Alibaba open-sources AI chip software, launches Zhenwu cloud supernode service

Alibaba’s Pingtouge open-sourcing its AI chip software stack and launching the Zhenwu rack-scale supernode marks a strategic pivot from IP vendor to full-stack ecosystem orchestrator. This move accelerates domestic AI frameworks’ decoupling from CUDA, forcing Chinese GPU makers to overhaul driver stacks and compiler toolchains, while compelling cloud providers to redesign heterogeneous resource schedulers. Amid tightening U.S. export controls on advanced compute, this reduces supply chain fragility but risks triggering stricter EDA and IP-core restrictions. NVIDIA may counter by deepening DGX Cloud integration with leading LLM training platforms, while Huawei’s Ascend could leverage closer co-optimization with telecom and government clients. Within 18 months, China’s AI chip sector will enter a ‘software-defined hardware’ shakeout—players lacking robust toolchains and developer communities will be rapidly marginalized.