NRG Energy, Inc. (NYSE: NRG) — A Cash-Returning Merchant Priced for a Data-Center Deal It Hasn’t Signed

Independent fundamental research. As-of date: 2026-06-20. Primary listing: NYSE (secondary NYSE Texas). CIK 0001013871. Fiscal year ends December 31. C-corporation, issues a Form 1099 (not a K-1).

⚡ Claude’s Take

This block is the author’s own subjective opinion and general information, not investment advice. The analysis that follows it takes no position and contains no price target; this section is the single exception.

Verdict: HOLD / accumulate-on-weakness toward the low-$110s–low-$120s. Not a short; not a chase at $135. NRG is the cheapest, most-levered, and only nuclear-less seat at the AI-power table — the fossil-and-retail mirror of Vistra (VST) and Constellation (CEG). It is a genuine cash machine: the integrated retail-load-plus-owned-gas model is a real volatility damper, FY2026 guidance is Adj EPS ~$8.90 / FCF-before-growth ~$14.50 per share (a ~10.7% FCFbG yield at $135), and management targets “at least 14%” compound growth in both metrics through 2030 — explicitly before any data-center contracts. After a ~6x run from ~$30 (late-2023) to a $183.47 all-time high (Feb-24-2026), the stock has given back ~26% to ~$135 on a March secondary-offering overhang, a soft-weather Q1 miss, and a broad de-rating of the entire AI-power complex. The euphoria premium is largely gone; what is left is a high-single-digit-EV/EBITDA, ~15x-earnings merchant trading at the ~68th–74th percentile of its own ten-year P/S and P/B history — fair-with-an-option, not deep value.

What caps my enthusiasm is the quality of the franchise underneath the cash flow. Consolidated ROIC has fallen from ~25% (2021) to ~9% (2025) as serial debt-funded, goodwill-heavy M&A — Direct Energy, the value-destructive Vivint deal, and now the ~$12B LS Power gas fleet (closed Jan-30-2026, doubling capacity to ~25 GW) — ballooned the asset base faster than returns; the comp plan rewards Adjusted EBITDA and FCF size with no ROIC and no per-share-EPS metric, precisely the incentive that green-lights scale-adding deals while returns erode, and say-on-pay just slid to ~87%. The balance sheet is sub-investment-grade at the corporate level (S&P BB / Moody’s Ba1 / Fitch BB+), re-levered toward ~3.0x by LS Power. And the headline reason to own the stock — a ~6 GW turbine-reservation pipeline that management frames as “>$2.5B of recurring EBITDA” — rests on 445 MW of actually-signed data-center load after ~18 months of “active discussions” and a fleet that sold its only nuclear asset (South Texas Project) in 2023, shutting NRG out of the 20-year hyperscaler nuclear PPAs that VST and CEG are signing. The premium prices optionality you cannot yet underwrite. Framing: a high-FCF, capital-returning operator in a structurally cyclical, no-moat industry, re-priced from mania to fair — own the cash-return engine on weakness; do not pay up for the AI option until a contract is on paper. The pro-cyclical buyback (heaviest dollars spent near the $166–183 highs) and zero insider buying since the December-2022 trough tell you management is not signaling deep value here either.

Conviction: medium. Flips bullish if NRG signs its first ≥1 GW “bring-your-own-power” data-center contract at disclosed >$90/MWh economics — that converts narrative optionality into a contracted annuity and re-rates the stock — and ERCOT/PJM forwards hold. Flips bearish if the data-center deals keep slipping (still zero gigawatt-scale signings into 2027), ERCOT energy-only margins and the PJM capacity-cap assumption re-base lower as Texas-Energy-Fund gas and batteries flood in, and LS Power integration/leverage strains the plan — that strips EBITDA and the multiple at once. Tag: the cheapest ticket to the AI-power show, sold by the one act without a nuclear encore — and the marquee deal is still unsigned.

📈 Stock Price Action — Five-Year Event Map

Over five years NRG round-tripped from a ~$30s commodity-merchant also-ran to a ~$183 AI-power darling and back ~26%: from a ~$29 low (late-2023, post-Vivint, mid-Elliott) to a $183.47 all-time-high close (Feb-24-2026) to ~$135.06 (Jun-18-2026), inside a 52-week range of $120.65–$183.47, now ~26% below the peak. This is a ~6x three-year advance (y3 total return ~+61% annualized, per the factor model) that has partially mean-reverted — the stock that was a one-way street up is now in a momentum reset, not a fundamental break.

