As generative AI models scale, traditional pluggable optical transceivers face a power-performance wall. Co-packaged Optics (CPO) and Silicon Photonics address this by integrating optical engines directly onto the switch substrate, dramatically slashing power and latency. While alternatives like. While DSPs effectively improve signal quality, their high power consumption and additional latency become major bottlenecks limiting system efficiency. To address this, Macom and NVIDIA first proposed Linear-drive Pluggable Optics (LPO) in 2022. Its core concept is to remove digital processing. In 2026, the large-scale expansion of AI computing power clusters continues, and the industry's supply-demand contradiction gradually shifts upward, transmitting from the traditional optical module assembly capacity to the core upstream optical chip segment. CPO, a technology that deeply co-packages the optical engine with the switch chip, offers a solution for. Traditional architectures based on front-panel pluggable optics are increasingly unable to meet these demands, prompting a shift toward more integrated approaches such as Near-Packaged Optics (NPO) and Co-Packaged Optics (CPO). From Pluggables to Co-Packaged Optics: Front-Panel Pluggable Optics. NVIDIA's networking innovations, including Spectrum-X Ethernet and NVIDIA Quantum InfiniBand, are designed to handle the high-bandwidth and low-latency demands of modern AI training and inferencing at scale.