U.S. tells Apple not to buy Chinese memory chips
The Trump administration does not want Apple Inc. to source memory chips from the Chinese mainland, U.S. Commerce Secretary Howard Lutnick said in an interview with The Wall Street Journal on Friday. Following a tour of an Apple manufacturing facility in Houston, Lutnick said that although Apple might still do so, the administration does not favor the move, adding that there have to be “other solutions to the memory issue, but it’s not great American companies using Chinese memory.” He confirmed that the stance had been communicated to Apple. The tech giant has been testing memory chips produced by Chinese manufacturers CXMT Corp. and Yangtze Memory Technologies Co. Ltd., sources said.
Alibaba open-sources Qwen AI models
Alibaba Group Holding Ltd. announced on Friday that it has open-sourced its Qwen3.8 series of AI models. The release includes Qwen3.8-27B, a native multimodal dense model with only 27 billion parameters that outperforms the previous Qwen3.7-Plus. Released under the Apache 2.0 license, the new model allows for free downloading, deployment and commercial use. It responds quickly, is easy to deploy and can run smoothly on consumer-grade graphics cards after quantization.
Z.AI launches new AI model
Chinese AI developer Z.AI Co. Ltd. launched its latest large language model, GLM-5.3, on Friday, highlighting its enhanced programming and network security capabilities. The Hong Kong-listed tech firm said that GLM-5.3’s base model remains unchanged from its predecessor GLM-5.2, but its intelligence upper bound has been significantly raised through extreme post-training scaling.
Embodied AI startup Light Origins raises new funds
Embodied AI startup Light Origins announced on Friday that it has completed a pre-A funding round, raising hundreds of millions of yuan. The round was led by CAS Investment Management, with participation from China Merchants Venture Capital and Xianghe Capital. Founded in 2024 by Jiang Xu, a former OpenAI algorithm expert who worked on models ranging from InstructGPT and ChatGPT to GPT-4, the company focuses on building foundation models for embodied AI.
.jpeg?w=600)
