China memory chips and open-source models draw global attention
China’s AI compute news centered on data infrastructure policy, domestic memory supply-chain tensions, and overseas debate over Chinese open-source models.
• National Data Administration director Liu Liehong visited Shanghai organizations including Shanghai Data Group, Shanghai Data Exchange, Zhiyuan Innovation, Infinigence AI and Lingxi Cloud to research data infrastructure, data circulation and market-oriented allocation reform for data elements.
• China’s Cyberspace Administration said in its National Informatization Development Report (2025) that chip R&D and manufacturing made solid progress, RISC-V applications are accelerating, OpenHarmony has been installed on more than 1.2 billion devices, and generative AI users reached 602 million.
• Apple has reportedly lobbied U.S. officials to use Chinese memory chips from CXMT and YMTC as it seeks to reduce costs, while Micron has campaigned against the move.
• Several of China’s top six smartphone makers have rejected memory suppliers’ proposed price increases for coming quarters, with disagreements also emerging in the AI server sector.
• Nearly 200 Silicon Valley companies, including Proton and Y Combinator, urged the White House not to cut off U.S. access to Chinese open-source AI models such as Kimi K3 and Qwen 3.8.
• NVIDIA, Microsoft, Meta and other companies backed an open letter supporting open-weight AI models, with related coverage citing Chinese models including DeepSeek and Kimi as examples in the policy debate.
• Elon Musk said China is strong in robotics and AI, praised Moonshot AI’s Kimi K3 model, and argued that power supply capacity is a key factor in the AI race.
• The CAC report also said several Chinese large-model products rank among the world’s leaders in performance, while 6G completed first-phase technical trials and brain-computer interface technology is seeing faster adoption.