ByteDance and Kuaishou lead US rivals in video generation — their short-form video training corpora from TikTok and Kuaishou are a structural data moat compute alone cannot close.
China's Ling-2.6-1T—a publicly available, inspectable trillion-parameter model—claims to use fewer tokens than comparable US efficient models.
Huawei projects $12B AI chip revenue in 2026 (+60%), driven by Ascend 950PR orders as Nvidia loses ground in China to export controls.
DeepSeek V4 is out: 1.6T open-source parameters, 1M token context, 3.7× fewer FLOPs than V3.2, scoring 120/120 on Putnam 2025.
Huawei confirms its Ascend 950-based supernode will fully support DeepSeek V4, completing China's domestic frontier AI hardware-software stack.
Curated AI insights — sent when there's something worth your inbox.