Cycle narrative: (1–2) NRG spent 2021–22 as a cheap, levered merchant whose December-2022 Vivint acquisition was so poorly received it drew Elliott Management. (3) The 2023 activist campaign forced out the CEO, monetized the nuclear stake, and pivoted NRG to a buyback-and-discipline story — the stock doubled off the lows. (4–5) The AI-data-center electricity-demand narrative then re-rated the entire independent-power complex; NRG’s May-2025 LS Power acquisition doubled its gas fleet and let it raise its long-term growth target, pushing the stock to ~$165 by year-end. (6) The Feb-24-2026 ATH coincided with the record FY2025 print, the LS Power close, and the unveiling of a “>14% to 2030” plan — peak optimism. (7) From there the stock fell ~26% on a confluence of technical and fundamental pressure: LS Power dumping 14.3M shares in a March secondary, a genuine Q1-2026 earnings miss on mild Texas weather and higher interest expense, and a sector-wide unwind of the AI-power trade (Vistra, Constellation, GE Vernova all sold off together). Each price move is a Fact; the attributed driver is Interpretation, cross-referenced to earnings dates, 8-K events, and the news feed.

1. Executive Summary

NRG Energy is an integrated competitive power producer, retail electricity supplier, and home-services company headquartered in Houston, Texas. It pairs the largest competitively-served residential electricity book in the United States (~6 million energy customers under Reliant, Direct Energy, Green Mountain Energy, NRG, and Cirro) with an owned, dispatchable generation fleet that — following the January-30-2026 close of the ~$12B LS Power acquisition — roughly doubled to ~25 GW, now >75% natural gas with zero nuclear, plus a ~2.4-million-subscriber Vivint Smart Home segment acquired in 2023. FY2025 revenue was $30.7B, Adjusted EBITDA ~$4.1B, and Adjusted EPS $8.24 (+21% YoY). (FACT — FY2025 10-K; ROIC.ai.)

The investment tension. Merchant power generation is, on the base rate, a structurally bad business — a megawatt-hour is the canonical commodity, price is set by the marginal unit, barriers to entry are low, and the sector’s capital cycles have repeatedly bankrupted their participants (NRG itself reorganized in 2003-04; Calpine, Dynegy, Mirant, GenOn all touched bankruptcy). What partially redeems NRG is (1) the integrated model — a short physical generation position hedged by a long retail-load position, a genuine natural hedge that damps the wholesale cyclicality that destroyed the pure merchants; and (2) a demonstrated capital-return competency post the 2023 Elliott campaign — a share count cut ~22% (245M → 190M) and a 7–9%-growing dividend. But the same period bloated the asset base with serial debt-funded, goodwill-heavy M&A, dragging consolidated ROIC from ~25% (2021) to ~9% (2025) — roughly the cost of capital.

The de-rate is the setup. From a $183.47 all-time-high close (Feb-24-2026), NRG has fallen ~26% to ~$135 on a March secondary-offering overhang, a soft-weather Q1-2026 miss, and a complex-wide AI-power unwind. The drawdown stripped the euphoria premium and re-priced NRG to ~8.6x forward EV/Adjusted-EBITDA, ~15x forward Adjusted EPS, and a ~10.7% FCF-before-growth yield — the cheapest of the three AI-power merchants (vs VST ~10x, CEG ~14x EV/EBITDA), consistent with NRG’s lower-quality fuel mix (no nuclear), higher leverage (sub-IG), and the unproven data-center option. On its own ten-year history NRG is not cheap — ~68th-percentile P/S and ~74th-percentile P/B — but it is not at a fantasy multiple either. (The 99th-percentile, 112x trailing P/E is a GAAP artifact of derivative mark-to-market and must be ignored; value NRG on Adjusted EBITDA and FCFbG only.)

What the analysis finds: (1) Narrow, fuel-disadvantaged moat — the integrated retail+gas cost/hedging edge is real but commoditized and capped by low retail switching costs; NRG sold its only nuclear asset and is therefore shut out of the most-prized 20-year hyperscaler PPA vector that VST/CEG dominate. (2) GAAP earnings are noise — non-cash hedge mark-to-market swung GAAP net income from −$202M (2023) to +$1,125M (2024) to +$864M (2025) while Adjusted EBITDA rose; only Adjusted EBITDA and FCFbG are decision-useful. (3) A real cash machine on a sub-IG balance sheet — FY2026 FCFbG guided ~$3.05B (~$14.50/sh), funding ~$11B of planned 2026-30 buybacks, but net debt has risen to ~$18-19B pro-forma and the corporate credit is BB/Ba1. (4) A favorable industry capital-cycle, for now — physically scarce supply (turbines sold out to ~2030, coal retiring, queues clogged) meeting inflecting demand (data centers, electrification), but with two bounding risks: AI-capex/overbuild mean-reversion (ERCOT, energy-only, reverts fastest) and political price intervention (PJM caps, ERCOT SB6 curtailment). (5) Mixed capital allocation — a genuine return-of-capital culture undercut by pro-cyclical buyback timing (heaviest dollars near the highs), a re-leveraging M&A appetite Elliott tried to curb, a comp plan with no ROIC metric, and zero insider buying since December 2022.

