DeepSeek is systematically culling its dependence on NVIDIA hardware, which remains a significant source of consternation for China’s AI labs due to the wild gyrations in its mass availability. To do so, China’s premier AI lab is now reportedly turning towards Huawei and its Ascend 950DT SuperPoDs/Supernodes to conduct a full training run for the upcoming V5 model. If successful, this would be China’s first AI training run that is completed on a wholly domestic hardware stack, complemented by the CUDA-like ‘TileLang’ kernel compiler. DeepSeek has ordered around 160,000 Huawei Ascend 950DT chips, with deliveries slated to ramp up in […]
Read full article at https://wccftech.com/china-going-for-domestic-full-stack-ai-dominance-deepseek-will-train-its-upcoming-v5-model-from-scratch-on-huaweis-ascend-950dt-based-atlas-superpods/