Baidu reports Chinese buyers shift to local AI chips amid ongoing supply constraints

Baidu’s earnings call highlights a growing preference among Chinese buyers for local AI chips, driven by persistent supply chain disruptions. The company’s Kunlunxin inferencing chips, which are CUDA-compliant, are positioned as a direct alternative to restricted Nvidia hardware. This suggests that Chinese firms are prioritizing availability and cost over potential performance trade-offs, at least in the short term.

The supply constraints stem from US export controls, which have blocked Nvidia from selling high-end AI chips in China. While the US later relaxed some restrictions, Beijing’s retaliatory measures, including a veto over foreign chip purchases, have kept the market uncertain. Baidu’s remarks indicate that Chinese buyers are no longer waiting for resolution and are instead investing in domestic solutions.

Baidu’s push to spin out Kunlunxin reflects confidence in the long-term viability of local AI chips. The company’s cloud revenue growth, driven by GPU infrastructure, underscores the commercial potential of this shift. However, the performance and scalability of domestic chips remain unproven compared to established alternatives, which could limit adoption in high-demand scenarios.

For engineers, this trend means evaluating local chip options for AI workloads in China, particularly for inferencing tasks. The reliance on domestic hardware may also accelerate innovation in China’s semiconductor ecosystem, but it introduces new dependencies and potential compatibility challenges. Firms operating in China will need to weigh these factors against the risks of continued supply instability.