The central question the price hinges on: does the ~6 GW / $2.5B-EBITDA data-center prize convert from equipment reservations and non-binding letters of intent into signed, long-dated contracts — and does it do so before the favorable power-price window the plan assumes begins to mean-revert? That is unknowable until the contracts (or their absence) arrive. The analysis below discusses valuation only as embedded expectations and scenarios; the single exception is the labeled opinion block above.

2. Business Overview

What NRG is. NRG Energy is an “integrated power” company — a competitive (merchant) generator bolted to the largest competitively-served residential retail electricity book in the United States, plus a smart-home subscription business. It is not a regulated utility: it owns no rate base, earns no allowed return, and bears commodity-price and weather risk. It buys and sells power, natural gas, and environmental/financial products; owns and dispatches power plants; serves ~6 million retail energy customers and ~2.4 million Vivint smart-home subscribers; and trades commodities to hedge and optimize the portfolio. Founded 1989, headquartered in Houston, ~16,000+ employees post-LS-Power. (FACT — FY2025 10-K, Item 1; company profile.)

Reporting segments and where the money actually is. NRG reports five segments: Texas, East, West/Services/Other, Vivint Smart Home, and Corporate. Revenue ($30.7B FY2025) is ~96% retail/commodity pass-through and tells you almost nothing — the signal is in gross margin (~$6.5B FY2025) and segment Adjusted EBITDA:

So generation + retail ≈ 84% and Vivint ≈ 17% of gross margin, but Vivint contributes only ~$33M of segment net income after ~$810M of D&A and subscriber-acquisition-cost amortization — i.e., it is a large gross-margin contributor whose GAAP bottom line is consumed by the cost of acquiring subscribers. (FACT — FY2025 10-K segment tables.) On the company’s adjusted basis, FY2025 consolidated Adjusted EBITDA was ~$4.09B and Vivint (“Smart Home”) contributed ~$1.09B of it. (FACT — company FY2025 results.)

The integrated model — the central structural argument. Generation is a short physical position (you must sell power); retail load is a long demand position (you must buy power to serve customers). Combined, the two partially offset, so the integrated entity carries lower earnings volatility than either standalone. This is genuine diversification and a real volatility damper — it is the mechanism that lets NRG, VST, and the other survivors run merchant fleets that the pure-play merchants of the 2000s could not. Whether it is a moat (a barrier to entry) versus merely diversification is a separate question (it is closer to diversification than moat). (FACT/INTERPRETATION — 10-K; standard industry analysis.)

The retail book. ~6 million retail energy customers across Reliant (Texas, ~19% competitively-served residential share — second to Vistra’s TXU), Direct Energy (multi-state + Canada), Green Mountain Energy (renewables brand), NRG, and Cirro. Capital-light and high-return on a unit basis, but customer counts are flat-to-declining (~5,632K energy customers ending 2025 vs ~5,759K in 2023) — growth has been price/margin, not units. (FACT — 10-K customer-count disclosures.)

Vivint Smart Home. Acquired March 2023 for ~$2.8B equity / ~$5.2B enterprise value; ~2.42M subscribers (2025, +8.7% YoY), ~90% retention. The thesis was cross-selling energy + smart-home + the “consumer of the future”; three years in, the cross-sell remains largely unquantified, the only concrete proof-point being a ~1 GW Texas residential virtual-power-plant. (FACT — 10-K; company disclosures.)

Generation fleet. Pre-LS-Power ~12 GW (Texas-heavy, gas/coal); post-LS-Power (closed Jan-30-2026) ~25 GW across ~9+ states, >75% gas, ZERO nuclear (NRG sold its 44% South Texas Project nuclear stake to Constellation for $1.75B in November 2023). Plus ~1.5 GW of new ERCOT gas being built under Texas Energy Fund 3%-interest loans. (FACT — 10-K; LS Power 8-K.)

Recurring vs. commodity. Vivint subscriptions and the retail base load are quasi-recurring; generation margins are commodity- and weather-driven (the swing factor, as Q1-2026’s mild-Texas-weather miss demonstrated). Reported revenue and GAAP earnings are dominated by non-cash derivative mark-to-market and pass-through commodity costs — neither is decision-useful.

Verdict (business model): A merchant generator + the largest U.S. competitive-retail book + a smart-home subscription business. The integrated retail/generation pairing is a real, value-adding volatility hedge and cash generator. It is not a regulated annuity; it is a commodity merchant with a partial physical hedge, now substantially larger and more gas-weighted, run by a management team with a strong return-of-capital culture and a re-leveraging acquisition appetite.

3. Industry Dynamics

The base rate is bad — but the cycle has turned. Through both the Greenwald (barriers-to-entry) and Marathon (capital-cycle) lenses, competitive power generation is historically a structurally bad industry: a perfect commodity, no differentiation, marginal-cost pricing, low entry barriers, and repeated boom/bust cycles that destroyed equity (the early-2000s merchant build-out bankrupted Calpine, NRG, Mirant; the 2010s gas glut crushed coal economics). In Greenwald’s terms, no supply/demand/scale advantage means ROIC mean-reverts to WACC. The bull case must specify what overrides that prior — and for how long.

