Power Bottlenecks Push SOFCs to GW-Scale

Rapid growth in AI compute is reshaping how data centers select power solutions, shifting the priority from the lowest levelized cost of electricity (LCOE) to faster time to power. Among new-build power solutions for AI data centers, generation costs rank from lowest to highest as follows: nuclear power < heavy-duty gas turbines < aeroderivative gas turbines < coal-fired power < SOFCs < solar + ESS + diesel genset. However, lower-cost solutions generally require longer construction lead times: nuclear power > heavy-duty gas turbines ≈ coal-fired power > aeroderivative gas turbines > SOFCs ≈ solar + ESS + diesel genset. We estimate that the implied opportunity cost of idle AI compute capacity is currently around USD500/MWh, substantially higher than the LCOE premium of only tens of USD/MWh for rapidly deployable solutions such as SOFCs relative to grid power or conventional gas turbines. On the policy side, following FERC's standardization of large-load interconnection rules in June, Ne