On the day when all A-share optical module stocks hit the daily limit up, the capital actually only picked a few. If you hold Zhongji Innolight, Suzhou TFC Optical Communication, or Dongshan Precision, first figure out what you're actually buying.

Conclusion first: On 6/15, the ChiNext board surged 5.30%, and CPO optical modules all hit the daily limit-up, seemingly indicating a broad-based rally in AI hardware. However, capital inflows were far from broad-based—across the entire market, 184 stocks saw net main fund inflows exceeding 100 million yuan, with Zhongji Innolight (Zhongji Xuchuang) alone netting 3.677 billion yuan, ranking first in the entire market. If you hold optical communication stocks in your portfolio, you should pay more attention to this than the "limit-up": Capital is not buying the AI concept; it is meticulously selecting links along the optical interconnect chain, betting on the inflection point in the penetration rate of 1.6T optical modules and CPO.

Let's frame this chain as a sequential story. For greater computing power, GPUs need to be connected with faster "optical fiber pipelines," which are optical modules; the larger the cluster, the usage doesn't increase linearly but non-linearly. This chain is currently stuck at a switching point: 800G pipelines are almost fully deployed, and the next-generation 1.6T is being bid on and sampled by North American cloud providers; meanwhile, a new form factor, CPO, is competing for market share. So the direction is clear (1.6T+CPO needs to ramp up), but who will get the biggest piece of the pie is still being decided.

Who is most valuable? Not the ones with the most limit-ups, but the upstream optical chips (EML/silicon photonics, the core devices for light emission and guidance) that constrain supply, and the module leaders with 1.6T market share. PCB and copper cables are like "trays and short wires," the limit-up is driven mostly by sentiment—the first to hit the limit-up are not necessarily the first to make money.

For the stocks you might hold:

Zhongji Innolight (Zhongji Xuchuang): The leader. The price has already factored in full 800G production and 1.6T leadership, but not the pace of CPO substitution—watch how much market share it can secure in Q3 North American 1.6T bidding and the proportion of 1.6T revenue in its quarterly report.

Eoptolink (Xin Yisheng): The most volatile but also most reliant on a single major North American customer. It surges when the customer orders more and gets revalued when the customer cuts orders—its strength and weakness lie in its concentration.

Dongshan Precision (Dongshan Jingmi): This limit-up (239.3 yuan, market cap 438.3 billion yuan) carries significant sentiment premium. It makes PCB carriers, not the core beneficiary link. Whether it's truly sustainable depends on high-layer PCB orders in the interim report—if orders don't materialize, the sentiment-driven portion will fall back before the gains do.

Next, watch three things: first, the bidding and delivery schedules for 1.6T from North American cloud providers (Amazon, Meta, Microsoft); second, the proportion of 1.6T revenue and gross margin in Zhongji Innolight and Eoptolink's quarterly reports; third, whether NVIDIA's next-generation platform uses CPO. These three things will tell you whether this wave is a real step-up in penetration or just a temporary frenzy.