What overrides it now: a genuine supply/demand inflection (the Marathon up-cycle). For ~20 years U.S. electricity demand was roughly flat as efficiency offset growth. That has decisively reversed: data centers (AI training/inference), electrification (EVs, heat pumps), and reshored manufacturing are inflecting load just as the supply side is physically constrained. NERC’s ten-year forecast shows U.S. peak demand up ~69% over the next decade in the high case; gas-turbine order books at GE Vernova/Siemens/Mitsubishi are effectively sold out to ~2029-2030; coal continues to retire; and interconnection queues are clogged with years-long backlogs. This is the favorable early-to-mid phase of a Marathon capital cycle — high prices that cannot yet attract sufficient new supply because the supply takes 4-6 years to build. The survivors are consolidating into it (Constellation–Calpine ~$26.6B; NRG–LS Power ~$12B; Vistra–Energy Harbor/Lotus/Cogentrix). (FACT/INTERPRETATION — NERC LTRA; trade press; deal filings.)

Two markets, two structures. NRG’s earnings concentrate in (a) ERCOT (Texas) — an energy-only market with no capacity payments, where generators earn volatile scarcity-energy and ancillary margins, and where the February-2021 Winter Storm Uri tail-risk (and the post-Uri price caps that truncate upside) lives; and (b) PJM and the Northeast (the East/LS-Power fleet) — a capacity market where the 2024 and 2025 base-residual auctions cleared at or near the administrative price cap (~$270-330/MW-day), a multi-year tailwind to dispatchable capacity. The mix matters: ERCOT offers the most upside torque and the fastest mean-reversion; PJM offers a capacity annuity that is politically contested (the Pennsylvania governor’s price-cap “collar,” FERC reforms). (FACT — market design; PJM auction results.)

The two bounding risks (where the capital cycle bites back). (1) Overbuild / AI-capex cooling — if hyperscaler capex slows or speculative load-queue requests (ERCOT’s interconnection queue holds >36 GW of large-load requests by 2033, far more than will materialize) fail to convert, the demand ramp underdelivers and ERCOT energy margins mean-revert first; Texas-Energy-Fund-subsidized gas and rapidly-growing battery storage are already adding supply. (2) Political price intervention — energy is politically sensitive; PJM capacity caps, ERCOT SB6 large-load curtailment rules (signed June 2025), and post-Uri scarcity-price caps all truncate the right tail. The VST analysis flagged ERCOT forwards already softening and batteries “returning virtually nothing” — the same Texas supply response that pressures NRG’s single largest market.

Verdict (industry): Structurally good right now — a favorable, supply-constrained capital-cycle up-phase with a real, multi-year demand inflection — overlaying a structurally bad long-run base rate. The window is real but finite and politically/competitively bounded; NRG’s heavy ERCOT (energy-only) and gas weighting place it on the higher-torque, less-protected, faster-mean-reverting side of the opportunity than nuclear-heavy CEG.

4. Competitive Position

Name the moat. NRG’s durable advantages, in Greenwald’s taxonomy, are (a) a modest economies-of-scale + customer-captivity edge in competitive retail (the largest competitively-served residential book, brand recognition in Reliant/Green Mountain, billing/servicing scale), and (b) a cost/hedging advantage from pairing that retail load with owned, flexible gas generation (the integrated natural hedge). Both are real; neither is wide.

Pressure-test the retail “moat.” Retail electricity has low switching costs (customers can change providers in Texas in days), many competitors, and flat-to-declining unit counts at NRG. The advantage is a cost-to-serve and customer-acquisition-efficiency edge from scale and brand, not customer lock-in. It is a moderate cost/captivity advantage, demonstrably not a barrier that produces excess returns indefinitely — exactly the kind of “narrow at the micro end” edge that erodes when low-cost entrants (and, increasingly, AI-native energy retailers) attack. (INTERPRETATION — Greenwald framework applied to disclosed retail metrics.)

The fuel-mix disadvantage versus VST and CEG. This is the decisive competitive fact. The most-prized contracting vector in the AI-power era is the 20-year hyperscaler PPA on zero-carbon baseload nuclear — Constellation’s Three Mile Island restart for Microsoft, Vistra’s up-to-6.6 GW nuclear arrangement with Meta. NRG sold its only nuclear asset (the STP stake) in 2023 and is therefore structurally shut out of that vector. Its pitch is instead flexible, fast gas + “bring-your-own-power” co-location + the CPower commercial-and-industrial virtual-power-plant — real capabilities, but commoditized and lower-margin than a nuclear annuity. NRG is the nuclear-short member of the AI-power trio. (FACT — STP sale 8-K Nov-2023; competitor PPA announcements.)

Direct comparison.

