AI & Enterprise
China\'s top AI still dependent on Nvidia as shift to domestic chips slows
Many of China’s most advanced AI models are still being trained on Nvidia chips, as local developers struggle to switch because of the cost and technical burden of moving away from Nvidia’s CUDA software ecosystem, the South China Morning Post reported. Engineers said migrating training workflows to Huawei’s Ascend platform can require extensive code rewrites and raise time and costs by at least 50 percent. The report said domestication is progressing faster in inference than in training.
