DeepSeek V4 Launches on Huawei Ascend — China's Full-Stack AI Independence Validated

DeepSeek released its V4 model preview on April 24, 2026 — one year after R1 disrupted global AI markets. V4-Pro features 1.6 trillion parameters, a 1 million token context window, and validated native compatibility with Huawei Ascend NPUs, marking a decisive step in China's CUDA-independence push. Huawei confirmed Ascend 950 chips fully support V4 through its 'Supernode' cluster technology. API pricing reflected a ~73% inference cost reduction vs V3. SMIC shares surged 9% and Hua Hong Semiconductor 15% on the news. The launch arrived one day after Commerce Secretary Lutnick confirmed zero Nvidia H200 chip purchases by China — together validating China's deliberate strategy of building competitive AI capability entirely within the domestic hardware ecosystem. DeepSeek remains fully open-source; Tencent and Alibaba simultaneously entered talks to invest at a $20B+ valuation. At Auto China 2026 opening the same day, Xpeng unveiled a flying car, IRON humanoid robot, and GX Level 4 robotaxi powered by in-house Turing AI chips — crystallizing China's transition to a fully integrated AI-hardware-mobility platform competitor.