• vs. VST (Vistra) — the closest peer (factor similarity 0.96): both are integrated gen+Texas-retail merchants. VST is ~3x NRG’s generation by some measures, owns ~6.4 GW of nuclear (with the federal 45U production-tax-credit floor) and has signed more hyperscaler PPAs. NRG is smaller in generation, larger in retail customers, and carries the Vivint home-services optionality VST lacks. NRG is cheaper (~8.6x vs ~10x EV/EBITDA) for these reasons.
• vs. CEG (Constellation) — ~21 GW of nuclear, a federally-backstopped baseload annuity; trades at a deserved premium (~14x). NRG is the fossil/retail opposite end.
• vs. regulated utilities (AEP, DUK, SO, XEL, etc.) — NRG has no rate base, no allowed return, far higher beta (1.48), and far higher commodity/weather risk, but also far higher torque to the demand inflection and a buyback regulated utilities cannot match.

Verdict (competitive position): A narrow, fuel-disadvantaged moat — an integrated retail+gas cost/hedging advantage that is durable-ish but commoditized, structurally short the nuclear vector that defines the premium end of the industry. NRG is a competent operator and aggressive capital-returner in a no-barriers commodity industry, riding a favorable cycle — not the owner of a wide, durable franchise.

5. Growth History and Forward Opportunities

Historical growth — acquisition-driven, then disciplined. Revenue scaled from $9.1B (2020) to $27.0B (2021, the Direct Energy acquisition) to $31.5B (2022), then settled ~$28-31B (2023-25) — i.e., the top line has been roughly flat for four years once Direct Energy lapped; revenue is a poor growth gauge for a commodity pass-through business. The meaningful growth has been in Adjusted EPS and FCF-before-growth per share, manufactured by margin/optimization, the Vivint addition, and an aggressive buyback: Adjusted EPS reached $8.24 in FY2025 (+21% YoY), the third straight beat-and-raise. (FACT — company adjusted results; ROIC.ai revenue series.)

The forward plan. At the Q4-2025 call (2026-02-24) NRG re-based and raised its long-range plan: “at least 14%” compound annual growth in BOTH Adjusted EPS AND FCFbG-per-share, 2026 → 2030 — taking Adjusted EPS from the $8.90 (2026 guide) base to >$14 by 2030, and FCFbG/share from ~$14.50 to >$22 by 2030. This was raised from “at least 10%” when the LS Power deal was announced (May-2025). (FACT — Q4-2025 call; investor materials.)

Three critical caveats on that plan. (1) It assumes ~80% organic / ~20% buyback-driven growth. (2) It assumes flat power and capacity prices — specifically that the next two PJM capacity auctions clear at the ~$325/MW-day cap (an assumption, not a result; a downside risk if reform lowers it). (3) Crucially, it excludes any new data-center deals and any CT-to-CCGT conversions — i.e., the headline AI-power upside is on top of the ≥14% plan, not in it. This is double-edged: it means the base plan does not require the data-center deals to deliver double-digit growth (a real positive), but it also means the entire equity premium over a ~15x merchant multiple rests on optionality the plan itself does not count. (FACT — management guidance, framed as hypothesis pending validation.)

The data-center opportunity — signed vs. reserved vs. aspiration. This is the engine of the stock’s re-rating and the single most important thing to size honestly:

• SIGNED (hard reality): ~445 MW of data-center load (a ~295 MW agreement Aug-2025 + ~150 MW in PJM), at pricing above the prior $70-90/MWh range; management has raised its new-deal price target to >$80/MWh (>$90-95 for new turbine builds).
• RESERVED equipment: a 5.4 GW GE Vernova + Kiewit turbine reservation (first ~1.2 GW — two 7HA units — targeting ~2029 commercial operation) plus up to ~2 GW of LS-fleet uprates/conversions = “>6 GW reservable.”
• PIPELINE: ~5.4 GW of non-binding LOIs/joint-development agreements (+35% q/q in Q3-2025).
• ASPIRATION: >6 GW of contracted load ⇒ management’s “>$2.5B of recurring annual EBITDA” on up-to-20-year contracts — explicitly NOT in guidance.
• STATUS (Q1-2026, 5/6): ZERO incremental gigawatt-scale signings after ~18 months of “active discussions.” New CEO Gaudette identifies the bottleneck as infrastructure/interconnection, not price, and frames 2026 as needing a first ≥1 GW “bring-your-own-power” signing to hit 2029 commercial-operation dates. (FACT — earnings transcripts; flagged as the thesis pivot.)

Other growth vectors. ~1.5 GW of new ERCOT gas under Texas Energy Fund 3%-loans (first ~0.4 GW at TH Wharton online ~mid-2026); the CPower C&I virtual-power-plant from LS Power; Vivint subscriber growth (~+9%/yr) and energy/home cross-sell; and the LS Power gas fleet’s PJM capacity-auction torque.

