Optical Modules for Data Centers

Modern data centers increasingly rely on 800G and 1.6T optical modules, including QSFP-DD, OSFP-XD, and co-packaged optics, to meet high-bandwidth, low-latency requirements for AI and hyperscale workloads.

Key Optical Module Types

1. QSFP-DD and OSFP Modules

• QSFP-DD (Quad Small Form-factor Pluggable Double Density) and OSFP (Octal Small Form-factor Pluggable) are widely used for 400G and 800G deployments, providing high-density, pluggable solutions for leaf-spine and top-of-rack architectures .
• OSFP-XD and next-generation OSFP-224 are designed for 1.6T deployments, addressing power, heat, and signal integrity challenges at 224 Gbps per lane . 2. Co-Packaged Optics (CPO)
• CPO integrates optical transceivers directly onto the switch ASIC, reducing electrical losses, power consumption, and improving signal integrity .
• Silicon photonics-based CPO switches are being adopted by hyperscalers like NVIDIA to eliminate pluggable transceivers entirely, achieving significant power savings . 3. Linear Pluggable Optics (LPO)
• LPO modules remove the DSP chip from the transceiver, relying on the host platform's DSP, which reduces latency by up to 90% and lowers power consumption from 7–9W to 2–4W for 400GbE transceivers .
• LPO is ideal for short links, such as switch-to-switch or GPU-to-GPU connections within racks, supporting high-performance computing and AI fabrics . 4. Coherent Pluggable Optics
• Coherent optics, including 400G ZR/ZR+ and future 800G ZR+ modules, are used for data center interconnects over distances from 40 km to over 1,000 km .
• These modules simplify DCI architecture, reduce power and space requirements, and support low-latency, high-bandwidth connections between clusters .

Deployment Trends

• 800G modules are now the baseline for new AI and hyperscale data center deployments, with shipments growing rapidly and expected to dominate new buildouts .
• 1.6T modules are entering production and will see targeted deployment between 2026 and 2028, particularly in large AI training clusters .
• eXtra-dense Pluggable Optics (XPO) offer up to 12.8Tbps per module with liquid cooling, enabling up to 204.8Tbps in a single rack unit, supporting pay-as-you-grow models and diverse optical architectures (DR, FR, LR, SR, ZR) for AI workloads .

Considerations for Data Center Operators

• Bandwidth and Latency: AI and GPU clusters require terabits per second of aggregate bandwidth with minimal latency, driving adoption of high-speed optics .
• Power and Cooling: Advanced modules like XPO and CPO reduce power consumption and improve thermal management, critical for dense deployments .
• Scalability and Serviceability: Pluggable modules allow quick replacement and upgrades without servicing the entire switch, supporting operational flexibility .
• Future-Proofing: Operators should plan for 800G as the baseline and consider 1.6T modules for high-performance AI clusters to accommodate exponential growth in data traffic . In summary, data centers require a mix of high-speed pluggable optics (400G–1.6T), co-packaged optics, and linear pluggable solutions to meet the demands of AI workloads, hyperscale cloud networks, and low-latency interconnects, with careful attention to power, cooling, and scalability.