the reactor runs, the budget doesn't

Problem statement

Advanced nuclear reactor designs (small modular reactors, molten salt reactors, high-temperature gas reactors) promise safer, more flexible nuclear power, but their projected fixed operations and maintenance (O&M) costs of ~$13/MWh are 6–10× higher than combined-cycle natural gas plants (~$2/MWh). These costs stem from regulatory-driven staffing requirements, conservative maintenance schedules inherited from the existing light-water reactor fleet, and the absence of real-time condition monitoring that could enable predictive rather than preventive maintenance. Without radical O&M cost reduction, advanced reactors cannot compete economically even if their construction costs are brought to target.

Why this matters

Nuclear power is the largest source of carbon-free electricity in the U.S. (roughly 20% of generation), but the existing fleet is aging and no pathway to replacement is economically viable at current cost structures. Advanced reactors are designed to be simpler and inherently safer, which should in principle enable leaner operations — but the regulatory and operational frameworks haven't evolved to reflect these design improvements. ARPA-E's GEMINA program specifically targets a 10× reduction in O&M costs to make advanced reactors competitive with natural gas on a total cost basis, recognizing that construction cost alone isn't the binding constraint.

What’s been tried and why it hasn’t worked

The existing nuclear fleet operates with large staffs (500–1,000 workers per plant) due to NRC requirements established for 1970s-era designs. Digital instrumentation and control (I&C) upgrades have been slow because nuclear-qualified digital systems require extensive verification and validation, and regulators have been conservative about approving reduced staffing based on automation. Condition-based maintenance has been piloted at a few plants but relies on sensor data from aging instrumentation not designed for the purpose. The fundamental disconnect is that advanced reactor designers promise simpler, safer operations but have no demonstrated operational data to support reduced regulatory requirements — and they can't get operational data without building plants that they can't finance at current cost projections.

What would unlock progress

ARPA-E's GEMINA program invests in digital twin technology for advanced reactors — high-fidelity computational models that simulate reactor behavior in real time, enabling autonomous monitoring, predictive maintenance, and reduced human intervention. Digital twins could provide the evidence base regulators need to approve reduced staffing and simplified maintenance schedules. Advances in embedded sensors (radiation-hardened, self-powered), machine learning for anomaly detection in reactor systems, and formal verification methods for safety-critical autonomous control could collectively enable the paradigm shift from preventive to predictive operations.

Entry points for student teams

A team could develop a simplified digital twin of a reactor subsystem (e.g., a heat exchanger or coolant pump) using physics-based modeling coupled with simulated sensor data, and demonstrate predictive maintenance capability by injecting simulated degradation modes. Nuclear engineering, controls engineering, and data science skills are most relevant.