# Is Three Mile Island's 835-MWe Restart Cleared for 2027?

On June 1, 2026, the Federal Energy Regulatory Commission approved a transmission waiver that removes the last major grid obstacle blocking [Constellation Energy](https://smrintel.com/companies/constellation-energy) from delivering full power from its restarted Three Mile Island Unit 1 — now the Christopher M. Crane Clean Energy Center — by 2027. The FERC ruling lets Constellation transfer 760 megawatts of grid-connection rights from its Eddystone plant near Philadelphia to the Crane site, bypassing a transmission upgrade bottleneck that would otherwise have constrained output until at least December 2030. The entire 835-MWe output is pre-sold under a 20-year power purchase agreement with Microsoft, signed in September 2024, making the Crane restart arguably the most commercially concrete large-nuclear-for-data-centers deal currently in execution anywhere in the United States.

The restart timeline has accelerated: Constellation now targets the second half of 2027, a year ahead of the 2028 date it originally projected. The $1.6 billion project received a $1 billion Department of Energy loan in November 2025. More than 600 jobs are returning to Dauphin County, Pennsylvania.

---

## Why This Is TMI Unit 1, Not the Reactor That Failed

The nomenclature matters and the source text is explicit on this point. Three Mile Island always housed two pressurized water reactors. Unit 2 suffered the partial meltdown on March 28, 1979 — the worst commercial nuclear accident in U.S. history — and never ran again. A separate company owns Unit 2 and continues decommissioning it today.

Unit 1 is an entirely distinct machine. It operated for roughly four decades after 1979 and was shut down in 2019 for a purely economic reason: cheap natural gas made it unprofitable to run, not any safety concern. Constellation is restarting Unit 1, renamed the Christopher M. Crane Clean Energy Center in honor of former Exelon CEO Chris Crane, who died in 2024 after a career centered on nuclear operations.

The distinction is not cosmetic. Restarting a reactor with a multi-decade operating history and known equipment condition is a fundamentally different risk profile than licensing a new design. Constellation's pitch to both investors and regulators is that the asset exists, the operating record exists, and the fuel cycle expertise exists — the variable is hardware reconditioning, not first-of-a-kind engineering.

---

## The Transmission Problem FERC Just Solved

PJM Interconnection's grid analysis found that the existing 500-kV and 765-kV transmission infrastructure near the Crane site could not safely carry the full 835 MWe until new lines are built — work PJM did not expect to complete until December 2030, with further slippage risk.

For a nuclear operator, that scenario is operationally damaging. Running a large reactor chronically below rated output to match constrained transmission capacity creates turbine vibration and thermal cycling stress that degrades equipment reliability over time. Extended part-power operation is not how a 835-MWe unit is designed to run.

Constellation's solution was jurisdictional arbitrage within PJM's own rules. It sought a waiver allowing it to move 760 megawatts of grid-interconnection rights from its Eddystone generating station, a plant near Philadelphia that Constellation is retiring, across the state to the Crane site. FERC approved the transfer on June 1, 2026, over the explicit objection of PJM's independent market monitor.

The market monitor's opposition is worth noting. Grid monitors typically flag such transfers because they can shift congestion costs onto other market participants or alter the locational marginal price signals that PJM uses to dispatch generation. FERC overriding the market monitor's recommendation suggests the commission weighed broader policy considerations — federal interest in nuclear restart, decarbonization commitments, data center load growth — against market structure purity. That tension will recur as more nuclear-for-data-center deals attempt to navigate legacy transmission infrastructure.

---

## What the Microsoft PPA Actually Means for the Grid

Under the 20-year PPA signed in September 2024, Microsoft takes the entirety of Crane's output — all 835 MWe, which Constellation characterizes as equivalent to roughly 800,000 homes' electricity consumption. The power flows into the PJM Interconnection footprint, which spans the Mid-Atlantic and Midwest and serves Pennsylvania and approximately a dozen other states. Microsoft's data centers across that footprint consume the power, though PJM's wholesale market structure means physical electrons don't flow point-to-point; Microsoft receives the equivalent in load-matched clean energy credits within the PJM system.

The commercial logic driving Microsoft here is straightforward and worth stating plainly for utility executives evaluating similar structures. [Baseload power](https://smrintel.com/glossary/baseload) from a nuclear plant operates at a [capacity factor](https://smrintel.com/glossary/capacity-factor) that variable renewables cannot match. AI inference and training workloads run continuously and cannot be cheaply curtailed when wind or solar generation drops. A 20-year fixed-price PPA on firm, carbon-free generation is, from a hyperscaler's perspective, a hedge against both power price volatility and ESG accounting risk. Amazon, Google, and Meta have each signed separate nuclear deals over the past two years, according to the source material, for the same underlying reasons.

