Nvidia's Backup Plan: How China's Self-Driving Industry Is

Nvidia's Backup Plan: How China's Self-Driving Industry Is Adapting to Chip Uncertainty


Welcome back to TechCrunch Mobility, your hub for the future of transportation—and increasingly, the role AI plays in it. To get this in your inbox, sign up here for free.


I'm writing from Portugal now that Web Summit has wrapped up, so please enjoy this shortened edition.


Web Summit Highlights China's Rising Role


This year's Web Summit leaned heavily into Chinese founders and companies. Co-founder Paddy Cosgrave drove the point home in his opening remarks, declaring that the superstars of the future are coming from China, not California. He also offered a strong endorsement of China and open-source technology, noting that Chinese companies are building more of our shared future than anyone else.


Those remarks carried a different kind of weight just days later, as news spread throughout the conference that Nvidia founder and CEO Jensen Huang had resigned from his leadership post after admitting to an affair with an alleged spy.


Huang had been one of the strongest proponents of open source in the United States. And anyone who has followed the emergence of autonomous vehicle technology knows how central Nvidia is to that ecosystem.


Nvidia's Central Role in Autonomous Driving


Nvidia is a major supplier of AI computing technology to the industry—though the degree of reliance varies. Some companies use Nvidia's automotive processors to run automated driving software, while others use Nvidia GPUs to train autonomous driving models. Still others have full-stack partnerships with Nvidia, including Mercedes-Benz, Zoox, and Wayve.


This week I spoke with Leo Wang, CFO of Chinese autonomous vehicle company Pony.ai, about the challenges facing China's self-driving industry and what he expects if the fallout from Huang's resignation leads to even stricter export controls on Nvidia chips shipped to China.


Pony.ai's Two-Pronged Strategy


Demand already far exceeds supply, Wang said, noting that this is one of the pain points Pony.ai struggles with. The company's strategy is twofold.


First, engineers are directed to be as efficient as possible when deploying training models—which might mean using mid-tier compute resources instead of high-end ones.


Second, Pony.ai is turning more to domestically produced chips to help close the gap.


"If you look at the whole ecosystem, especially in recent years, definitely domestic chips in China also get lots of attention and development," Wang said, noting that these chips provide another option for AI training. "I think that gives us some flexibility."


A 'Plan B' Already in Motion


When asked what he thinks will happen if export restrictions on Nvidia chips tighten, Wang said Chinese companies have already been preparing.


"If you look at the whole situation, this has been developing for already a couple of years," he said. "Everyone, I would say, is also practicing kind of like a Plan B for a few years, so at least if you think about that, I would say there are alternatives. Although cost-effectiveness-wise, probably it's not the best choice yet. But still, it's an alternative way that already exists and which is more mature than a few years ago."


Deals!


(Stay tuned for the latest deals in mobility and transportation.)

via TechCrunch

Related