Carbon markets are uniquely driven by policy interventions-cap adjustments, regulatory announcements, and compliance deadlines-that introduce abrupt shocks propagating through trading activity, liquidity, and price formation in highly nonlinear, event-driven patterns. Conventional time-series and deep learning models, relying on synchronous sampling and fixed time windows, struggle to capture such dynamics. We present an evolutionary neuromorphic deep learning framework for policy shock detection in carbon markets. The architecture integrates spiking neural networks with a deep learning abstraction layer, encoding market microstructure events and policy announcements as spike trains to model temporal accumulation and nonlinear shock diffusion. An evolutionary optimization layer automatically adapts neuronal and synaptic parameters to maximize shock sensitivity under noisy market conditions. Experiments on 18 months of European Union Emissions Trading System, high-frequency data demonstrate that the model achieves 87.8% recall with 19-minute detection latency-63% faster than Long Short-Term Memory and Transformer baselines maintains robust performance across market regimes with 91.7% recall during crisis periods. Cross-market validation shows 86-93% performance retention. In economic simulations, model delivers a 0.79 Sharpe ratio versus 0.67 for Transformers, preventing €1.8M in losses for a €100M portfolio. Results establish neuromorphic Deep Learning as a viable paradigm for modeling event-driven dynamics in policy-driven financial systems.