Computing Power Optical Module Optical Communication

Optical modules like LPO, NPO, and CPO are critical for high-speed, low-power data transmission in AI and high-performance computing, bridging electrical systems with optical networks efficiently.

Overview of Optical Modules

Optical modules serve as the interface between electrical systems (servers, switches, GPUs) and optical fiber networks, converting electrical signals into light for transmission and back into electrical signals at the receiver end. This enables high-bandwidth, low-latency communication over long distances, which is essential for AI training, cloud computing, and 5G networks .

Key Types of Optical Modules

1. Linear Pluggable Optics (LPO)

• LPO modules are pluggable and cost-effective, suitable for short-distance interconnects in back-end networks.
• They reduce power consumption and latency but rely on the linearity of host-side SerDes and lack advanced DSP-based error correction, making them more suitable for single-vendor or closed systems .
• Example: 800G OSFP 2xDR4/DR8 modules provide energy-efficient, low-latency interconnects . 2. Near-Packaged Optics (NPO)
• NPO places the optical engine adjacent to the xPU (GPU, NPU, or switch chip) on the same PCB, connected via extremely short electrical paths.
• This reduces channel loss, improves signal integrity, and supports high bandwidth (800G+), while maintaining independent thermal management for easier cooling and maintenance .
• NPO acts as a transitional technology toward more integrated solutions like CPO. 3. Co-Packaged Optics (CPO)
• CPO tightly integrates the optical engine with the switch ASIC or processor in the same package, reducing electrical I/O distance from centimeters to millimeters.
• Advantages include higher bandwidth density, lower power consumption, improved signal integrity, and reduced latency .
• CPO is ideal for front-end, high-speed, high-density interconnects such as spine-layer networks in AI data centers.
• It represents a fundamental shift from module-level to chip-level optical integration, enabling next-generation AI cluster performance .

Supporting Technologies

• Silicon Photonics: Fabricates optoelectronic devices on silicon, combining photonics and electronics for superior performance .
• Electro-absorption Modulated Lasers (EML): High-speed lasers for 800G and 1.6T applications.
• Thin-Film Lithium Niobate (TFLN): Emerging material for ultra-high-speed modulators.
• Advanced Modulation Formats: Techniques like PAM6 at 174 Gbaud increase data throughput without faster hardware .

Evolution and Trends

The evolution path moves from pluggable modules → LPO → NPO → CPO, reflecting a trend toward higher integration, lower power, and reduced latency. While pluggable modules remain flexible and essential for current deployments, CPO is the long-term direction for AI data centers, enabling unprecedented bandwidth and energy efficiency . In summary, optical modules are key enablers of modern computing power, supporting AI and big data workloads by providing scalable, low-latency, and energy-efficient interconnect solutions.