Grid Power, Tariffs and Contracts
On September 20, 2024, Constellation Energy announced a twenty-year power purchase agreement with Microsoft tied to the restart of Unit 1 at the Three Mile Island site in Pennsylvania, to be renamed the Crane Clean Energy Center. The reactor had been shut down in 2019 for economic reasons. In November 2025 the Department of Energy closed a loan of about one billion dollars to Constellation for the project, and Constellation has said it hopes to return the unit to service as early as 2027. As of September 2026 the Nuclear Regulatory Commission was still working through the licensing steps. The deal is a good way into the question of where a data center's electricity actually comes from.
Start with the ordinary case. Most data centers buy electricity from a local utility under a tariff, which is a rate schedule approved by a state regulator. The tariff sets charges for energy consumed, for peak demand, and for delivery over the utility's wires. In states with retail competition, a large customer may buy the energy itself from a competitive supplier while still paying the utility for delivery. Either way, the electrons that reach the building come from the grid as a whole, a pool into which many power plants feed.
A power purchase agreement, or PPA, is a long-term contract to buy electricity, or its financial value, from a specific generator at an agreed price. The Microsoft and Constellation agreement is one. So is the twenty-year deal Meta announced with Constellation in June 2025 for output from the Clinton nuclear plant in Illinois. PPAs come in two broad forms. In a physical PPA, the buyer takes delivery of the power in the same market where it operates. In a virtual or financial PPA, the generator sells its power into the wholesale market and the two parties settle the difference between the market price and the contract price; the buyer keeps buying from its local utility as before.
Why would a company sign such a contract? One reason is price certainty over many years. Another is the environmental claim: the buyer can say it has paid for a specific amount of low-carbon generation. A third, increasingly important, reason is that a long contract can help a generator justify building, restarting or extending a plant that would not otherwise exist, adding supply to a grid that needs it. That last point is where the Crane restart is distinctive: it brings a reactor back into operation rather than relabeling output that would have been produced anyway.
It is worth being precise about what a PPA does not do. A virtual PPA with a plant in one state does not physically deliver electricity to a data center in another. The data center still depends on the local grid and the local utility for its actual supply, its connection date and its reliability. Critics of some corporate clean-energy claims make exactly this point, and some buyers have responded by matching purchases to the hours and regions in which they actually consume power. Google, which announced in 2020 a goal of running on carbon-free energy every hour of every day on every grid where it operates by 2030, is the best-known example.
The nuclear deals have also exposed a question of fairness. When an existing plant that already serves the grid signs a contract dedicating its output to one customer, other customers may lose access to that supply, at least on paper, and regulators have to decide whether the arrangement shifts costs onto them. Restarts and new plants avoid part of that objection because they add supply. Contracts for existing plants invite it.
For a reader evaluating a company's power position, the useful questions are concrete. Is the contract physical or financial? Does it add new generation, or does it claim existing output? What happens if the plant is late, and who bears the cost of replacement power in the meantime? The Crane agreement, for example, depends on a restart whose timing was still subject to regulatory approval when this lesson was written.
Faraday's read is that a power purchase agreement is best read as a financial and reputational instrument first and a physical supply arrangement second, with the important exception of contracts that bring genuinely new capacity onto the grid. The next lesson looks at the arrangements that try to take the grid out of the picture altogether.