The Australian chief executive of a major DeepSeek rival has questioned the Chinese artificial intelligence start-up’s claims that it developed its Silicon Valley-beating model for less than $10 million amid increasing scepticism from investors that its arrival will hurt tech sector earnings.

DeepSeek claimed it could process AI queries with the same sophistication but vastly fewer resources – expensive chips manufactured by companies like Nvidia – than its big rivals such as OpenAI, Alphabet and Anthropic. In its supporting documentation, the company said it had spent $US5.6 million ($9 million) to train its model, which is known as R-1.