<p>This study examines whether US climate policy uncertainty (CPU) affects AI-sector returns beyond the influence of broad economic policy uncertainty (EPU) and macroeconomic income conditions. Using monthly data from 2018M6 to 2025M9, AI performance is proxied by the Indxx artificial intelligence index. At the same time, CPU and EPU are measured using established uncertainty indices, with income-side GDP included as a control. Motivated by evidence of non-normality and nonlinear dependence, the analysis adopts a distribution- and horizon-sensitive econometric framework, combining quantile wavelet Zivot–Andrew’s break-aware stationarity testing with wavelet quantile regression and multivariate wavelet quantile regression; robustness is assessed using kernel-regularised quantile regression with marginal effects. The results show that the uncertainty–AI nexus is state-dependent and horizon-dependent: CPU and EPU effects vary markedly across return quantiles and time horizons, implying that mean-based linear models obscure economically meaningful dynamics. Importantly, the mechanism analysis indicates that CPU retains incremental explanatory power for AI returns even after conditioning on EPU and GDP, confirming that climate-policy uncertainty is not merely a proxy for general policy uncertainty or macro fundamentals. The findings highlight that AI functions as both a macro-sensitive technology asset and a transition-relevant sector, whose response to climate-policy uncertainty depends on market states and time horizons. Novelty stems from isolating climate-policy-specific uncertainty from general policy uncertainty in explaining AI-sector performance, and from documenting these effects within a multiscale, quantile-based framework that captures distributional heterogeneity and time-horizon dependence.</p>