Open in interactive viewer → charts, metric popovers & call review

NRG Sells the Fleet It Just Bought and Buys Back the Stock It Says Is Cheap

An unsolicited $4B bid lets NRG shrink to PJM and ERCOT — while the tape quietly fades the very data-center growth story management keeps promising.
NRG · Earnings Call · 2026-11-09

The flip

Nine months after closing the LS Power acquisition — the deal that took non-current debt from roughly $9.8B to more than $23B in a single quarter — NRG is selling just under 4.9 gigawatts of that same portfolio back to the market. Six facilities across New York, Texas and New England, $4B of gross proceeds, 8.0x expected 2027 adjusted EBITDA. NRG keeps the entire PJM generation fleet and CPower. asset sale is the company's top-ranked keyword this quarter and debt reduction sits right behind it, inverting the language of the last four quarters, when “LS Power acquisition” and “Power transaction” dominated the vocabulary. The concept has simply turned over. Management is careful to frame this as inbound, not strategic retreat:

We were not pursuing a sale. LS Power approached us with an unsolicited offer at 8.0x expected 2027 adjusted EBITDA. At that valuation, we saw a compelling opportunity to further focus our portfolio and create more value per share by redeploying the proceeds than by continuing to own those assets.

Robert Gaudette, President and CEO · 2026-11-09
The arithmetic behind that claim is the most concrete thing on the call. Jack County carries an implied value of roughly $1,400 per kilowatt; NRG says it is developing 1.5 GW of new gas in Texas for “just above $1,000 per kilowatt.” Sell the older machine at a premium, build the newer one at a discount, pocket the spread. Proceeds split $2B to debt, $1B to incremental buyback, $1B held for contracted growth or further deleveraging. That capital structure was built for exactly this. Non-current debt of $23.4B, up 138% year over year against a liabilities-to-assets ratio of 87.8%. Deleveraging is not optional here; it is the price of the buyback.

The deadline that keeps moving

The 1.2 GW Texas data center project has been the forward story for a year, and it is still forward. February's call promised a minimum of 1 GW of announcements in 2026, explicitly excluded from guidance. August's call delivered “principal terms” and this caution from the CEO: “we are commercially aligned... Moving forward is not done, and we will continue to push until we are” — Robert Gaudette, President, NRG Business · 2026-08-04 November's call delivers the same finish line, one quarter later: “We've agreed on the terms. It's just a matter of time. I feel very confident we're going to get it done” — Robert Gaudette, President and CEO · 2026-11-09 That is three consecutive quarters in which the definitive agreement is a quarter or two away. And crucially, the 2027 guidance of $5.75B adjusted EBITDA is built without it — the project is upside, not plan. Analysts noticed; the CFO declined to size the $1B placeholder for it, and on the 2029 COD mechanics he was blunt that they would not revisit the number. The genuinely new argument this quarter is political. Rather than treating Texas landowner and ratepayer backlash as a risk, management reads the election as a moat: “we have more demand for our turbines today after the election than we actually have in our joint development agreement with Kiewit and GE Vernova” — Robert Gaudette, President and CEO · 2026-11-09 That is a contrarian claim, and a testable one — if hyperscalers really are migrating toward “responsible” developers with community ties, the year-end announcement should confirm it. The batch process and SB 6 compliance are framed as boxes already checked.

What the tape is saying instead

Here is the discord. Across the global theme tape, AI data center exposure is being actively sold: down roughly 21% over 30 days with 8 advancers against 56 decliners, and down 23% over 90 days across a similarly lopsided breadth. Data-center capacity, co-packaged optics, high-bandwidth memory — all deep in the decliner column. NRG's own stock is down 30.9% in the last 90 trading days and 47% from its February 24 peak of $184.03. The company's growth equity story and the market's data-center de-rating are the same trade, pointed in opposite directions. ERCOT does not help. The CFO conceded prices have run “nearly 40% below our original planning assumption year-to-date” — Bruce Chung, CFO · 2026-11-09, and the CEO has been making the same structural point for two calls running — that Texas is absorbing a battery and solar build with no weather to mark the curve. In August he put it plainly: “the ERCOT market is not valuing anything right now. Prices are low.” — Robert Gaudette, President, NRG Business · 2026-08-04 The offset is PJM, and it is the one genuinely fresh positive in the quarter. higher PJM prices enters the company's keyword set for the first time, with the explicit nuance that inherited LS Power hedges mute the benefit in 2027 but extend hardly at all beyond it — so 2028 and later get the full uplift. Add a December capacity auction management expects to clear at the top of the collar, and the retained PJM fleet is the reason the sale leaves the company with more market exposure, not less. The drag is RGGI compliance costs, now $100M for 2027, $30M worse than expected.

What went quiet

The loudest absence is the consumer platform. In late 2024 the keyword set was dominated by Virtual Power Plant, Renew Home, the essentials bundle, Home Energy wallet share. As recently as Q1 2026, demand response ranked sixth. Today none of it cracks the top thirty. Smart Home gets a single sentence for 14% EBITDA growth, CPower survives the sale and is barely mentioned, and the CEO's closing remarks read like a rearguard action — “our strategy starts with the customer” is a line you say when the narrative has drifted elsewhere. What replaced it as a novel theme is defensive. Asked twice whether AI agents threaten the retail book, the retail president answered: “AI agents relatively early. We've not seen any shift in consumer behavior.” — Brad Bentley, Executive Vice President, Retail · 2026-11-09 The moat is contract duration — more than 80% of Texas customers on 24-to-36 month terms, a share rising four years straight — plus an under-penetrated East where fewer than 30% of customers have shopped. That is a credible answer, but it is the first time AI has been framed as a threat to NRG rather than a tool, and it aligns with a global “personal agent” motif now appearing across telecom and insurance reporters alike. The bet, in one line: a $3B two-year buyback on a $20.5B market cap, executed into a 47% drawdown, with a 14%-plus per-share CAGR reaffirmed “with or without” the asset sale. Either the market is mispricing a diversified, now-deleveraging cash generator — or the buyback is catching a falling knife in a state whose power curve refuses to cooperate.