Verdict (growth quality): Mixed-to-good. The ≥14% per-share base-plan growth (flat-price, buyback-supported, deal-excluded) is genuinely high-quality and does not depend on the AI narrative. But the premium growth — the >$2.5B data-center EBITDA — is at present 445 MW of signed reality dressed in 6 GW of equipment reservations and non-binding letters; high-potential, low-proven. Growth is real; the headline growth is mostly optionality.

6. Financial Quality

GAAP earnings are noise — value NRG on adjusted cash metrics. NRG’s GAAP net income swung from +$2,187M (2021) to +$1,221M (2022) to −$202M (2023) to +$1,125M (2024) to +$864M (2025) — driven overwhelmingly by non-cash mark-to-market on commodity derivative hedges (and, in 2023, Texas-storm and impairment items), not by the operating business, which produced rising Adjusted EBITDA across the same span. The trailing GAAP P/E of ~112x (99th percentile of NRG’s own history) is therefore a meaningless artifact and must be discarded; the decision-useful metrics are Adjusted EBITDA, Adjusted EPS, and FCFbG. (FACT — ROIC.ai income statements; standard IPP treatment; the same issue applies to Vistra.)

Adjusted earnings power and trajectory.

(Adjusted EBITDA/EPS/FCFbG are company-defined; GAAP figures per ROIC.ai. 2026E reflects ~11 months of LS Power.)

Margins. Reported gross margin is low and commodity-distorted (~19% FY2025) because revenue is mostly pass-through; the operating signal is that Adjusted EBITDA rose ~17% in 2025 on roughly flat revenue — operating leverage from the integrated optimization and Vivint. (FACT — ROIC.ai; company results.)

Returns on capital — the quality red flag. This is where the franchise weakness shows. ROIC.ai’s return-on-invested-capital reads ~25% (2021) → 13.6% (2022) → negative (2023) → 13.1% (2024) → 9.4% (2025); return-on-capital ~8.8% (2025). The reported ROE (~45-90%) is not a quality signal — it is inflated by a thin, buyback-shrunken, negative-tangible-book equity base (tangible book value per share ~−$32 in 2025; goodwill+intangibles ~$12.3B exceed total equity of ~$1.7B). Stripping the M&A goodwill, the core retail+generation unit economics are better than 9%, but the consolidated, capital-deployed return has fallen to roughly the cost of capital as serial debt-funded acquisitions ballooned the denominator — the defining quality issue (and the precise risk the comp plan, with no ROIC metric, fails to police). (FACT — ROIC.ai profitability ratios; balance sheet.)

Cash flow. GAAP operating cash flow is volatile due to collateral/derivative working-capital swings (−$221M in 2023, +$2,306M in 2024, +$1,913M in 2025); the company’s adjusted FCFbG strips these and is the cleaner read — ~$2.21B FY2025, guided ~$3.05B FY2026 (~$14.50/sh). Maintenance capex is modest (the fleet is largely existing assets), which is why FCF conversion is high; growth capex (new turbines) is the back-end-loaded swing item. (FACT — cash-flow statements; guidance.)

Balance sheet. Total debt rose $8.3B (2020) → $11.0B (post-Vivint 2023) → $16.6B (end-2025, pre-positioned for LS) → ~$18-19B pro-forma after the LS Power close and an April-2026 $3.5B refinancing; net debt ~$11.7B end-2025 and rising. Interest expense was ~$741M in 2025 and ran ~$285M in Q1-2026 alone (annualizing >$1.1B) — a real near-term EPS drag. Net-debt/EBITDA targets ~3.0x (up from ~2.5x pre-LS). Corporate credit is sub-investment-grade: S&P BB / Moody’s Ba1 / Fitch BB+ (all Stable; senior secured is IG at BBB−/Baa3). Liquidity is ample, but NRG is not an IG corporate issuer — a material distinction from VST (BBB−) and CEG. (FACT — 10-K; rating-agency actions; LS Power filings.)

Verdict (financial quality): A genuine, high-yielding cash machine (~10.7% FCFbG yield, rising) with strong operating-leverage and a real volatility hedge — undercut by (a) noise-dominated GAAP optics, (b) a consolidated ROIC that has fallen to ~the cost of capital on a goodwill-bloated, negative-tangible-book balance sheet, and © sub-IG leverage re-extended by LS Power. The cash is real; the return on the capital deployed to generate it is mediocre and deteriorating.

7. Capital Allocation

A genuine return-of-capital culture. Post the 2020 “transformation” and the 2023 Elliott campaign, NRG has been an aggressive, credible capital-returner. Share count fell from 244.2M (2020) to 190.4M (2025), −22% (~$4.2B of buybacks 2021-25), and the dividend grew from $1.20 (2020) to $1.76 (2025) to $1.90 (2026, +8%), on a stated 7-9%-annual-growth target and a ~43% payout. A new $3.0B buyback authorization (Oct-2025) supports a long-range plan of ~$11B of repurchases 2026-30. The stated framework routes ~80% of excess cash to shareholders / ~20% to growth, after debt service. (FACT — proxy; cash-flow statements; capital-allocation disclosures.)

