clusters continue to scale to hundreds of thousands of GPUs, optical interconnects have transcended their traditional role as supporting infrastructure. They are now a core variable determining the performance ceiling of AI infrastructure. The technical consensus and industry directions established at this year's OFC are expected to profoundly influence the optical communications landscape for years to come.