Capital Wealth
TUE CLOSE · OCT 6   S&P 500 7,818.93 ▲0.58%  ·  DJIA 51,521.28 ▲0.49%  ·  NASDAQ 27,599.79 ▲0.45%  ·  10-YR 5.27%  ·  2-YR 4.79%  ·  WTI $89.44 ▲0.01%  ·  GOLD $4,159.20 ▲0.75%  ·  VIX 15.01 ▼3.3%
Business & Finance · Energy · IN04

Google Will Pay 20 Years for 890 MW Squeezed From 11 Reactors Constellation Already Runs — Constellation Spends $4.3 Billion, CEG +12%, Held

Nobody has to pour concrete. Constellation will squeeze a new reactor’s worth of power out of 11 plants it already runs, and Google will pay for 20 years of it. The AI power trade keeps compounding — and the desk keeps holding, not chasing.

By Sean Anees Saifi · Capital Wealth · Published Wednesday, October 7, 2026 · Source: The Wall Street Journal, Wednesday, October 7, 2026, whose market figures are the Tuesday, October 6 close (Business & Finance)
Key Points
890 MW
output added across 11 reactors under the 20-year Google deal
$4.3B+
Constellation’s planned investment in the uprates
+12%
CEG on Tuesday, to $300.40; VST +11%, NRG +7%
190 MW
Amazon’s Calvert Cliffs uprate, signed last week
Two concrete cooling towers release white plumes of vapor into a clear blue sky, with steel transmission towers and power lines crossing harvested golden fields and a band of autumn trees in the foreground.
An uprate squeezes more megawatts from reactors that already exist — faster than new construction, which last ran more than $16 billion over budget.
In one line: Google will pay 20 years for 890 MW squeezed out of 11 reactors Constellation already runs, CEG jumped 12%, and the desk holds CEG and NEE at weight — the AI power trade compounds, but we don’t chase a 12% day.

The cheapest new reactor is the one you already have. That’s the whole idea behind Tuesday’s deal, reported by Jennifer Hiller: Google — the Alphabet (GOOGL) unit — will buy 890 megawatts for 20 years from Constellation Energy (CEG), which gets it not by building anything but by upgrading 11 reactors it already runs in Illinois, Pennsylvania and New Jersey. The added output is roughly a large new reactor’s worth. Constellation will spend more than $4.3 billion to do it; the stock rose 12% on the day, closing at $300.40, and took the whole power patch along — Talen up 12%, Vistra (VST) 11%, NRG 7%. The Dow Utility Average’s 2.59% gain was the best of any index Tuesday.

“These are expensive things to be done,” Constellation CEO Joe Dominguez told the Journal — the company has known for years it could do these uprates; it’s Google’s 20-year check that makes them bankable. The first is due in 2028, the rest through 2032 or 2033, and a separate 15-year agreement supplies another 2,700 megawatts in the PJM market across the Mid-Atlantic and Midwest. Google’s energy chief, Peterson Corio framed it as speed to firm, clean power at “a moment where data centers are under a lot of criticism.”

It’s the third such deal in a week of papers. Monday’s Journal had Amazon (AMZN), fresh from abandoning a data-center campus beside Maryland’s Calvert Cliffs plant over local opposition, paying Constellation instead to add 190 megawatts there — about 147,000 homes’ worth — in exchange for up to 690 megawatts long term from across Constellation’s fleet, plus a look at new reactors on the site. Meta has a 20-year, 433-megawatt uprate deal with Vistra; Microsoft is funding the Three Mile Island restart; Google and NextEra Energy (NEE) are reopening Iowa’s Duane Arnold by early 2029.

Because the alternative is grim. The U.S. has switched on only a few new commercial reactors this century, and the most recent big build ended up more than $16 billion over its budget and seven years past its deadline. Squeezing output from plants built in the 1970s — Calvert Cliffs’ two reactors date from then and face relicensing in 2034 and 2036 — is the quickest nuclear megawatt there is, and a 20-year hyperscaler contract is what pays to keep those plants alive.

Because the alternative is grim. Only a handful of new commercial reactors have come online in the U.S. since 2000, and the last big project finished more than $16 billion over budget and seven years late. Squeezing output from plants built in the 1970s — Calvert Cliffs’ two reactors date from then and face relicensing in 2034 and 2036 — is the quickest nuclear megawatt there is, and a 20-year hyperscaler contract is what pays to keep those plants alive.

Our read

Investments and Risk (IN04). We own this trade already, which is the point: Constellation is held in 9 of our 81 model books, NextEra in 21, Vistra in 15, and the buyers — Alphabet, Amazon, Microsoft, Meta — are core mega-cap holdings. All of it is held at weight; nothing added. A 12% day is the kind that invites a chase, and we’d remind ourselves that a stock that can jump 12% on one contract can give it back on one delay — nuclear operators trade on deals that arrive in lumps, and the lumps cut both ways. The thesis — scarce firm power, buyers with 20-year horizons and fortress balance sheets — was just as true last week at a lower price, and it’ll be judged over the life of the contracts, not the week.

For a retiree reading this as a utility story: these aren’t your grandfather’s utilities. The regulated names (NextEra, Duke, Southern) still pay you to wait; the merchant generators (Constellation, Vistra, Talen) are now priced like growth stocks on data-center demand. Know which one you own and why. If Tuesday pushed a position past its place in the plan, that’s a sizing question — best settled on a calm afternoon, not the next headline. We’d rather check the gutters now than during the storm.

What It Means For Your Portfolio

Hold — CEG and NEE at weight; don’t chase a 12% day

No portfolio action — Constellation (9 books), NextEra (21) and Vistra (15) stay at weight alongside the hyperscalers signing the checks; the AI power thesis is intact and already owned.

General planning principles, not advice for anyone in particular. Separate your utilities into the two kinds they’ve become: regulated, dividend-paying names you hold for income, and merchant generators whose price now rides on data-center contracts. The second kind belongs in the growth sleeve and should be sized like one.

When a holding jumps double digits on a single contract, resist adding; check whether the position has drifted above its target weight and trim back if so. Twenty-year deals are judged over 20 years — the market will reprice them many times before then.

Book a 15-Minute Review → Back to Edition No. 180 →