Chinese startup DeepSeek is reportedly developing its own AI inference chip, marking a significant strategic shift from its previous focus on AI model breakthroughs.

The chip, designed for inference tasks rather than training new models, aims to reduce reliance on Nvidia and Huawei chips, which have been critical for training DeepSeek's globally popular models.This move comes amid U.S.export controls restricting Chinese access to advanced semiconductors, prompting Beijing to push for domestic alternatives.

While Huawei has captured a substantial share of China's AI chip market through its Ascend series, competition from Alibaba and Baidu is intensifying.DeepSeek's efforts include partnerships with chip-design and foundry companies, as well as increased hiring of engineers.

The company's in-house chip development aligns with global trends, following examples like OpenAI's Jalapeno and Anthropic's potential custom chips.However, challenges remain, including manufacturing hurdles due to U.S.restrictions on advanced foundries and high-bandwidth memory access.DeepSeek's recent $7 billion funding round marks its first embrace of external capital, signaling a new phase in its growth strategy.