China clouds, chips and policy push AI compute expansion
Chinese cloud providers, chipmakers and local governments all issued new AI compute updates, from capacity expansion and chip roadmaps to agent software and procurement deals.
• China Unicom said its AI computing scale has exceeded 45 EFLOPS, while Anhui issued guidelines backing coordinated development of computing power clusters, edge computing and a quantum-AI scheduling platform.
• Alibaba Cloud’s third data center in South Korea went live, bringing its global footprint to 30 regions and 104 availability zones, and Tencent Cloud said it will launch its first cloud region in Malaysia.
• Baidu said AI businesses accounted for 50% of revenue for the second straight quarter, and GPU cloud revenue in its AI cloud infrastructure rose 283% year on year for the fourth consecutive quarter of triple-digit growth.
• Robin Li said Kunlunxin has completed R&D and commercialization of three generations of AI chips.
• Lei Jun said Xiaomi’s next-generation Xuanjie chip is coming soon, while Lu Weibing said Xring O1 has shipped more than 1 million units across three devices and a new generation of Xring chips will also be released soon.
• VeriSilicon said AI computing power-related new orders account for 90%, and another report said its orders reached 15.1 billion yuan.
• StarFive launched Dubhe-100, a server-grade RISC-V HPC CPU IP designed for HPC scenarios and emerging applications such as AI agents.
• Doubao updated Work Task mode with a Windows virtual desktop that can operate the computer, and WeCom 5.0.10 opened CLI and MCP access to mainstream AI agents across 10 office modules.