We develop a stochastic human immunodeficiency virus type 1 (HIV-1) infection model to analyze combination antiretroviral therapy (cART) dynamics in the brain microenvironment, explicitly accounting for two infected cell states: (1) productively infected and (2) latently infected populations. The model introduces two key epidemiological thresholds $–\overline{\mathcal{R}}_{c1}$ (productive infection) and $\overline{\mathcal{R}}_{c2}$ (latent infection) – and defines the stochastic control reproduction number as $\overline{\mathcal{R}}_c={\rm max} \{\overline{\mathcal{R}}_{c1},\overline{\mathcal{R}}_{c2}\}.$ Our analysis reveals three distinct dynamical regimes: (1) viral extinction $(\overline{\mathcal{R}}_c <1):$ the infection clears exponentially with probability one; (2) latent reservoir dominance $(\overline{\mathcal{R}}_c=\overline{\mathcal{R}}_{c2}>1):$ the system almost surely converges to a purely latent state, characterizing stable viral reservoir formation; (3) persistent productive infection $(\overline{\mathcal{R}}_c =\overline{\mathcal{R}}_{c1} >1):$ the infection persists indefinitely with a unique stationary distribution, for which we derive the exact probability density function. And numerical simulations validate these theoretical predictions, demonstrating how environmental noise critically modulates HIV-1 dynamics in neural reservoirs. Our results quantify the stochastic balance between productive infection, latency establishment, and cART efficacy, offering mechanistic insights into viral persistence in the brain.