DeepSeek gives China's chipmakers a leg up in the race for inexpensive AI
DeepSeek's synthetic intelligence (AI) models are giving some Chinese chipmakers, like huawei, a better chance to compete against U. S. companies in the AI market. For years, huawei and others have struggled to create chips that can rival Nvidia's high-quality products. While Nvidia focuses on training models using large amounts of data, DeepSeek's models improve efficiency in producing conclusions, which is expected to close the gap between Chinese and U. S. AI processors.
Recently, chipmakers such as Hygon, Tencent-backed EnFlame, Tsingmicro, and Moore Threads have announced that their products will support DeepSeek models, although details are limited. Predictions indicate that DeepSeek's low-cost, open-source features could help boost AI adoption and real-life applications in China, particularly in response to U. S. export restrictions on advanced chips. Prior to DeepSeek's rise, Huawei's Ascend 910B was seen as suitable for less data-intensive tasks, such as inference.
Many Chinese companies, including automakers and telecom firms, have plans to integrate DeepSeek models into their operations. However, despite their cost-competitiveness in inference, Chinese chips primarily serve the domestic market, as Nvidia remains dominant in both AI training and inference. U. S. restrictions prevent the sale of Nvidia's advanced chips in China, but less powerful models are still available for inference tasks. Nvidia’s CUDA platform remains a critical factor in its dominance, while huawei offers its own alternative called Compute Architecture for Neural Networks (CANN), facing challenges in shifting developers away from CUDA.
click and follow Indiaherald WhatsApp channel