Background: Molecular HIV surveillance (MHS) is routinely used by public health departments to monitor HIV transmission dynamics and identify rapidly growing transmission clusters (groups of one or more genetically linked individuals) for targeted prevention and response. Current approaches to assessing cluster growth typically rely on fixed-window metrics, such as counts of new diagnoses over 12 months, and do not capture the continuous evolution of clusters or time-varying clinical and demographic factors. These limitations reduce the ability of surveillance systems to prioritize cluster response on the basis of key epidemic drivers, such as incident infection or viral load, that may be associated with future cluster growth.Methods: We analyzed de-identified HIV surveillance data, including partial pol sequences from routine drug-resistance genotyping, from San Diego County (SDC) (2017–2020; N = 1,241 cases forming 1,118 clusters, including singletons, by the end of 2020) and Miami-Dade County (MDC) (2017–2022; N = 3,448 cases forming 2,972 clusters by the end of 2022). Molecular clusters were defined using a ≤0.5% pairwise genetic-distance threshold. Incident cluster growth was modeled with an Andersen-Gill recurrent event model incorporating time-varying individual-level and cluster-level covariates; model estimates were used for counterfactual prediction.Findings: Most clusters remained singletons (single-individual clusters with no subsequent genetically linked cases) throughout follow-up (93.5% in SD; 92.9% in MD). Among growing clusters, only a subset showed incident or mixed growth (2.2% of clusters in SD; 2.9% in MD) with at least one incident infection during follow-up, indicating ongoing transmission during the surveillance period. In both jurisdictions, a higher proportion of incident infections and greater cumulative log viral load burden were associated with faster incident cluster growth. In SD, a 10-percentage-point increase in the proportion of incident infections was associated with a 53% higher incident cluster growth rate (hazard ratio [HR], 1.53; 95% CI, 1.43–1.62); in MD, the corresponding increase was 30% (HR, 1.30; 95% CI, 1.26–1.34). Counterfactual analyses showed that sustained reductions in incident infections and cumulative viral load substantially attenuated predicted incident cluster growth. Reducing the proportion of incident infections from 10% to 1% corresponded to approximately 61 fewer linked cases in SD over four years and 103 fewer in MD over the six years following identification of the incident case. A scenario that reduced cumulative log viral load burden from 3.9 to 1.3 corresponded to approximately 96 fewer cases in SD and 521 fewer in MD over the same 4-year and 6-year period, respectively. By contrast, counterfactual reductions in bacterial sexually transmitted infection prevalence produced comparatively modest effects.Interpretation: Using routinely collected HIV surveillance data, a recurrent event framework quantified how incident infection and viral load shape the pace of HIV cluster growth over time and enabled counterfactual evaluation of plausible intervention scenarios. This approach may strengthen the ability of health departments to prioritize timely interventions to reduce onward transmission.