Mistral enters physical AI race with robotics model
Mistral, the Paris-based AI startup valued at over $6 billion, released its first robotics foundation model on Wednesday, marking a strategic shift from language models into embodied AI. The model enables robots to understand natural language commands and adapt to unstructured environments without task-specific programming. Mistral's move puts it in direct competition with Tesla's Optimus, Boston Dynamics, and Google's DeepMind in the rapidly growing physical AI market.
China may allow top AI firms limited Nvidia H200 purchases
The Information reported that the Chinese government is considering allowing a select group of domestic AI companies — including Baidu, Alibaba, and ByteDance — to buy limited quantities of Nvidia's H200 chips, which are currently restricted under US export controls. The potential policy shift suggests a pragmatic approach: China seeks to maintain AI competitiveness while avoiding a full-scale chip war escalation. Nvidia shares rose 2.3% on the news.
Google's Veo 3 generates cinema-quality video from text
Google launched Veo 3, its latest AI video generation model, capable of producing 4K-resolution clips with consistent character rendering, complex scene composition, and multi-language audio syncing. Early demonstrations show the tool can generate 60-second cinematic sequences from simple text descriptions. The release intensifies competition with OpenAI's Sora, which launched in select Asian markets including Thailand earlier this week.