Baidu reports that Chinese buyers are accelerating purchases of domestic AI chips amid persistent supply chain constraints.
Chinese web giant Baidu told investors it expects strong prospects for its Kunlunxin chip division, citing a lack of viable alternatives for domestic buyers. The company has previously announced plans to spin out and publicly list Kunlunxin, which produces CUDA-compliant inferencing chips. These processors power Baidu’s own cloud services and have been supplied to Huawei and ZTE, who integrate them into equipment sold to Chinese telecommunications carriers.
Speaking during the company’s second-quarter earnings call, Dou Shen, executive vice president of Baidu’s AI Cloud Group, confirmed that Baidu is actively pursuing a listing for Kunlunxin and expects to provide concrete details shortly. “From a business perspective, we remain very confident in Kunlunxin’s long-term growth and commercial potential for a few reasons,” he said. First, demand for AI inferencing continues to rise, a trend Baidu expects to sustain over the long term. Second, China’s domestic market holds “significant growth potential” because the supply of AI chips is “likely to remain constrained for some time.” He added, “Against this backdrop, customers are increasingly seeking high performance, reliable, and cost-efficient domestic AI chips.”
These remarks come amid shifting U.S.-China semiconductor policies. While the U.S. government has permitted Nvidia to resume certain product sales to China, Beijing has asserted a veto over such purchases by domestic companies. Following earlier U.S. export restrictions that barred Nvidia from selling in China, the company reported a $10.5 billion revenue loss over six months. In its most recent financial disclosure, Nvidia stated it had generated zero revenue from China following the policy reversal and remains “uncertain whether any imports will be allowed into the country.” Consequently, Baidu notes that Chinese buyers are turning to domestic alternatives due to ongoing supply constraints.
Nvidia CEO Jensen Huang has urged the Trump administration to facilitate chip sales to China, arguing that doing so would cement American dominance in artificial intelligence. Conversely, Beijing has promoted the adoption of homegrown technology to reduce reliance on U.S. products. Baidu maintains that controlling its entire stack—spanning foundational models, cloud infrastructure, and custom silicon—will allow it to deliver AI services at competitive prices and secure a market edge. Competitor Alibaba advances a similar strategy, though it is widely considered to hold a lead in large language model development.
Baidu’s second-quarter results highlighted robust expansion within its AI segment. Revenue from cloud infrastructure rentals increased 50 percent year-over-year to nearly $1.1 billion, while GPU cloud revenue surged 283 percent year-over-year, accelerating from the 184 percent growth recorded in the previous quarter. Despite these gains, Baidu’s scale remains modest compared to global hyperscalers like AWS, Google, and Microsoft, and it trails several domestic rivals. Nevertheless, the company’s AI cloud division serves as its primary growth driver, especially as overall corporate revenue grew just 4 percent year-over-year to $3.9 billion.
Beyond cloud computing, Baidu continues to dominate the autonomous ride-hailing sector with its Apollo Go service. Executives reported that the platform completed more than one million fully autonomous trips globally during the second quarter. On the consumer side, Baidu executives highlighted improved accuracy across its AI offerings, noting low hallucination rates in consumer-facing applications. Daily active users for the ERNIE assistant rose 83 percent year-over-year, while the volume of daily user conversations more than tripled.