On June 17, reports surfaced revealing significant advancements in ByteDance's data center expansion plans. Industry sources disclosed that ByteDance is engaged in negotiations with Iluvatar CoreX to procure a minimum of 50,000 AI chips, predominantly earmarked for inference operations. These chips are set to be integrated with Iluvatar CoreX's ZhiKai series cloud inference GPUs, while training tasks will leverage the TianGai series. Should the agreement materialize, Iluvatar CoreX will join Huawei and Cambricon as ByteDance's third GPU supplier. At present, both ByteDance and Iluvatar CoreX have refrained from commenting on the matter. The intensified efforts by internet behemoths like ByteDance to bolster their in-house data centers transcend mere short-term computing power augmentation. They signify a profound shift in the structural landscape of AI computing power demand, a strategic elevation towards autonomous and controllable computing capabilities, and the concurrent evolution of hardware deployment strategies that separate training and inference functions, alongside a sweeping transition to domestic computing power solutions.