Huawei and DeepSeek Challenge Nvidia with New AI Software
DeepSeek and Huawei are taking on Nvidia's CUDA programming interface with their new software, TileLang. The company claims it is simpler to program than CUDA while still utilizing chips to their upper limits.
Towards this goal, DeepSeek has built the new software alongside Huawei for use on its Ascend chips. Huawei plans to release its next-gen Ascend chips early next year, and TileLang will be used to program them.
CUDA was a major breakthrough in parallel computing when Nvidia unleashed it in 2006, and it has become the industry's go-to for building neural networks on GPU chips. However, CUDA alternatives have been around for a decade and change, with AMD's ROCm and Intel's oneAPI failing to gain traction among developers.
TileLang could help China separate itself from the US and Nvidia specifically, as Huawei has largely conceded China's AI-chip market to the Chinese tech giant. However, it seems unlikely to turn many heads among the millions of developers that depend on CUDA.