DeepSeek V4 Delay Tests China's AI Ambitions

The global tech industry awaits DeepSeek's next-generation V4 model, over a year after the Chinese startup stunned markets with its low-cost R1 chatbot in January 2025. The model is expected to be multimodal and potentially more affordable than rivals, with analysts warning it could again shake US tech valuations. A key question surrounds which chips power V4. Reports suggest it may run on Huawei processors, signaling a potential breakthrough in China's push for AI self-sufficiency amid US export restrictions on Nvidia hardware. Some reports allege DeepSeek may have also used smuggled Nvidia Blackwell chips for training, claims Nvidia has dismissed as farfetched.