Print this page
Published in AI

Nvidia top boffin says US export bans boosted Huawei AI dev

by on06 June 2025


Dally says US rules handed China a tech talent pipeline

Nvidia chief scientist Bill Dally [pictured] has suggested that the US export control ban on AI gear to China has helped Huawei and its mates in the Middle Kingdom rather than holding them back.

Speaking to the gathered throngs at the second AI EXPO, hosted by Washington-based think tank the Special Competitive Studies Project, Dally said that the AI race is not just a corporate scrap but a geopolitical one. He noted that the US ban on high-end chips like Nvidia’s H20 series had ironically given China more wiggle room and incentive to poach top AI researchers.

Dally pointed out that China already had about a third of the globe’s top AI minds in 2019. That number’s now nearing half. He said many of these researchers used to write software for Nvidia. Now, thanks to the US rules, they’re working for Huawei instead.

“If it weren't for US regulation, Huawei wouldn't have been forced to grow,” Dally said.

Despite Nvidia still being ahead in raw AI firepower, Dally reckons Huawei has cobbled together some decent solutions. With US competitors locked out of China, the local tech scene is booming in a vacuum.

The Taiwan Economic Daily noted that the Chinese government is throwing its full weight behind AI, setting up institutes like the Beijing Academy of Artificial Intelligence and the Beijing Institute of General Intelligence. Meanwhile, the US has yet to come up with any serious government-level research effort on the same scale.

Dally said that future leaps in AI will need advances in  hardware and software. Nvidia has made AI training far more energy-efficient in the past decade, mostly through clever hardware tweaks rather than just smaller chips, he said. That trend is set to continue.

As for the biggest problem with AI according to Dally, it’s fake media content. Rather than relying on increasingly complex software to tell what’s real, he thinks the smarter move is to label original footage properly to make sure people can trust what they see.

Last modified on 06 June 2025
Rate this item
(0 votes)