DeepSeek and Huawei Open Software Tools to Challenge Nvidia's AI Chip Dominance
DeepSeek and Huawei Technologies unveiled a suite of open-source programming tools on September 30 designed to reduce developers' reliance on Nvidia's dominant CUDA software ecosystem, directly attacking one of the semiconductor giant's most formidable competitive advantages. The package includes six software modules—DeepGEMM, FlashMLA, TileKernel, DeepSelect, DeepEP, and TileLang—optimized for Huawei's Ascend 950 accelerators and freely available to download. TileLang, DeepSeek's high-level programming language, is positioned as a simpler alternative to Cuda, with DeepSeek describing it as offering "a simpler programming model" that reduces barriers to entry.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.