DeepSeek reverts to Nvidia for R2 model after Huawei AI chip fails

The Deepseek plan to train the new AI, R2, on Huawei Ascend chips and forced to retreat on NVIDIA while delaying the launch.

For several months, the narration that Beijing pushed was from the unimaginable technological progress and a march towards self -sufficiency. However, the reality usually has biting. The recent problems of the Chinese Ai Deepseek are an example of a textbook for the place where ambition meets with the solid wall of artistic restrictions.

After the successful launch of the R1 model in January, Dibsic found itself under pressure from China to the national issue. According to three people talking to the Vinanchel Times times, the message was clear: Use Huawei chips, not NVIDIA.

When it comes to training the new R2 model already, the sources say Deepek faced “continuous technical problems” with AI chips in Huawei. The problems were so essential that the project stops. The person who has knowledge of the situation said that this is the main reason for the planned launch in May, putting the company on the back foot in a market that no person is waiting.

To understand why this is a big problem, you have to know the difference between artificial intelligence training and inference. Training is a difficult part, such as sending a student to the university for years of intense learning. It requires huge amounts of energy and stability. Inference is the relatively “easy” part, such as asking a question on postgraduate studies.

Deepseek discovered that although Huawei chips may be ready for the final exam, they were not yet in the arduous university session. The company had no choice but to return to the powerful NVIDIA systems to do training. The sources say that Deepseek is still trying to get the R2 to work with Huawei chips for the less demanding inference.

Two people confirmed that Huawei has sent its team from engineers to Deepseek offices to help them raise the R2 and operate it on their chips. But even with the presence of experts in the room, they could not get a successful training.

Talk to anyone in the industry, and they will tell you that this is not a big surprise. Even the CEO of Huawei Ren Zhengfei said earlier this year that the United States is “exaggerated in Huawei’s achievements” and that the company is “not great yet”, noting that its best chips are still behind a generation.

However, Beijing is still pushing its technology giants activity for the benefit of local devices, as the Venangal Times times have informed that companies must now justify orders from the H20 chip that is compatible with the export of NVIDIA. It is part of a strategy to build a local hero, but can force companies to make internal choices technically.

Regardless of the issues facing the use of Huawei chips for training, Deepseek Liang Wenfeng was said to have told his team that he was not satisfied with the general progress towards the R2 model. He is said to push them up and create something that can keep the company among the leaders of the artificial intelligence industry.

For all directions from top to bottom and national pride, engineering laws are still applied. Deepsik’s story is a reminder that in the global race of artificial intelligence, there are no shortcuts. China plays the long game, but for the time being, the crown is still firmly at the head of NVIDIA.

See also: Ren Zhengfei: The future of artificial intelligence in China and the long game Huawei

Do you want to learn more about artificial intelligence and large data from industry leaders? Check AI and Big Data Expo, which is held in Amsterdam, California, and London. The comprehensive event was identified with other leading events including the smart automation conference, Blockx, the digital transformation week, and the Cyber Security & Cloud.

Explore the upcoming web events and seminars with which Techforge works here.