But the buyback is pro-cyclical. The dollars were spent heaviest near the highs: 2025’s program averaged ~$129/share, and Q4-2025 repurchases ran at ~$166/share — near the $183 ATH — while the lightest buying was at the 2022 lows (~$35). This is the opposite of value-accretive timing (contrast VST, which retired ~28% of its shares at an average cost under $36). NRG is returning cash credibly but not buying its stock cheaply. (FACT/INTERPRETATION — buyback disclosures vs. price history.)

M&A track record — two bad, one good, one TBD. (1) Direct Energy (~$3.6B, Jan-2021): a good deal — scaled the retail book, integrated cleanly. (2) Vivint Smart Home (~$2.6B cash / ~$5.2B EV, Mar-2023): value-destructive on entry — the stock fell ~20% on announcement, it added ~$2.5B of goodwill, it was off-strategy, and it directly triggered the Elliott campaign; operationally it has outperformed the worst fears (~$1.09B segment EBITDA, 2.42M subs) but has never been vindicated on the price paid, and the cross-sell thesis remains largely unproven. (3) LS Power (~$12B EV, closed Jan-30-2026): disciplined-leaning on price (~$900/kW vs >$2,000/kW replacement cost, immediately accretive, with an OB3 100%-bonus-depreciation cash-tax windfall) but re-levering and partly dilutive — financed with $6.4B cash + 24.25M new shares + ~$3.2B assumed debt, the equity issuance reversing ~3 years of buyback (190M → ~213M shares) and re-opening the empire-building tension Elliott tried to close. (FACT — deal 8-Ks; company financing disclosures.)

Incentive alignment — the key governance flag. The annual incentive plan weights Adjusted FCFbG 45% + Adjusted EBITDA 45% + non-financial 10%; the long-term plan is 67% relative-TSR performance shares + 33% time-vested RSUs. There is no ROIC metric and no per-share-EPS metric in the plan — “return on invested capital” appears only in permitted-metric boilerplate. The plan therefore pays for absolute EBITDA/FCF size — exactly the incentive that rewards debt-funded, goodwill-heavy, scale-adding M&A even as ROIC falls (as it has, 25% → 9%). Relative-TSR is a partial offset (it rewards beating peers), but the absence of any capital-efficiency or per-share gate is precisely the structure a skeptic would flag on a serial acquirer. CEO (Coben) FY2025 total comp was ~$21.8M (up from ~$12.4M in 2023), and say-on-pay support fell to ~87% (2026) from ~96% (2025) — a ~9-point slide signaling shareholder pushback. (FACT — 2026 DEF 14A.)

Insider behavior — neutral-to-mildly-negative. Across ~473 Form 4s over five years, there were only 7 open-market purchases (code P), all clustered at the December-2022 trough (~$31-32) — including future-CEO Coben (15,000 sh @ $31.70) and then-CEO Gutierrez (15,000 @ $32.03) buying the Vivint-drop bottom — totaling ~$1.55M. There has been zero open-market insider buying since December 2022, across the entire $120 → $183 → $135 cycle. The March-2026 “sells” were LS Power’s secondary, not operating insiders. The signal: management bought the 2022 bottom with conviction and has not signaled value at any point since. (FACT — Form 4 corpus.)

Verdict (capital allocation): Mixed. A real, credible return-of-capital culture and a disciplined-on-price LS Power deal sit against pro-cyclical buyback timing, a value-destructive Vivint acquisition, a re-leveraging/re-diluting acquisition appetite, an incentive plan that ignores ROIC and per-share metrics, and zero insider conviction since 2022. Management allocates capital actively and shareholder-friendly-ly, but not always intelligently on a per-share, returns-on-capital basis.

8. Changes and Headwinds — Last Two Years

Leadership transition (major). After the Elliott-driven 2023 change (Mauricio Gutierrez out; Chairman Larry Coben to interim then permanent CEO), NRG is now in a second CEO transition: Coben steps down at the April-30-2026 annual meeting and is succeeded by Robert Gaudette (named Jan-6-2026; President immediately, CEO effective Apr-30). Gaudette is a ~20-year NRG insider who ran the commercial/C&I/wholesale platform behind the data-center strategy — signaling continuity with more emphasis on contracted, long-duration cash flows and openness to partnering with regulated utilities. CFO Bruce Chung stays. Board churn continued (Howell resigned; Abraham not re-standing; Kapoor and Wright added). NRG also dual-listed on NYSE Texas. (FACT — 8-Ks Jan-2026; proxy.)

LS Power closed (Jan-30-2026). The ~$12B acquisition of 18 gas/dual-fuel plants (~13 GW, mostly PJM) + the CPower C&I VPP closed on time, doubling the fleet to ~25 GW and >75% gas. All approvals (HSR/FERC/NY PSC) obtained; financing completed Sept-Oct 2025; management calls it “immediately accretive, exceeding underwriting.” An April-28-2026 $3.5B refinancing retired the acquisition bridge (“Lightning” notes) toward the 3.0x leverage target. (FACT — 8-Ks; company results.)

