The Restart Play: Buying Time You Cannot Build

The single most important thing to understand about the current nuclear cycle is that its fastest-moving pieces are not new reactors at all. They are old ones being switched back on. If you want firm, clean power this decade rather than next, the surest way to get it is to take a plant that already has a license, a cooling source, a switchyard, and a trained workforce, and return it to service. You are buying time you physically cannot build.

The emblem of this strategy is the Three Mile Island Unit 1 restart in Pennsylvania. Note carefully which unit. Unit 2 was the site of the 1979 partial meltdown, the accident that froze U.S. nuclear expansion for a generation. Unit 1 is the undamaged neighbor, a perfectly functional reactor that ran until 2019, when it was shut down for economics, not safety, in an era of cheap gas and flat demand. In 2024, its owner Constellation Energy announced it would refurbish and restart the unit under a twenty-year power purchase agreement with Microsoft, and rename it the Crane Clean Energy Center. That deal is the template: an existing asset, a single enormous creditworthy offtaker, and a long contract that underwrites the refurbishment cost. Faraday's read: the symbolism of restarting a reactor at Three Mile Island, of all places, tells you how far the political and commercial mood has swung.

The restart logic extends beyond full reactivations. There is also uprating, the practice of squeezing additional megawatts out of a running plant through equipment upgrades and revised operating limits, and there is life extension, relicensing plants to run sixty or eighty years rather than forty. Each of these adds firm, clean capacity without a greenfield construction project. And there are other dormant units in the fleet that owners are studying for potential return, though not every shuttered plant is a candidate. A reactor that has begun decommissioning, with components cut up and removed, is generally gone for good. The economics of restart depend on how far down the decommissioning path the plant has already traveled.

Why does this strategy exist at all? Because the alternative, new construction, is slow and expensive enough that a customer who needs power before 2030 has few other firm and clean options. A restart can plausibly deliver power in a few years. A new large reactor cannot. So the restart play is not really competing with SMRs. It is competing with waiting, and against the option of just buying gas. For the earliest data-center deals, restart is winning that comparison because it is the only nuclear path short enough to matter on the AI buildout's timeline.

But restarts have a ceiling, and it is worth being clear-eyed about it. There are only so many restartable plants. Once you have reactivated the handful of recently shuttered, not-yet-dismantled units, the well runs dry. Restarts are a bridge, not a destination. They can absorb the first wave of demand and buy the industry perhaps a decade. They cannot, by themselves, supply the sustained growth in firm, clean power that the load forecasts imply if AI demand keeps compounding. That is precisely the gap the small modular reactor is meant to fill, which is why the same customers signing restart deals today are also signing options and orders for SMRs that will not deliver until the 2030s.

So when you read a restart announcement, read it for what it is: a fast, real, but finite source of firm clean power, underwritten by a single deep-pocketed buyer, exploiting an asset that already cleared the hardest hurdles years ago. It is the most certain thing in the sector precisely because it involves the least that is new. The regulator is not licensing a novel design; it is reviewing the return to service of a plant it already knows. The construction risk is refurbishment risk, not first-of-a-kind risk. And the timeline is measured in a handful of years rather than a decade or more. In a sector famous for optimistic schedules that slip, the restart is the closest thing to a sure bet, and that certainty is exactly why the marquee early deals took this form.