Zhipu AI Unveils New Model Trained Entirely on Huawei Chips
Zhipu AI has unveiled a new artificial intelligence model trained entirely on Chinese-made Huawei chips, marking significant progress toward Beijing’s goal of complete technological self-reliance in AI development.
GLM-Image Model Built on Huawei Ascend Hardware
The company, listed as Knowledge Atlas Tech Joint Stock (HK:2513), announced that the open-source image generation AI called GLM-Image was trained end-to-end using Huawei’s Ascend Atlas 800T A2 hardware.
Zhipu is recognized as one of China’s “AI tigers,” a group of startups leading the country’s AI development efforts. The company recently completed a successful public listing debut in Hong Kong, demonstrating strong market confidence in Chinese AI capabilities.
Market Response and Stock Performance
Following the announcement, Zhipu AI shares surged more than 17% to HK$222.60, reflecting investor enthusiasm for domestic AI chip solutions. The news also sparked a broader rally across Chinese chipmaking stocks.
This development represents another milestone in Beijing’s strategy to achieve complete self-reliance in AI technology, with semiconductors at the core of this ambitious goal. Advanced chips are essential for AI development due to the intensive computational requirements of training and operating large language models.
China’s Strategic Push for AI Independence
China has significantly restricted the sale of U.S.-made AI chips, particularly from NVIDIA Corporation, in domestic markets. This policy aims to foster local technological development while reducing dependence on American technology solutions.
While Huawei offers several processors positioned as competitors to NVIDIA’s offerings, industry analysts note that these chips remain less powerful than NVIDIA’s most advanced Blackwell line, which is currently blocked from sale in China under U.S. law.
