This paper investigates AI-based compression techniques for transmitting high-dimensional signals under strict bandwidth and latency constraints in the fronthaul link of wireless systems. Conventional strategies such as compressed sensing, scalar quantization, and fixed codec pipelines rely on limited prior information, suffer from rapid performance degradation at high compression ratios, and are difficult to tune across channels and deployment environments. In this paper, we investigate AI-based compression techniques and analyze two representative high-compression approaches: CSI feedback via end-to-end learning and resource block (RB)-level precoding optimization and compression combining. Based on these insights, we propose a fronthaul compression strategy tailored to cell-free architectures, aiming for high compression ratios, controllable performance loss, RB-level rate adaptation, and low-latency inference suitable for centralized, cooperative transmission in next-generation networks.