Moonshot AI pauses new Kimi subscriptions
Moonshot AI announced Sunday that it has paused subscriptions for new retail users of its Kimi K3 model due to an unexpected computing-power shortage. User requests over the past 48 hours far exceeded projections, nearing the limits of the company's existing computing clusters, the Chinese AI startup said. The company said it will dedicate all current computing power to existing subscribers while expanding capacity, and it planned to gradually reopen subscriptions as new computing power becomes available.
Alibaba hypes up Qwen3.8
Alibaba Group Holding Ltd. announced Sunday that its Qwen3.8 large language model, featuring 2.4 trillion parameters, will soon be released and open-sourced. Following Apple Intelligence’s completion of its domestic filing, Alibaba told Caixin that Qwen will be integrated into Apple Intelligence as an AI capability to provide services for users in China across iOS, iPadOS, macOS and visionOS. The Qwen3.8-Max preview model is already available on Alibaba’s Token Plan, Qoder and QoderWork.
SenseTime plans to launch computing satellites
Hong Kong-listed SenseTime Group Inc. announced Saturday at a forum during the World Artificial Intelligence Conference (WAIC) that it had formed a strategic partnership with Adaspace Technology Co. Ltd. to launch four satellites in 2026 for early technical validation and prototype testing. The company plans to build a network of 1,000 satellites within five years to provide at least 10,000 petaflops of space-based computing power.
Biren Technology releases new super-node architecture
Hong Kong-listed Shanghai Biren Technology Co. Ltd. unveiled its next-generation near-packaged optics (NPO) optical interconnect and distributed decoupled architecture super-node solution at the 2026 WAIC on Friday. Ding Yunfan, the company’s vice president of AI framework architecture, said that a single GPU’s computing power can no longer independently handle complex tasks, and the super-node architecture was designed to connect massive numbers of GPUs at high speeds to address this issue.
Tencent Cloud sees enterprise AI agent demand double
Demand for enterprise-level AI agents has at least doubled year-on-year so far this year, Wu Yunsheng, vice president of Tencent Cloud and head of Tencent Cloud AI, told Caixin and other media at the 2026 WAIC on Saturday. The core challenge in implementation lies in integrating AI with enterprises’ existing data and workflows, Wu said, adding that much of this information remains in traditional systems that require modification or new interfaces.
Contact editor Kelsey Cheng (kelseycheng@caixin.com)
