iFlytek Chairman: Domestic Computing Power Trails Global Standards by Two Generations
Speaking at the 2026 Global 1024 Developer Festival, iFlytek Chairman Liu Qingfeng stated that China's domestic computing power currently lags about two generations behind global standards, but expressed confidence that it will eventually catch up and lead. This assessment from a major industry leader provides a realistic evaluation of China's domestic AI infrastructure gap under international export controls. It highlights the strategic urgency for domestic companies to build autonomous and controllable AI platforms for large language model training. Liu noted that very few domestic companies are currently dedicating serious efforts to large-scale model training within the domestic ecosystem. Furthermore, he raised philosophical concerns regarding the long-term convergence and challenges between carbon-based and silicon-based life as AI advances.
## BACKGROUND
Domestic computing power refers to China's self-reliant computing capacity framework, encompassing everything from chip design and server manufacturing to operating systems and software ecosystems. Large language model training is a resource-intensive compute process that relies heavily on advanced AI hardware accelerators and robust software frameworks.