Huawei's Ascend supernode has crossed the easy threshold: it can be announced. Reuters reported that DeepSeek's V4 preview was adapted for Huawei chip technology, and Huawei said its Ascend 950 supernode would support the model. [1][2]

The paper's Sunday brief on Huawei's ecosystem test asked whether CANN and Ascend could carry real inference traffic. Monday sharpens the test. The question is no longer whether DeepSeek can run on the stack. It is whether any customer beyond the Chinese national champion circle treats it as infrastructure.

South China Morning Post reported Huawei's claim of day-zero adaptation and noted that V4 throughput problems may persist until Ascend 950PR supernodes ship at scale. [3] That is the operational sentence. Sovereignty narratives travel faster than racks.

X wants the phrase "de-CUDA" to do the work. Mainstream coverage prefers self-reliance. The customer question is colder: who signs, who deploys, who absorbs outages, and who lets a non-Nvidia frontier stack become procurement rather than slogan?

-- DAVID CHEN, Beijing