The broader SMR sector should read this deal carefully. Most advanced reactor developers are pursuing [first-of-a-kind (FOAK)](https://smrintel.com/glossary/foak) designs that won't generate commercial power until the early 2030s at the earliest. Constellation's Crane restart demonstrates that a hyperscaler willing to write a 20-year check will do so for proven technology on an aggressive timeline. That commercial urgency — Microsoft needing firm power *now*, not in 2034 — is what drove the FERC waiver fight and the accelerated 2027 target. SMR developers competing for the same corporate PPA market need to understand they are benchmarking against a restarted 835-MWe proven plant, not against each other.

---

## Restart Hardware: What $1.6 Billion Buys

Constellation has conducted thousands of inspections across the plant's steam generator, main generator, rotor, turbines, and condensers. The company awarded a contract for three new main power transformers — described in the source as one of the largest single equipment purchases in the project — and committed an additional $35 million to accelerate transformer delivery to the site in 2026, according to World Nuclear News. Transformer procurement is a genuine long-lead constraint in nuclear restart projects; U.S. manufacturing capacity for large power transformers is limited, and delivery timelines frequently slip.

The company has also begun training and licensing the reactor operators who will staff the control room, upgraded the control room simulator, and restored the main office building. The DOE's $1 billion loan, closed in November 2025, reduces Constellation's cost of capital on the remaining construction spend and, per the source, allows savings to be redirected into plant upgrades rather than debt service.

CEO Joe Dominguez has publicly stated the plant can be "returned to service better than ever" — a claim that will be tested against the 2027 target and the post-restart [capacity factor](https://smrintel.com/glossary/capacity-factor) the unit achieves in its first operating cycle.

---

## Key Takeaways

- **FERC approved on June 1, 2026** a waiver allowing Constellation to transfer 760 MW of grid-interconnection rights from the retiring Eddystone plant to the Crane site, over PJM's market monitor's objection.
- **The full 835-MWe output** is locked into a 20-year PPA with Microsoft, covering all of the restarted unit's generation for data center load across PJM.
- **Restart target is now second-half 2027**, one year ahead of the original 2028 projection; the $1.6 billion project received a $1 billion DOE loan in November 2025.
- **This is TMI Unit 1**, not Unit 2 — the reactor that ran cleanly for decades and was closed in 2019 due to market economics, not safety.
- **FERC overriding PJM's market monitor** sets a notable precedent for how federal policy priorities may trump grid market structure when nuclear restart and data center load are both in play.
- **SMR developers face a benchmark problem**: corporate buyers with immediate load needs are signing deals for proven restart capacity, not future advanced reactor output.

---

## Frequently Asked Questions

**Is the Three Mile Island reactor being restarted the same one that had the 1979 accident?**
No. Three Mile Island had two separate reactors. Unit 2 suffered the 1979 partial meltdown and was permanently shut down. Unit 1, now renamed the Christopher M. Crane Clean Energy Center, operated independently for roughly four decades after 1979 and was closed in 2019 due to low natural gas prices, not any safety issue. Constellation is restarting Unit 1 only.

**Who is buying the power from the restarted Three Mile Island plant?**
Microsoft holds a 20-year power purchase agreement for the entire 835-MWe output of the Crane Clean Energy Center, signed in September 2024. The power serves Microsoft data centers within the PJM Interconnection grid region.

**What did FERC approve on June 1, 2026?**
FERC approved a waiver from PJM's interconnection rules that allows Constellation to shift 760 megawatts of grid-connection rights from its Eddystone plant near Philadelphia to the Crane site. Without the waiver, transmission upgrade work expected to take until at least December 2030 would have prevented the plant from delivering its full output on the 2027 restart timeline.

**How much is the TMI Unit 1 restart costing and how is it financed?**
Constellation has described the project as a $1.6 billion restart. In November 2025, the Department of Energy closed a $1 billion loan for the project, with Constellation stating the lower-cost federal debt allows savings to be reinvested in plant upgrades.

**What does this deal mean for small modular reactor developers?**
The Crane restart demonstrates that corporate buyers with urgent, firm-power needs will move quickly and pay for proven technology over promised future capacity. SMR developers targeting hyperscaler PPAs are, in practical terms, competing against restart economics and a 2027 delivery timeline — a materially different competitive environment than if large-scale nuclear options were limited to new builds alone.