The artificial intelligence behind ChatGPT-like products and autonomous driving is driving enormous demand for Nvidia’s chips in China. In the past week, however, analysts cut their Nvidia price targets after news the U.S. plans to ban the sale of more high-end semiconductors to China. The country accounts for at least one-fifth of Nvidia’s big data center business . Chinese companies also dominate the burgeoning electric car market , where Nvidia has had a fast-growing business of selling chips for assisted and fully autonomous driving. When it comes to such chips for cars, Nomura analysts said there’s little reason to worry. Their analysis of U.S. rules found regulators are focused on two technical specifications — which the Nvidia Drive AGX Orin chip meets in part. But the chip still isn’t powerful enough to cross a key performance metric, and it isn’t used in data centers, the report said. “Therefore, Orin X chips remain safe and are not impacted by the new regulation, indicating the rest of the existing auto chips should not be impacted by the new rule as well,” Nomura China Technology analyst Joel Ying and a team wrote Thursday. Auto chip market BYD, Nio , Li Auto and Xpeng are among the China-based electric automakers using the Orin chip. Xpeng, which currently offers the closest equivalent to Tesla’s Full Self-Driving in China, is set to hold a tech day on Oct. 24. Meanwhile, other companies are launching alternatives to Nvidia’s products for AI computing. U.S.-based Kneron is using a different approach to AI chips that the company’s CEO Albert Liu claims is based on neuroscience — instead of the graphics processing that Nvidia uses. He told me at CNBC’s East Tech West conference last week that Kneron’s revenue for the fourth quarter is forecast to grow by double-digit percentage points from the third quarter, and “multiple times” from the fourth quarter last year. The company is working with Apple supplier Foxconn for the development of automotive AI , according to an announcement in late September. Weeks earlier, Kneron had unveiled its KL730 chip and claimed it is 150% to 200% times more energy efficient than peers. Its applications include advanced driving assistance systems. Other customers include Quanta Cloud Technology, South Korean search giant Naver, and Japanese and German auto giants, Liu said. He didn’t name the automakers, but claimed that overall Kneron is shipping “millions of chips” annually. “Everyone noticed that GPU is not the perfect solution for AI so it’s more easy to convince people to use our solution,” he said. Kneron raised $49 million in late September. Simply testing AI models also requires significant processing power, which can get expensive to operate amid a shortage of chips. “Charging by GPU per hour is a common global industry practice with global cloud players charging USD1-3 per GPU hour for use of NVIDIA’s A100 80G chip,” HSBC analysts said in an Oct. 17 report. Nvidia said in an SEC filing the new U.S. restrictions would affect sales of its A100 chips and many other products to China, but did not mention Orin. The new U.S. rules are set to take effect in about a month. Homegrown chip companies While Nvidia may get a pass on automotive chips in China for now, the new measures do indicate more advanced ones in the future may require a license from the U.S. government if they are shipped to China, the Nomura analysts said. In the automotive chip category, they noted Nvidia’s Thor chip and Qualcomm’s Snapdragon Ride Flex chip both fall into that more advanced category. Chinese companies have meanwhile been building homegrown alternatives. Autonomous truck driving company Inceptio CEO Julian Ma told me in August the company is using a chip from Chinese startup Horizon Robotics. Ma said Inceptio has enough computing power to support it for the next three years. The startup currently sells trucks with assisted-driving software to logistics companies in China. Other kinds of automation in China today that stock analysts are watching do not even need such advanced computing power. In the past week, Nomura and HSBC analysts both raised their price targets for mainland China-traded Inovance, which HSBC describes as “the largest domestic factory automation solution supplier in China in terms of 2022 revenue.” HSBC has a price target of 83 yuan, up from 76 yuan previously. That marks upside of more than 30% from Inovance’s close on Thursday. Nomura, which like HSBC has a buy rating on Inovance, raised its price target to 76 yuan, up from 74 yuan previously. “Management attributed the healthy revenue growth to the solid growth of its [new energy vehicle] and automation businesses, offset by demand weakness in its elevator business during the quarter,” Nomura analysts wrote in an Oct. 17 report. The analysts noted Inovance has grown its market share this year for motor controllers and powertrain systems in China. “We believe Inovance’s market share gain in the domestic [new energy market was mainly fuelled by wallet share expansion in key customers such as GAC,” the Nomura report said, noting the firm also has a buy rating on the Hong Kong-listed automaker.
The U.S. is trying to tighten the screws on Chinese AI. What that means for stocks
