DeepSeek says it has partnered with Huawei to develop Ascend chip programming tools
DeepSeek, a Chinese artificial‑intelligence firm, announced a partnership with Huawei Technologies to create programming tools for Huawei’s Ascend AI chips. The collaboration will focus on building a suite of development utilities, including TileLang, an open‑source alternative to NVIDIA’s CUDA framework.
Key points
- DeepSeek and Huawei announced a partnership to build programming tools for Ascend AI chips
- The effort includes TileLang, an open‑source CUDA‑compatible language
- Both firms said development work has already started, but no timeline was given
TileLang is intended to give developers a familiar, high‑performance language for writing code that runs on Ascend processors, potentially easing the transition for teams accustomed to CUDA. Both companies said the effort aims to broaden the software ecosystem around Ascend chips and accelerate AI workloads on Huawei hardware.
The partnership was disclosed in a Reuters brief on Wednesday, with DeepSeek stating the joint work is already underway. No further details on timelines, licensing or commercial terms were provided.
The story so far
2 episodes →- DeepSeek says it has partnered with Huawei to develop Ascend chip programming toolsthis story
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- AMD says PerfOpt in Linux 7.4 may boost AI/LLM performance on Radeon iGPUs by up to 23% · 1 src
- Netlist sues for US import ban on Micron chips in Google, Nvidia AI hardware · 1 src
- Google to launch first Project Suncatcher satellite with TPUs on SpaceX Transporter-18 in October · 16 src
- MachGen AI cuts video generation time by 6x with diffusion-focused infrastructure · 1 src
- Meta brings Muse agent to AI glasses, launches VR and audio models · 32 src
Comments
via GitHub Discussions