WAIC live coverage: The competition for end-to-end computing power breaks out, and Lingsi holds the key chips for the implementation of the next generation of AI terminals

Zhidongxi Author | Cheng Qian Editor | Mo Ying Zhidongzhi reported on July 21 that at the World Artificial Intelligence Conference (WAIC 2026), which just concluded, there was a wave of concentrated explosion of terminal intelligent hardware. Throughout the entire conference, the most intuitive shift in the industry track is clearly visible: the focus of the AI ​​track is shifting from gathering together cloud supercomputing clusters, judging by parameter scale and cloud computing power, to mobile phones, AI PCs, humanoid robots, smart cockpits, whole-house central control and other pan-terminal hardware. The implementation form of AI has also changed accordingly. It is no longer limited to static dialogue windows in web pages and apps, but has begun to enter real devices with the ability to perceive, understand, interact and execute, and is more closely connected to the physical world. Changes in interaction scenarios also determine the underlying logic of future smart hardware: they are not just hardware terminals, but also computing power carriers equipped with local AI capabilities. This also pushes the key watershed of industrial competition to a lower level - whether efficient, stable, and scalable large-model local reasoning can be achieved under the constraints of low power consumption, miniaturization, and low cost; whether a chip platform can be built that can adapt to multiple types of terminal scenarios. Although the terminal forms are ever-changing, the underlying computing power chip is a unified necessity. Therefore, only manufacturers that have the ability to self-develop end-side AI inference chips and can actually introduce the chips into terminal products will have a better chance of winning the entry into this round of end-side AI explosion. Through the wave of device-side AI that broke out in this WAIC, it is clear that the core of the future development of the AI ​​industry is moving from cloud capability demonstration to device-side scenario implementation, and from software experience competition to chip base competition. Device-side AI chips will become the key infrastructure that connects all categories of physical intelligent terminals and supports the implementation of all-area intelligence. This is also the reason why Lens Technology deserves to be re-observed after WAIC 2026. As a company that has been deeply involved in end-side AI chips for a long time, Lens plans to release the Nebula series of end-side large-model AI inference chips at the end of this year, targeting the underlying computing power demand behind this round of terminal intelligent hardware explosion. 1. The overall rise of end-side AI and the underlying computing power