🍊 Has opened the race. CANN go CUDA.
Nvidia blames the USA:
A loss on the scale of Boeing: Nvidia boss Jensen Huang warns in no uncertain terms about the loss of the Chinese market.
Things are actually going brilliantly for Nvidia: its CEO Jensen Huang has given himself a hefty pay rise for the first time in 10 years for the 2025 financial year.
But now he is warning of the consequences that the enormous loss of the Chinese market would mean not only for Nvidia, but also for the dominance of the USA in the technology market.
A looming loss on the scale of Boeing, not the airplane, but the company
According to CompaniesMarketCap, Nvidia is the third most valuable company in the world with a current market capitalization of 2.863 trillion. In the 2025 financial year, which ended on January 26, 2025, sales in China accounted for around 13 percent, or around 17 billion US dollars, according to Reuters.
This is despite the fact that Nvidia's high-quality GPUs in particular, such as the A100 and H100, are already subject to export restrictions to China. This is one of the reasons why the release of the Chinese chatbot DeepSeek caused such a stir.
The trade dispute between the USA and China could therefore mean billions in losses for Nvidia. Huang himself put the expected loss of the Chinese sales market at 50 billion dollars, according to MarketWatch at the Milken Institute Global Conference:
50 billion dollars is like Boeing - not the airplane - the whole company.
According to Huang, the gap that Nvidia would leave in the Chinese market would quickly be filled by a competitor: Huawei.
CANN from Huawei gives CUDA from Nvidia serious competition
In an interview with CNBC at the end of April, Huang had words of praise for the Chinese competitor. Huawei is one of the most impressive tech companies in the world.
China is hot on the heels of the USA in the long, never-ending race to become the global leader in the development of artificial intelligence.
Jensen himself emphasizes in an interview with Bg2 Pod that Nvidia's software platform and programming architecture CUDA (Compute Unified Device Architecture) is a decisive success factor for the company.
CUDA is deeply embedded in almost all leading deep learning platforms such as PyTorch or TensorFlow - anyone who trains AI models on GPUs there usually automatically uses Nvidia's CUDA libraries.
However, an independent alternative is now emerging in China: Huawei's CANN (Compute Architecture for Neural Networks) is the software ecosystem behind the Group's Ascend chips and is set to play the same role as CUDA in the West in the medium term.
However, CANN is still being criticized by developers, according to the China Academy: CANN does not yet come close to CUDA in terms of performance and user-friendliness. However, Huawei is working on improving the system.
The Chinese competition therefore seems to be well prepared to fill the gap that Nvidia is potentially leaving in China and perhaps secure a certain advantage in the neck-and-neck race with the USA.
https://www.gamestar.de/artikel/nvidia-huawei-jensen-huang-china-usa,3432614.html
