Abstract
Abstract
This study develops a hybrid empirical-scenario framework for long-horizon small modular reactor (SMR) infrastructure finance. The empirical fixed-income layer uses 249 official U.S. Treasury daily par-yield observations from 2025, producing 248 consecutive realized curve changes. The tested instrument is the project's own 30-year, unit-principal, level-payment amortizing debt rather than a par-coupon bullet bond. Under linear interpolation, KRD has mean absolute one-day repricing error of 0.0024 percentage points, versus 0.1653 for effective duration driven by the average node change, 0.0646 when duration is driven by the 30-year node, and 0.1680 for a first principal-component factor-duration comparator. KRD wins 99.6%, 98.4%, and 99.6% of the respective comparisons. The result remains robust under natural cubic spline, PCHIP, and fixed-decay Nelson-Siegel construction, and circular moving-block bootstrap inference yields a finite-sample corrected one-sided p-value of approximately 0.00005 for every comparator and block length tested.
The lifecycle layer is calibrated to the published empirical record rather than to project-specific estimates. Construction-cost-overrun and schedule-delay distributions are fitted to the nuclear subsample of 401 electricity infrastructure projects reported by Sovacool, Gilbert and Nugent (2014), which records a mean cost escalation of 117.3%, a 75th percentile of 179%, and overruns affecting more than 97% of projects. Against that distribution the 75% escalation realized by the NuScale and UAMPS Carbon Free Power Project before its 2023 cancellation falls at the 39th percentile, and the median outcome is doubling of initial cost. A representative 300 MWe project is then modelled with debt sized at a 1.35 target coverage. The central result is prior to any overrun: at the $89 per MWh target price the developer itself disclosed in January 2023, the as-planned project supports $1,281m of debt against a $3,056m funding requirement, leaving 58% of project cost to be met by equity, grant, or federal cost-share, and full debt finance at the 6.0% base term rate would require a tariff between $148 and $162 per MWh. Once the calibrated overrun distribution is applied, modelled post-COD coverage falls below the 1.20 analytical threshold on 90.1% of 50,000 simulated paths.
Connecting the two layers, this study asks whether a financing structure that appears acceptable under a parallel-rate assumption remains resilient when the curve twists, steepens, or changes during construction. Sensitivity analysis shows the funding-gap result is robust across a 4% to 8% term-rate range, over which the tariff required for full debt finance moves from $125–$136 per MWh to $173–$190 per MWh, so that a 200-basis point error in the assumed term rate moves the required tariff by between $26 and $28 per MWh against a contract price of $89. The empirical KRD claim and the calibrated lifecycle claim remain explicitly separated throughout: a simulated reduction in default probability is not an achieved saving, and a scenario coverage ratio is not a forecast for any specific named reactor project.
JEL classification: G12; G31; L94; Q48.