Chinese AI firm DeepSeek is building its own AI inference chip to lower reliance on American-made semiconductors. Reuters reported the move, noting it aims to reduce dependence on foreign suppliers such as US company Nvidia, which has faced difficulties reestablishing sales channels in China after US President Donald Trump began allowing certain chip exports last year.
Why an in-house chip matters
Bringing chip design in-house — a strategy similar to that used by OpenAI — would give DeepSeek tighter control over its technology stack, enabling closer hardware–software integration, potential cost reductions, and greater independence from external vendors. The Chinese government has been urging domestic companies to address shortages of access to powerful chips, a limitation that has hindered China’s development of more advanced AI models.
Market timing and performance
Reuters also highlights that DeepSeek’s pricing has made the service more attractive to consumers and businesses. According to OpenRouter’s data, the share of tokens consumed by DeepSeek models on the OpenRouter platform has doubled over the past six months — a surge that followed DeepSeek’s release of its V4 model in April. Since January, nearly all Chinese model providers on the platform have gained market share, while all US companies have lost share.
Implications
Developing an inference chip could help DeepSeek reduce dependency on foreign suppliers, increase operational control, and improve cost efficiency. The effort aligns with broader Chinese state and industry initiatives to strengthen the domestic semiconductor supply chain, particularly for components required by high-performance AI applications.
(Source: Reuters — article by Rachyl Jones.)
