The Energy Department Offered Constellation $5 Billion for Nuclear Uprates. Google and Amazon Already Bought Almost All the New Power.

Chris Wright picked a fitting place to make this announcement. On October 27th, the Energy Secretary stood inside Constellation Energy's Byron Clean Energy Center in Ogle County, Illinois, and told the room that his department was offering the company a conditional loan commitment of up to $5 billion. The plant behind him already supplies some of its output to Google under a contract Constellation signed three weeks earlier. The loan Wright announced that day exists, in large part, to help pay for work Google and Amazon had already agreed to buy.

The mechanics: DOE's Office of Energy Dominance Financing, the renamed Loan Programs Office, offered the commitment under its UPRISE program, launched in March with a target of adding 2.5 gigawatts of nuclear capacity by 2027 and 5 gigawatts by 2029 through uprates, license renewals, and restarts of shuttered plants. The money would fund uprates — equipment and licensing changes that let an existing reactor produce more than its original license allowed, without building anything new — at eight plants across three states: Braidwood, Byron, Clinton, Dresden, LaSalle, and Quad Cities in Illinois, Calvert Cliffs in Maryland, and Limerick in Pennsylvania. Combined, DOE and Constellation put the added output at close to 1.1 gigawatts, enough, DOE says, for more than 930,000 homes. The office's own figures put its total remaining loan authority at more than $289 billion, so a $5 billion ceiling here is a fraction of what it could still deploy toward the same kind of work elsewhere in the fleet.

It's worth being precise about what an uprate actually is, because the word does a lot of work in this story. A reactor's original license sets a maximum thermal output, and an uprate is a formal, NRC-reviewed change to that ceiling — new turbine blades, upgraded steam generators and condensers, recalibrated instrumentation — that lets the same core produce more electricity without touching the fuel, the containment structure, or the footprint of the plant. None of the heavy civil engineering that makes a new reactor take a decade or more is involved. What's left is equipment procurement and a license amendment docket at each unit, which is a real process with real risk of delay, but a fundamentally shorter and cheaper one than pouring new concrete.

Google and Amazon got there first

We've covered most of this capacity already, because it isn't new. On October 6th, Google signed a 20-year power purchase agreement for 890 MW of uprate capacity across 11 Constellation units, with Limerick's two units carrying the single largest share, 348 MW. A week before that, Amazon signed its own 20-year deal for 690 MW at Calvert Cliffs, roughly 190 MW of it from a planned uprate. Constellation's own figure for what Google and Amazon have already committed to buy is more than $7 billion for about 1,080 megawatts — a number that lines up almost exactly with the 1.1 GW this loan is meant to help finance. One industry analysis also puts Meta as the buyer effectively taking the output of the Clinton plant, though I haven't seen that confirmed in a Constellation filing, so treat it as reported rather than established.

In other words: by the time Wright stood up in Ogle County, the hard commercial work — finding a buyer willing to sign a two-decade contract at a price that makes the uprate pencil — was mostly done. What the federal government is offering is cheaper capital on top of deals that already exist. The timing lines up with why those buyers moved so fast in the first place: PJM, the grid operator covering all three states this loan touches, came up 6.8 gigawatts short of its own reliability target in a September backstop capacity auction that never actually opened, with its own independent market monitor naming data-center load as the biggest driver of the shortfall. Firm, carbon-free megawatts that can be running in a few years rather than a decade are genuinely scarce in that market right now, which is a big part of why Google and Amazon were willing to sign at all.

What the loan is, and isn't

Start with what it isn't: a done deal. DOE's commitment is conditional, and Constellation's own statement is careful to say it "does not, by itself, represent a final Constellation decision to proceed with any individual uprate or at every facility under consideration." Before any money moves, the company has to satisfy DOE's technical, legal, environmental, and financial conditions and sign definitive financing documents. The program caps federal financing at 80% of eligible project costs, so Constellation and its private partners are still carrying real exposure, not handing the whole bill to taxpayers.

This is how America wins the race for energy abundance — by pairing private-sector demand and investment with smart federal financing to unlock more power from the nuclear plants we already have.Joe Dominguez, Chairman, President, and CEO, Constellation Energy

The Energy Department is accelerating President Trump's American nuclear renaissance and unlocking the full potential of our nation's nuclear fleet.Chris Wright, U.S. Secretary of Energy

The question worth asking

I don't think the uprates themselves are a bad idea. Squeezing 1.1 GW out of reactors that are already built, staffed, and connected to the grid is about the fastest, cheapest way to add firm carbon-free capacity that exists right now — faster than a gas turbine order backlog, far faster than a new reactor. Constellation's last uprate round, about 135 MW split between Byron and Braidwood, cost roughly $800 million and didn't need a federal loan at all. What's different this time is scale, and whose money is in the stack. The same week DOE offered Constellation this commitment, it also closed a separate $5.1 billion loan to Alliant Energy for gas plants, battery storage, and generation upgrades in Iowa and Wisconsin, split between its Wisconsin Power & Light and Interstate Power and Light subsidiaries. DOE called that its tenth financial close under this administration, and taken together, the two announcements come to roughly $10.1 billion in federal credit for about 3.7 gigawatts in a single week. That's a lot of government lending flowing toward a sector whose biggest customers are four of the most cash-rich companies on earth, and that already closed more than $7 billion of its own private financing before the government showed up with a loan at all.

None of this touches household electric bills in Illinois, Maryland, or Pennsylvania, which is the real, uncomplicated good news here — Google and Amazon's 20-year contracts are the commercial backbone of the 1,080 MW this loan supports, not a rate case in front of any of those states' utility commissions. But a federal loan is still somebody's money, backstopped by the same taxpayers who aren't on the hook for anyone's electric bill in this particular deal, and the honest version of this story is that it's increasingly hard to tell where "accelerating America's nuclear fleet" ends and "subsidizing contracts the private sector already signed" begins. The 80% financing cap matters here precisely because it means the answer isn't "all of it" — Constellation and its lenders still have to carry a real share of the risk if an uprate runs over budget or a license amendment stalls at the NRC, which is a different proposition than a blank check.

What to check next

On GridPermit's evidence ladder, this sits below the Google and Amazon power purchase agreements it's meant to help finance, not above them: a conditional term sheet with named conditions still to be satisfied ranks lower than a signed 20-year contract with a dollar figure and a delivery date attached. The things actually worth tracking are whether DOE and Constellation reach definitive financing documents, whether the Nuclear Regulatory Commission grants the license amendment each uprated unit still needs, and whether the Meta-Clinton connection gets confirmed anywhere beyond secondhand analysis. The power itself was already sold. What's still open is only the paperwork behind it, and paperwork is where projects like this actually stall or slip.