Co-Packaged Optics (CPO) Nears Mass Production: Interfaces, Yield, and Maintainability Emerge as Next Key Challenges

September 20, 2026

DIGITIMES Taiwan

Overview

As AI computing bandwidth demands escalate, optical interconnect architectures are shifting from pluggable optical modules to near-packaged optics (NPO) and co-packaged optics (CPO), moving closer to mass production. Successful high-volume deployment of these technologies requires addressing critical challenges beyond optical-engine performance, including precision alignment, serviceability, manufacturing yield, and advancing supply chains for light sources, packaging, and testing.

In Depth

Key Findings

Amidst the escalating bandwidth demands of AI computing, optical interconnect architectures are transitioning from conventional pluggable optical modules to chip-proximate near-packaged optics (NPO) and co-packaged optics (CPO), now approaching mass production. With this evolution, the next critical challenges revolve around establishing robust interfaces, improving manufacturing yield, and ensuring practical maintainability.

Technical / Clinical Details

NPO and CPO technologies aim to resolve data transfer bottlenecks and significantly enhance power efficiency by integrating the optical engine near or within the same package as the electrical chip. This enables high-speed communication between GPUs and large-scale data processing within AI clusters. However, the widespread adoption of these technologies hinges on overcoming several technical and manufacturing hurdles. Specifically, establishing precise alignment techniques between optical fibers and optical engines is paramount; even slight misalignments can lead to signal loss, necessitating highly accurate assembly processes. Maintainability is another challenge: unlike traditional pluggable modules, CPO’s integration with the chip makes replacement or repair more complex in case of failure. Furthermore, improving manufacturing yield is directly tied to cost reduction and mass production. This requires the development of a robust, comprehensive supply chain, including reliable light source provision, advanced packaging technologies, and efficient testing processes.

Background & Context

The rapid advancements in AI and High-Performance Computing (HPC) are placing unprecedented demands on data transfer volumes and processing speeds within data centers. Traditional electrical interconnects are facing limitations due to signal degradation, increased power consumption, and physical distance constraints, hindering the scalability of AI infrastructure. Consequently, the shift to optical interconnects, particularly integrated solutions like CPO, is an inevitable trend. Numerous major semiconductor manufacturers and cloud service providers are investing in CPO technology, anticipating its role as a foundational element for AI-era data centers.

Strategic Significance & Outlook

As CPO technology moves into mass production, the standardization of interfaces, optimization of manufacturing yield, and simplification of maintainability will be crucial factors determining market penetration. If these challenges can be overcome through industry-wide collaboration, CPO is poised to become the dominant technology in AI data centers, enabling further advancements and applications of AI. Innovation in related technologies such as light sources, packaging, and testing, along with the strengthening of the supply chain, will be key to CPO’s long-term success.

Source: https://www.digitimes.com/news/a20260914PD208/cpo-production-bandwidth-optics-manufacturing.html

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.