800G and 1.6T Optical Transceivers Coexist as AI Clusters Push Bandwidth and Thermal Limits 2026-09-01
As AI training and inference clusters continue to scale, data-center interconnects have moved beyond pure bandwidth competition into a systems challenge that balances power, cooling, supply stability, and deployment compatibility. Industry data indicate that 800G optical transceivers remain the volume workhorse in 2026, with unit shipments expected to grow substantially year-over-year, while 1.6T modules have entered the early volume ramp and are becoming a planning baseline for new high-density AI fabrics.
Why 800G and 1.6T Must Coexist
Rising rack power density and GPU counts are driving higher per-port bandwidth requirements in leaf-spine and scale-out networks. Multiple research firms project that shipments of 800G-and-above optics will reach tens of millions of units in 2026, with 800G still accounting for the majority and 1.6T beginning real deployments based on 200G-per-lane architectures.
For operators and system integrators, the transition is not a simple module swap. Existing 800G equipment remains in depreciation cycles, while new projects must reserve a clear upgrade path to 1.6T. Key considerations include whether switch ASICs support 200G/lane electrical interfaces, fiber plant density, cabinet thermal capacity, and the power and heat characteristics of the modules themselves.
Power and Thermal Performance Drive Decisions
Fully retimed 1.6T modules typically consume 23–30 W, noticeably higher than the 14–18 W range common for current 800G modules. This directly affects rack power budgets and cooling strategy. Some supply-chain players are advancing linear pluggable optics (LPO) or receive-side light-retiming (LRO) designs to reduce power further for dense deployments.
In short-reach AI cluster interconnects, multimode VR8 and single-mode DR8 / 2×DR4 variants remain primary choices. Modules must comply with OSFP MSA, CMIS management, and real-world thermal conditions to avoid signal-integrity issues or thermal throttling that increase GPU idle time.
ALT Group High-Speed Optical Response
In response to AI servers’ dual requirements for high-speed transmission and rapid deployment, ALT Group (Aeon Lighting Technology) offers 800G OSFP VR8 short-reach solutions within its optical communications portfolio and continues development of 1.6T OSFP 2×DR4 silicon-photonic transceivers. Design priorities emphasize deployment compatibility across different AI server and network architectures, supporting scale-out links from TOR switches to GPU servers.
Looking ahead, 800G and 1.6T will coexist for the foreseeable future. Selecting optical transceivers with stable supply, well-characterized power profiles, and strong interoperability focus helps reduce uncertainty during AI cluster expansion and improves overall compute utilization.
About Aeon Lighting Technology Inc.
Aeon Lighting Technology Inc. is a trusted global partner in AI infrastructure solutions, with deep expertise in data center architecture, high-density power distribution, intelligent cooling systems, and network cabling. Guided by rigorous quality control and compliance with international standards, the company is committed to helping enterprises and partners worldwide accelerate digital transformation and advance toward the next generation of intelligent computing.
For more information, visit the Aeon Lighting Technology website: http://www.aeonlighting.com/