The 890 MW figure is the part worth dwelling on. It does not come from new construction. It comes from investing in new equipment and technology at existing Constellation reactors to extract additional output from plants already licensed, already interconnected, and already operating. The combined effect is equivalent to adding a new reactor to PJM without building one.
Reported at roughly $4.3 billion across the two contracts, the deal also incorporates operational load-shaping and demand-response capabilities, meaning Google has agreed to modulate consumption during grid stress. That provision is as notable as the megawatts.
Uprates as the Overlooked Resource
The nuclear conversation of the past two years has been dominated by small modular reactors, which remain years from commercial deployment. Uprates at existing plants deliver firm, carbon-free capacity on a timeline measured in months to a few years, inside existing licences, on existing sites, with existing transmission.
The constraint has never been engineering feasibility. It has been that no counterparty would fund the capital expenditure against uncertain merchant revenue. A twenty-year contract from an investment-grade buyer resolves exactly that problem, which is why this structure is likely to be replicated.
The Demand-Response Precedent
Parselion's assessment: the load-shaping commitment matters more than the headline capacity. A hyperscaler contractually agreeing to curtail during grid stress establishes a precedent that regulators in every PJM state will now cite. Operators negotiating new interconnection agreements should expect flexibility obligations to shift from a voluntary concession to a standard term, and should price the operational cost of that flexibility into their models now.