In a recent investor briefing, Zhongji Xuchuang disclosed that the company has already received clear guidance and confirmed share of 800G demand from key customers for next year, indicating a demand growth several times higher than this year, necessitating a rapid increase in production capacity. The company has been expanding its capacity since the second quarter of this year, and production capacity will continue to increase step by step in each quarter, with expansion expected to continue until the first half of next year, both domestically and overseas. If next year's customer demand for 800G further increases, the company still has ample capacity expansion space.

Currently, the mainstream technology solution adopted by 800G customers is still the pluggable traditional solution (EML+DSP or silicon photonics+DSP), and CPO is still too early for the 800G stage. LPO as an alternative solution is still being tested by some customers, but the majority currently prefer the traditional solutions mentioned earlier.

Zhongji Xuchuang stated that the demand for GPUs in the entire AI market is still ongoing, and the demand for accompanying 800G optical modules is continuous and stable.

Furthermore, major AI customers have already explicitly expressed demand for 1.6T to match future GPU demands with larger bandwidth and higher computing power. It is expected that the progress of 1.6T in 2024 will mainly involve testing, certification, and small batch demand, with true large-scale production starting in 2025.

At the same time, Zhongji Xuchuang expects that the demand for 800G will continue into 2025 for several reasons: Some customers' AI processes and deployment rhythms are set for 2025; the technological iteration of cloud computing data centers is driving the demand for 800G. Traditional cloud data centers in 2024 will still mainly demand 400G. Large-scale production of 800G in data centers is expected to begin in 2025. The trend of 800G demand overlaying with 1.6T in 2025 is already quite clear.