At the opening ceremony of Baidu World 2026, Robin Li stated that demand for AI computing power has now shaped a landscape where training and inference are equally important. The large-scale adoption of agent applications has driven an explosive surge in inference computing demand, positioning inference as the primary driver of long-term growth for AI chips. Kunlunxin unveiled its product roadmap: the M100 has already launched on schedule, focusing on ultimate cost-effectiveness; the flagship M300 will launch next year, delivering all-around breakthroughs across multiple dimensions including computing power, storage, and interconnect; and in 2028, multiple chips based on the all-new N architecture will be brought to market, precisely tailored to differentiated application scenarios. Robin Li stated that Baidu has long remained committed to technological investment, and Kunlunxin represents the culmination of more than a decade of such investment, with its commercial value gaining increasingly broad recognition.