Framework
Huawei's Compute Architecture for Neural Networks, the software toolkit for Ascend AI chips and Huawei's counterpart to Nvidia's CUDA.
Huawei claims its Ascend chips have outsold Nvidia in China as of October 2026, but DeepSeek keeps its training work on Nvidia hardware. Unit sales and training workloads are separate measures, and sizing either company's China business needs both.
Reality
DeepSeek open-sourced six software components for Huawei's Ascend AI chips on September 30, porting infrastructure it first built for Nvidia hardware. The ports cut the software bill for leaving CUDA, and The Neuron argues Ascend's case now turns on chip supply, reliability and full-scale model performance.
Perspective Coverage
4 publishersReality
The Atlas 960E fits 8 EFLOPS behind Huawei's own optical engines, and the Ascend 960DT arrives three quarters early, while rotating chairman Eric Xu says supply goes to Chinese customers first and dates China's catch-up with its own demand to 2030.
Perspective Coverage
4 publishersReality
CEO Liang Wenfeng said in investor discussions that Huawei starts delivering AI training chips around Q4 2026 or Q1 2027, and the Inner Mongolia fleet under consideration is 160 times the Ascend cluster DeepSeek ran in June 2026.
Reality
TrendForce now expects domestic silicon to take about 90% of China's AI accelerator market in 2026, up from 45%. The arithmetic says the constraint is fab and packaging output, not Washington.
Reality