Baidu Asserts Chinese Consumers Prefer Domestic Chips, According to GadgetLad

Kunlunxin: The Next Major Player or Just Another Chip?

The Chinese internet behemoth Baidu believes it has struck a rich vein with its Kunlunxin chip venture, primarily due to the notion that local customers have limited choices. Baidu intends to spin off Kunlunxin, the smart team behind CUDA-compliant inferencing chips utilized in its cloud services and sold to clients like Huawei and ZTE. These companies then integrate them into the equipment they market to Chinese telecom operators. During Baidu’s Q2 earnings discussion, Dou Shen, the executive vice president of their AI Cloud Group, suggested that Kunlunxin has a promising future listing, with more information to follow soon.

Increasing Demand and Domestic Edge

Shen is enthusiastic about Kunlunxin’s expansion and commercial prospects, pointing to the persistent demand for inferencing as a significant catalyst. He also believes the domestic market in China is primed for growth, as the supply of AI chips may be constrained for some time. Against this backdrop, clients are actively seeking high-performance, dependable, and cost-effective domestic AI chips.

The Nvidia Narrative: A Touch of Intrigue

These comments are particularly noteworthy considering the US government’s choice to allow Nvidia to resume sales of its products in China, alongside Beijing’s recent decision to control purchases by local enterprises. Nvidia is expressing concerns, asserting that the restrictions led to a loss of around $10.5 billion in just half a year. Even with the policy changes, Nvidia has not recorded any recent revenue from China and is uncertain about when it can re-enter the market.

China’s Push for Indigenous Tech

Nvidia’s CEO, Jensen Huang, believes the Trump administration should be pleased about exporting chips to China to solidify America’s leadership in AI. However, Beijing has been urging the adoption of local technologies, partially to reduce dependence on US products. Baidu’s earnings reflect strong growth in its AI operations, with cloud infrastructure rental income climbing 50 percent year over year to almost $1.1 billion, and the company’s GPU cloud revenue soaring 283 percent. While these numbers are modest compared to giants like AWS, Google, and Microsoft, they are impressive among Chinese competitors.

Baidu Against Its Rivals

Baidu is wagering that controlling its own models, infrastructure, and chips will enable it to offer AI services at competitive rates and provide it with a competitive advantage in the market. Alibaba is making similar declarations, though it arguably outstrips Baidu in model development. Nevertheless, Baidu owns the autonomous taxi landscape with its Apollo Go service, which has achieved over a million fully autonomous rides globally in Q2. On the web side, executives highlighted low hallucination rates in their consumer-AI services and an 83 percent year-over-year increase in daily active users of Baidu’s ERNIE assistant, with daily interactions more than tripling. While overall revenue only grew four percent year-over-year to $3.9 billion, Baidu’s AI cloud is undoubtedly the driving force behind this growth.

Chip Off the Old Block

Baidu is riding the wave of local chip enthusiasm, asserting that the absence of alternatives will keep Kunlunxin in high demand. With Beijing supporting homegrown technologies and Nvidia pondering its future in the Chinese market, the chip landscape remains as dynamic as ever. Stay tuned to see whether Baidu’s audacious strategies yield results or if they are merely full of hot air.