On the morning of the 24th, computing hardware such as CPO optical modules showed strong performance, with Zhongji Innolight rising nearly 3% intraday to a new record high, while Ingenic Semiconductor surged over 5%. The ChiNext AI sector bucked the broader market trend. Among popular ETFs, the largest and most liquid ChiNext AI ETF (159363) gained over 1% intraday, with frequent premium trades indicating active buying interest and a net subscription of 28 million units in real-time.
According to Reuters, NVIDIA has informed Chinese clients that it plans to deliver its second-most powerful AI chip, the H200, by mid-February next year—before the Lunar New Year. Sources revealed that NVIDIA will use existing inventory to fulfill initial orders, with shipments estimated at 5,000 to 10,000 chip modules, equivalent to 40,000–80,000 H200 chips.
Changjiang Securities highlighted continued focus on AI computing power in 2026. For optical modules, both AI training and inference are driving clear and sustained growth. The high demand for AI computing persists, with expanding needs for training and inference, further solidifying the network-side expansion logic. As model capabilities improve and token costs decline rapidly, inference is becoming the primary driver of computing demand, boosting intra-data-center and cross-rack interconnect needs.
Structurally, ASICs are gaining traction due to their superior total cost of ownership (TCO), potentially increasing optical modules' share in overall IT capital expenditures compared to general-purpose GPUs. Additionally, the shift from "in-rack" to "cross-rack, clustered" architectures in scale-up designs enhances optical interconnect intensity, opening new growth opportunities for optical modules in scale-up scenarios.
To capitalize on computing opportunities centered on optical modules, investors may consider the ChiNext AI ETF (159363) and its feeder funds (Class A: 023407, Class C: 023408). The underlying index heavily weights leading optical module stocks, with over 56% exposure. The portfolio allocates over 70% to computing power and more than 20% to AI applications, efficiently capturing AI-themed trends (data as of November 30, 2025).
As of December 23, the ChiNext AI ETF (159363) leads its peers with assets exceeding ¥3.8 billion and an average daily turnover of over ¥600 million in the past month—the highest among seven ETFs tracking the ChiNext AI Index.
*Data source: SSE/SZSE. Note: "First in the market" refers to the first ETF tracking the ChiNext AI Index.
**Risk Disclosure**: The ChiNext AI ETF passively tracks the ChiNext AI Index (base date: December 28, 2018; launch date: July 11, 2024). The index’s annual returns from 2020–2024 were 20.1%, 17.57%, -34.52%, 47.83%, and 38.44%. Index constituents are adjusted per its methodology, and past performance does not indicate future results. Constituent mentions are illustrative and not investment advice or indicative of fund holdings. The fund is rated R4 (higher risk) and suitable for aggressive (C4+) investors. Investment decisions carry risks; historical returns do not guarantee future performance.
Comments