DeepSeaK Advances AI Chip Design Amid US Export Restrictions

Chinese firm DeepSeaK has initiated efforts to design its own artificial intelligence chip, marking a significant shift from its reputation as a low-cost AI model provider to a contender aiming to control the hardware powering these models. This development coincides with increasing US restrictions on China's access to the latest Nvidia chips.

According to a Reuters report, DeepSeaK is focusing on creating a chip tailored for the inference phase of AI, the stage where trained models generate responses for users, rather than for training models from the ground up. Sources familiar with the matter told Reuters that the company has expanded its chip design engineering team in recent months. The project reportedly began about a year ago and remains in early stages, involving discussions with design firms, foundries, and memory suppliers.

The potential chip is expected to lessen DeepSeaK's dependence on Nvidia and Huawei components. US export controls have prompted many Chinese companies to seek domestic alternatives. While Huawei has benefited from the absence of advanced Nvidia chips in China, it faces competition from companies like Alibaba and Baidu, which are also developing proprietary chips. DeepSeaK positions itself within a dual competition: the race to develop AI models and the race to create the hardware that determines the cost of operating AI at scale.