Guosen Securities stated that the differentiation of the AI industry chain is intensifying: the certainty of optoelectronic modules and PCB is enhanced, while the middle and lower reaches of the chain have been downgraded.

date
05/07/2025
avatar
GMT Eight
On one hand, the upstream computing power hardware represented by PCBs and optical modules is starting to outperform midstream software services and downstream end applications; on the other hand, within the upstream computing power hardware, the North American computing power chain represented by "Easy China" is significantly outperforming domestic computing power chains represented by SMIC and Cambricon.
The Industrial team led by Zhang Qiyao recently published a research report on the AI industry, mentioning that there have been two major differentiations in the AI market recently. The main points of the report are as follows: Since June, although the overall AI sector has seen a recovery, there have been two distinct differentiations internally. On the one hand, the upstream computing power hardware represented by PCB and optical modules has started to outperform the midstream software services and downstream application sides. On the other hand, within the upstream computing power hardware, the North American computing power chain represented by "Yi Zhongtian" has significantly outperformed the domestic computing power chain represented by Semiconductor Manufacturing International Corporation and Cambricon. The core behind these two major differentiations is performance. As tariff concerns ease and factors such as North American cloud companies reiterating their increase in AI capital expenditure, and overseas demand for optical modules next year gradually becoming clearer, the performance certainty of the North American computing power chain, particularly in optical modules and PCBs, has increased. Since Q2, the performance of the North American computing power chain has seen significant upward revisions, while the domestic computing power chain and the AI mid-to-downstream industry chain have seen overall downward revisions. It can be seen that performance is also the main contradiction determining the recent performance of the internal AI industry chain. Since June, the performance of each sector within the AI industry chain is positively correlated with the extent of its upward revisions in Q2 performance. As the interim report period approaches, the technology sector is also beginning to price in performance certainty. Among the sub-sectors of the AI sector that have seen upward revisions in performance since Q2, the main ones include upstream computing power hardware (PCB, GPU, optical modules, IDC), midstream software services (AI Agent), and downstream application sides (digital marketing, publishing, humanoid Siasun Robot & Automation, drones).