Integrated sensing and communication (ISAC) in dense multi-cell networks suffers from severe inter-cell interference, degrading both sensing and communication performance. Conventional strategies focus on interference suppression. To address this issue, this paper proposed an interference empowering precoding (IEP) framework that exploits informative inter-cell interference as a distributed sensing resource while protecting communication quality-of-service (QoS) in ISAC. IEP operates on two timescales. In the slow timescale, a lightweight Transformer-based module infers the symbol-level knowledge coefficients of interfering waveforms while per-resource-element adaptation between constructive interference exploitation for sensing and robust suppression for communication is conducted in the fast timescale. A joint optimization problem was formulated to maximize the weighted sum of the communication sum rate and the log-determinant of Fisher Information Matrix with QoS and constructive interference alignment constraints under channel state information uncertainty. The non-convex problem is reformulated as a difference-of-convex program and solved using an alternating Weighted Minimum Mean Square Error and Convex-Concave Procedure algorithm with guaranteed convergence. Simulation results demonstrate that the proposed framework could sub-stantially enlarge the achievable communication-sensing trade-off region. Notably, it achieves approximately 24% reduction in the Position Error Bound and a 21.5% extension in detection range compared to conventional interference suppression schemes, without compromising communication throughput.