ETF abnormal movement | PP Technology Innovation 50 (03151) rises nearly 5%, with the performance of most domestic AI chip stocks skyrocketing. The market is highly concerned about the prospects of computing power training adaptation.

2026-04-30 15:43

Zhitongcaijing
PP Innovation 50 (03151) gained nearly 5% in the final trading hours, as of the time of writing, it increased by 4.88% to 10.96 Hong Kong dollars, with a turnover of 45.967 million Hong Kong dollars.
PP Science and Technology Innovation 50 (03151) rose nearly 5% at the end of the day. It is understood that PP Science and Technology Innovation 50 tracks the Shanghai Science and Technology Innovation 50 Index, with a high concentration of holdings in the semiconductor and other technology sectors. The top ten major holdings include Cambricon, Allwinner Information, SMIC, Unisoc, and Biwin Storage. As of the deadline, it rose 4.88%, closing at 10.96 Hong Kong dollars, with a turnover of 459.67 million Hong Kong dollars.
On the news front, on the evening of April 29th, "AI ASIC leader" Unisoc released its 2026 first quarter report, achieving operating income of 836 million yuan, a year-on-year increase of 114.47%; net profit attributable to the parent company was -341 million yuan, with a slightly expanded year-on-year loss. As of April 29th, the company's new signed order amount hit a new high, reaching 8.24 billion yuan, with AI computing-related orders accounting for 91.37%, and data processing orders accounting for 90.15%, mainly from cloud AI ASIC and IP. It is noteworthy that as of April 30th, Cambricon reported good performance in the first quarter, with its stock price approaching the limit up line. In addition to Cambricon, many other AI chip companies such as Moore Thread and Allwinner Information have also shown good performance.
Recently, the official launch of DeepSeek-V4 was announced. East China Securities believes that DeepSeek V4 is an attempt to use domestic computing power for training large models from 0 to 1. Previously, domestic large models used domestic computing power for inference, but this time, from model core to training architecture, and to the entire inference process, there is the shadow of domestic computing power, which is an important milestone. Therefore, regardless of how DeepSeek V4 performs, its strategic significance is very important, and the future prospects of domestic computing power training adaptation are the focus of attention.