AI-Driven Techno-Economic Risk Assessment for Small Modular Reactor Deployment Under Construction Cost and Schedule Uncertainty

Small modular reactors (SMRs) promise lower construction risk, but no mature commercial series exists from which cost and schedule risk can be measured. This study develops an AI-driven techno-economic risk framework linking correlated cost overruns and delays to the levelised cost of electricity (LCOE) of a 300 MWe plant. A literature-calibrated Monte Carlo generator produced synthetic projects, and gradient-boosting, random-forest and linear models were trained using only features observable at the investment decision. First-of-a-kind (FOAK) projects had a median LCOE of USD 155/MWh against USD 101/MWh for mature-series units, and 92% of FOAK outcomes exceeded a USD 100/MWh reference price. Design maturity, series position and cost of capital dominated the risk. Gradient boosting explained 54% of out-of-sample LCOE variance, only modestly above linear regression, but conditional percentile forecasts cut budget shortfall by 20% against a breach-matched fixed contingency allowance. Combined mitigation lowered the probability of severe-risk outcomes from 56% to 20% without making FOAK projects broadly competitive. Results are simulation-based and indicate relative risk ordering, not empirical forecasts. Ablation and replication checks supported the stability of the results, which point investors, policymakers and researchers towards design maturity, financing cost and better early-project data as priorities.