DeepSeek sidelines Nvidia, AMD in latest AI model rollout, Huawei gets early access: Report

In a move that underscores the deepening technological divide between the US and China, DeepSeek, the Chinese artificial intelligence lab whose low-cost model jolted global markets last year, has withheld early access to its upcoming flagship model from leading US chipmakers, Reuters reported on Thursday, citing sources familiar with the matter.

Breaking from industry practice ahead of a major release, DeepSeek did not share pre-release versions of its next model, V4, with Nvidia and Advanced Micro Devices (AMD) for performance optimisation. Instead, it granted early access to domestic suppliers, including Huawei Technologies, giving Chinese chipmakers a head start of several weeks to tune the software for their processors, the report said.

DeepSeek is expected to unveil V4 around the Lunar New Year holiday.

AI developers typically provide leading hardware makers with advance access to major model updates so that the software runs efficiently on widely used graphics processing units (GPUs) and accelerators. DeepSeek has previously worked closely with Nvidia’s technical staff.

More from Business

Nvidia beats estimates, but cash return question caps market reaction Sensex rises 215 pts to 82,490; Nifty tops 25,538 as IT heavyweights power early gains

Strategic shift amid export tensions

The move comes amid intensifying scrutiny in Washington over China’s access to advanced AI chips. A senior official in the administration of US President Donald Trump told Reuters that DeepSeek’s latest AI model was trained on Nvidia’s most advanced Blackwell chips using a cluster in mainland China — a step that could potentially violate US export controls.

According to the US official, DeepSeek may seek to remove technical indicators that reveal its use of

American AI chips and could publicly claim that it relied on Huawei processors to train its model.

If confirmed, such a strategy would reflect Beijing’s broader push to reduce dependence on US semiconductor technology while bolstering domestic champions such as Huawei.

ALSO READ: Nvidia beats estimates, but cash return question caps market reaction

Ben Bajarin, chief executive of research firm Creative Strategies, told Reuters the direct financial impact on Nvidia and AMD was likely to be limited.

Quick Reads

View All

Nvidia beats estimates, but cash return question caps market reaction

Sensex rises 215 pts to 82,490; Nifty tops 25,538 as IT heavyweights power early gains

“The impact to Nvidia and AMD for general data accelerators is minimal — most enterprises are not running DeepSeek, which serves as a benchmarking model more than anything else,” Bajarin said. He added that advances in AI coding tools have compressed the time needed to optimise software for hardware “from months to weeks”.

However, he suggested the decision fits into a broader Chinese government effort “to try to keep US hardware and models disadvantaged” within China.

China’s open-source surge

DeepSeek burst onto the global AI scene in January 2025 with a low-cost model that intensified competition with US labs and triggered volatility in technology stocks. Its models have since been downloaded more than 75 million times on Hugging Face, the open-source platform widely used by developers.

Over the past year, downloads of Chinese AI models on the platform have surpassed those from any other country, highlighting the rapid ascent of China’s open-source ecosystem.