On July 31, Huawei released the weights, inference code, and technical report for openPangu-2.0-Pro, a 505-billion-parameter language model that completed its full pretraining run on Huawei's own Ascend 910B NPUs, with no Nvidia hardware involved at any stage. It is the first model at 500 billion parameters or above to credibly make this claim, as prior frontier-scale open-weight models like DeepSeek V4 Pro, GLM-5.2, and Qwen 3.7 all touched data centers loaded with Nvidia A100s or H100s. The model is live on GitCode's Ascend Tribe community and accessible through Huawei Cloud ModelArts Studio. However, the Ascend 910B NPUs were manufactured primarily on SMIC's N+2 process, a 7-nanometer-class node using DUV lithography, as ASML is barred from shipping EUV equipment to China. TechInsights teardowns found that essentially every Ascend 910B and 910C chip examined contained dies fabricated at TSMC's 7nm node, not SMIC's. Huawei had acquired approximately 2.9 million of those TSMC-fabricated dies through Sophgo, a scheme US authorities determined violated export controls, with TSMC facing a potential penalty of $1 billion or more. As of early 2026, the TSMC die bank is effectively exhausted, and future Ascend