Video with Chinese subtitles: China Doesn’t Need Nvidia Anymore — DeepSeek V4 Just Proved It. 視頻有中文字幕: 中國不再需要輝達 – DeepSeek V4 剛剛證明了這一點.
https://rumble.com/v78ymrm-china-doesnt-need-nvidia-anymore-deepseek-v4-just-proved-it.html
https://www.tiktok.com/t/ZP8ggTUj7/
China just trained an AI model on Huawei chips — and it rivals OpenAI’s best. The US chip ban was supposed to stop this. Instead, it accelerated it.DeepSeek V4 was built on Huawei Ascend 950 AI chips, using Mixture of Experts architecture and inference-first engineering to do what US export controls said was impossible. This isn’t a chip story. It’s a sovereignty story — and the implications go far beyond Silicon Valley.In this video, we break down:
— How DeepSeek V4 was trained without a single Nvidia GPU
— Why the US chip ban may have handed China its biggest AI advantage yet
— The inference vs training distinction that every analyst is missing
— Why Nigeria, Indonesia, and Brazil are now building on Chinese AI infrastructure
— What this means for the future of global AI powerThe United States still has the best chips. China now has something more dangerous. An AI economy that no longer needs them.📌 Sources: Fortune, Stanford AI Index 2026, Counterpoint Research, Al Jazeera, Reuters etc.
中國剛用華為晶片訓練出一個AI模型，其效能足以媲美OpenAI的最強模型。美國的晶片禁令本來是要阻止這一切發生，結果反而加速了它的進程。
DeepSeek V4 是基於華為昇騰950 AI晶片打造，採用「混合專家」（Mixture of Experts）架構與推論優先的工程設計，實現了美國出口管制原本認為不可能的事。這不只是晶片的故事，更是一場關於技術主權的較量——而它的影響遠遠超出了矽谷的範疇。
在本影片中，我們將深入分析：
· DeepSeek V4 如何在完全不用輝達GPU的情況下完成訓練
· 為何美國的晶片禁令可能反而給了中國最大的AI優勢
· 每位分析師都忽略的「推論 vs. 訓練」關鍵差異
· 為何奈及利亞、印尼、巴西現在開始轉向中國的AI基礎設施
· 這對全球AI實力版圖的未來意味著什麼
美國仍擁有最好的晶片。但中國現在掌握的東西更為危險 – 一個不再依賴它們的AI經濟體系。
📌 來源：Fortune、史丹佛AI指數2026、Counterpoint Research、半島電視台、路透社等
Johnson Choi Biz 