The dependence of China from Nvidia: difficulties and perspectives in the AI sector | Hardware computer | Hardware catalog | Gpu vs cpu reddit | Turtles AI
AI’s Chinese companies must continue to rely on Nvidia for its calculation solutions, since national alternatives are not yet able to guarantee the same performance and competitive costs. The situation is complicated by American restrictions and the growing demand for calculation power.
Key points:
- CAICT invites Chinese companies to use Nvidia solutions for AI.
- National alternatives, such as Huawei chips, show significant limits.
- China saw 70% growth in the calculation power demand.
- US restrictions further complicate the acquisition of hardware for AI.
The Chinese CAICT research institute has recently underlined the importance for AI companies in the country to continue using the solutions offered by Nvidia, highlighting how local alternatives, in particular those of Huawei, do not represent a valid replacement. The difficulties related to the availability and high costs of implementing national products have been at the center of the analysis. In a context in which China tries to enhance its calculation capacity to support its AI industries, the growing pressure of US sanctions makes it difficult to dependence exclusively on national solutions. The calculation power demand in China has grown considerably, with an increase of 70% in the use of GPU for training and inference AI, which highlights an expanding and hungry market of resources. Companies such as Bytedance and Tencent, which have a vast base of consumers, require significant computational support, making access to NVIDIA chips essential. Despite the initiatives of Huawei and other local producers such as Birentech, the current skills of these solutions remain limited. The Huawei Ascend 910b GPU has achieved some success, with items that anticipate the arrival of more advanced models such as the Ascend 910C. However, the porting of existing applications developed on Nvidia platforms towards national infrastructure remains a complex and often unconvenient challenge. In this scenario, despite the uncertainties related to American policies, China must look for alternative paths. Although there are options such as the GPU rental or the acquisition of hardware through unofficial channels, these solutions are not sustainable in the long run.
The question of access to AI adequate resources is fundamental for the future of Chinese companies, the development of which largely depends on the reliability of current technologies.