At the 2026 AIDC Industry Development Conference and 3D Data Center On-Site Meeting, Wang Tao from Huawei remarked that AI large models have stepped into a novel stage, where the parameter scales have soared beyond the trillion-mark threshold. He forecasted that by 2027, models boasting 10 trillion parameters will set the baseline for ultra-large models, propelling AI towards a transformation into intelligent agents. As technological progress marches on, both the open-source ecosystem and industrial scale are witnessing explosive expansion, with China's AI computing power consumption escalating at an exponential rate. This surge is driving a rapid evolution of computing infrastructure, necessitating a restructuring of traditional distributed architectures. Ultra-large-scale computing clusters are steadily emerging as an industry benchmark. It is anticipated that by 2027, super node clusters outfitted with over 100,000 accelerator cards will constitute the fundamental setup for data centers, encompassing full-chain innovation. Presently, the AI industry is characterized by a "dual-wheel drive" phenomenon: on one hand, there are ongoing breakthroughs in foundational model parameters; on the other hand, application scenarios are continuously infiltrating the real economy, reshaping the global digital economy competition landscape.