China's AI model, DeepSeek AI, is facing difficulties due to the shortage of advanced American chips. The company initially claimed to use Huawei's Ascend chips, but the model is proving to be largely unsuccessful on a large scale. Now, reliance on Nvidia chips has further increased the project's challenges.
DeepSeek AI: China's ambitious AI project, DeepSeek, is now in trouble. A few months ago, at the time of its launch, it was considered an alternative to major AI models like ChatGPT and Google Gemini. Silicon Valley giants were also shocked by its performance. But today, the situation has changed. This model, which claimed to be based on indigenous technology, is now dependent on American hardware, and the chip shortage has exposed its real weakness.
Limitations of Huawei's Ascend Chips
DeepSeek used Huawei's Ascend chips to train its R2 model. The company claimed that these chips were sufficient for high-level AI training. But now, engineers have admitted that these chips are not reliable for training large-scale models. The Ascend chips have low stability, slow network speed, and a lack of software tools, which are affecting the project.
Because of this, DeepSeek has had to rely on American Nvidia chips to make its model more advanced. Although Huawei's chips are still being used for the interface, real heavy training is now impossible without Nvidia's advanced chips.
Impact of US Restrictions and Trade War
The ongoing trade war between the US and China has further deepened this challenge. The US has made it clear that it will not sell its most advanced AI chips, such as Nvidia A100 and H100, to China. These are the same chips on which the world's largest AI models are trained today.
Due to this restriction, DeepSeek is facing considerable difficulty in advancing its project. The company is unable to take full advantage of the CUDA software ecosystem, and the model's performance is continuously declining.
Questions being raised on indigenous technology
China's DeepSeek AI was poised to create a stir in the AI world at the time of its launch, but now this project has proven to be lagging behind due to the hardware shortage. Huawei's Ascend chips are insufficient for large-scale model training. The US ban on advanced Nvidia chips has further complicated the situation. As a result, DeepSeek's development speed has slowed down, and China's reliance on indigenous technology has come under question.
Impact of AI on a Global Scale
The delay in DeepSeek's R2 project has weakened China's position in the global AI competition. China is still dependent on advanced hardware compared to the leading AI models in the US and Europe. In this situation, China needs to further strengthen its technology and become self-reliant in hardware manufacturing.
Challenges and Opportunities in the Coming Times
Even though DeepSeek's project is currently stalled, China is getting important lessons from this experience. Huawei and other Chinese companies now need to improve the performance, network capabilities, and software support of their chips.
DeepSeek's story also demonstrates that algorithms alone are not sufficient for the successful development of an AI model. Reliable and high-performing hardware is essential.
