google.com, pub-8701563775261122, DIRECT, f08c47fec0942fa0
USA

China learns to build without Nvidia

An iFlytek liquid-cooled server equipped with Huawei Kunpeng 920 chips and Ascend AI chips is displayed at the World Artificial Intelligence Conference in Shanghai, China, on July 26, 2025.

Cphoto | Future Publishing | Getty Images

Hello, I’m Evelyn, writing to you from Beijing. Welcome to the latest edition of The China Connection, a concise summary of what I’m seeing and hearing from local businesses.

China’s push for tech self-sufficiency is quickly becoming a reality as companies focus on deeper business questions than geopolitics. What does this mean for Nvidia?

big story

Robovan startup Zelostech plans to use multiple chip suppliers from China and elsewhere over the next year or two rather than relying solely on China Nvidia for self-driving systems, the company told CNBC.

Finance and investment director Shi Yunjian said an important factor was cost. He said using Chinese-made chips, for example, would cost much less than the two Nvidia Orin chipsets currently used in every vehicle.

This is very important because scale becomes a competitive advantage. The more autonomous vehicles can be deployed, the more operational data they can collect and the easier it will be to convince regulators that the technology is ready for broader use.

Zelostech claims to already have More than 25,000 vehicles It operates in more than 20 countries and plans to expand rapidly. These do not carry people and many are smaller than a mail truck. Most operate in mainland China for logistics companies that mostly deliver packages.

In comparison, Alphabet-powered by Waymo Just under 4,000 vehicles Chinese rivals Baidu, WeRide and Pony.ai have yet to deploy fleets of similar scale.

Beyond Nvidia

Zelostech is hardly alone in pursuing Nvidia alternatives.

waymo uses special chipsChinese electric car giant BYD I attended last week Nio And xpeng while revealing own semiconductors for driver assistance systems.

This year, Nio said it plans a fivefold increase in spending on computing power. When I asked if this included Nvidia, CEO William Li said the company no longer buys chips but rents computing power powered by various processors.

An Xpeng vehicle co-developed with Volkswagen also uses the Chinese company’s “Turing chip”; The German automobile manufacturer has established a partnership with China. Horizon Robotics Developing driver assistance systems in China without Nvidia.

Nvidia’s driver assistance chips are not subject to the same U.S. export restrictions that apply to more advanced semiconductors used to train and run artificial intelligence models.

But even after Nvidia CEO Jensen Huang joined US President Donald Trump on a trip to Beijing in May, it’s clear China isn’t keen on letting in more Nvidia chips.

The change extends beyond vehicles. Chinese AI developers are increasingly optimizing their models to run on homegrown hardware rather than Nvidia’s widely used CUDA ecosystem.

With the latest MiniMax and Kimi models, DeepSeek’s V4 is compatible with local Chinese semiconductors.

“We believe the transition to domestic chips will accelerate in 2026E-28E,” Goldman Sachs analysts said in a May 5 report. They pointed out that DeepSeek V4 is powered by eight Chinese-made chips, including Huawei. Alibaba’sT-head chip unit.

A narrowing window

Huawei announced last week that it was using a new scientific approach to develop its chips and that it plans to do so. include them in future products. It’s the latest sign of the Chinese telecommunications giant’s comeback after years of U.S. restrictions.

Kevin Xu, founder of hedge fund Interconnected Capital, expects Chinese companies to continue to need Nvidia chips for the next three to five years.

But he argues that Beijing has an incentive to limit this dependence sooner rather than later. Chinese-made chips can only improve if companies use them in real-world scenarios and generate the feedback needed to make the technology useful for businesses.

“The more Nvidia chips enter the ecosystem, the weaker that relationship becomes,” he said.

Nvidia’s revenue from mainland China and Hong Kong is declining anyway; the company is doubling down on this, planning to spend as much as $150 billion a year in Taiwan.

The investment would likely reverse Taiwan’s initial plan to limit AI data centers and nuclear power, said Chris Cottorone, president of TriOrient Investments and co-chairman of AmCham Taiwan’s alternative assets committee. He also expects more local businesses to adopt AI.

Nvidia’s growing presence in Asia (but not in mainland China) is prompting the chipmaker to find other ways to maintain technological leadership.

The US company is trying to get a foothold in China’s world of “physical AI” by collaborating with Chinese humanoid startup Unitree on a research robot sold globally. reportedly attended A Tsinghua University board said the company and the school have not yet responded to requests for comment.

This is a sign that the tide is turning. China’s technological ambitions are no longer defined by access to Nvidia, but by companies that can produce without it.

I need to know

is coming

31 May – 2 June: Brazilian Foreign Minister Mauro Vieira visiting china

1 – 3 June: UK Foreign Secretary Yvette Cooper visiting china For the 11th bilateral strategic dialogue

June 3: RatingDog China Services PMI (May)

June 9: Trade data for May

Select CNBC as your preferred source on Google and never miss a beat from the most trusted name in business news.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button