PP CSI STAR 50 (03151) surged close to 5% during the late trading session. The ETF tracks the SSE STAR 50 Index, with its holdings heavily concentrated in the technology sector, particularly semiconductors. Its top ten holdings include Cambricon, Hygon Information, SMIC, VeriSilicon, and BIWIN Storage. At the time of writing, the ETF was up 4.88%, trading at HKD 10.96, with a turnover of HKD 4.5967 million.
On the news front, on the evening of April 29th, VeriSilicon, a leading AI ASIC company, released its first-quarter 2026 financial report. The company reported revenue of RMB 836 million, a year-on-year increase of 114.47%. However, its net profit attributable to shareholders was a loss of RMB 341 million, representing a widened loss compared to the previous year. Notably, as of April 29th, the company's new order bookings reached a record high of RMB 8.24 billion. Orders related to AI computing power accounted for 91.37% of the total, while orders from the data processing segment made up 90.15%, primarily driven by cloud-side AI ASIC and IP.
Furthermore, as of April 30th, Cambricon reported positive first-quarter earnings, with its stock price approaching the 20% daily gain limit during trading. In addition to Cambricon, other AI chip companies such as Moore Threads and Hygon Information also posted favorable financial results.
Recently, DeepSeek-V4 was officially launched. Analysts at Soochow Securities believe that DeepSeek V4 represents a significant step from zero to one in utilizing domestic computing power for large model training. Previously, domestic large models employed local computing resources primarily for inference tasks. DeepSeek's latest model, however, incorporates domestic computing power throughout the entire process—from the model kernel and training architecture to inference—marking an important milestone. Consequently, regardless of DeepSeek V4's performance, its strategic importance is substantial, with the primary focus being the prospects for adapting domestic computing power to training needs.