The de-rate (Feb → Jun 2026). From the $183.47 ATH the stock fell ~26% on a confluence: (a) LS Power’s March 2-4 secondary of 14.3M shares (NRG got no proceeds; the stock fell ~8% on pricing) — a technical overhang; (b) a genuine Q1-2026 earnings miss (Adjusted EPS $1.49 vs ~$1.78 consensus and $2.68 prior-year) on a mild Texas winter (heating-degree-days −30%), low volatility, and higher interest expense ($285M); and © a broad AI-power-complex unwind (Vistra, Constellation, GE Vernova all sold off together) — NRG’s 1.48 beta amplified it. Importantly, FY2026 guidance was reaffirmed throughout — the de-rate is best read as sector/momentum + technical + one soft-weather quarter, not a fundamental thesis break. (FACT — transcripts; price history; news feed.)

Other. STP nuclear stake sold ($1.75B, Nov-2023); Airtron divested (2024); Indian River coal retired; Rockland ERCOT gas acquired (Q2-2025); exploring a Gladstone (Australia) sale; PJM 2024/2025 capacity auctions cleared high (a multi-year East tailwind).

Verdict (changes): Net neutral-to-slightly-cautionary. LS Power and the capacity-auction tailwind strengthen near-term cash flow; the second CEO transition in three years, the re-leveraging, the secondary overhang, and the data-center signings still not arriving are offsetting cautions. The thesis is intact but the easy-money phase (discipline-driven re-rating) is over.

9. Risk Analysis (Risk Matrix)

Total-loss risk: negligible. NRG generates ~$3B of FCFbG, holds hard generation and retail assets, and has ample liquidity; the realistic downside is a multiple/EBITDA de-rate (a value trap), not impairment. The dominant thesis risks are the unproven data-center conversion and ERCOT/PJM price mean-reversion.

10. Valuation Discussion (Embedded Expectations)

Frame the multiples correctly. At ~$135 (Jun-18-2026), ~213M pro-forma shares, market cap ~$28-29B, and pro-forma net debt ~$18-19B, enterprise value is ~$46-48B. Against FY2026 guidance (Adjusted EBITDA ~$5.575B, Adjusted EPS ~$8.90, FCFbG ~$3.05B / ~$14.50/sh):

On its own ten-year history, NRG sits at the ~80th-percentile composite valuation, but the components matter: P/E 99th percentile is a GAAP-noise artifact (ignore); P/B ~74th and P/S ~68th are the meaningful reads — elevated but not richest-ever (contrast Vistra at a higher percentile of its own history). NRG is fully valued versus its own pre-AI history, fairly-to-cheaply valued versus its AI-power peers, and cheap only versus the fantasy multiples of early 2025. (FACT — AZI valuation_index percentiles; ROIC.ai multiples; the author VST/CEG reports.)

Embedded-expectations analysis — what the ~$135 price underwrites. A ~15x forward P/E / ~8.6x EV/EBITDA / ~10.7% FCFbG yield on a sub-IG, no-moat merchant is not demanding in absolute terms — it roughly prices the ≥14%-CAGR base plan (Adj EPS $8.90 → >$14 by 2030) succeeding, with the data-center upside as a near-free option. Decompose it: a pure merchant with flat prices and a buyback might warrant ~7-9x EV/EBITDA (where VST/NRG historically traded); the ~8.6x NRG multiple plus the AI-power-complex association implies the market is paying a modest premium for the data-center optionality and the demand-inflection tailwind — far less than the ~14-15x it was paying at the $183 peak. The market is underwriting correctly that NRG is a cash machine growing FCFbG ~14% on a flat-price plan; it is leaving the >$2.5B data-center EBITDA as upside it no longer fully pays for (the right posture after 18 months of no signings). The risk is not that the base plan is mispriced high — it is that the flat-price assumption itself (ERCOT margins, PJM cap) re-bases lower, taking the base plan’s EBITDA and the multiple down together.

Scenarios (illustrative; not price targets).

• Bear (~30%): ERCOT/PJM forwards re-base lower (TEF gas + batteries + a PJM cap rollback), data-center deals keep slipping with zero gigawatt signings, LS Power leverage/integration strains the plan; Adjusted EBITDA stalls/declines and the multiple compresses toward a pre-AI ~6-7x merchant level. The ~10.7% FCFbG yield and the buyback cushion the equity, so this is a de-rate to the ~$95-115 zone, not a collapse — a value trap, not impairment.
• Base (~50%): the ≥14% flat-price plan roughly delivers (Adj EPS $8.90 → low-double-digit growth, FCFbG ~$14.50 → ~$16-17/sh), PJM capacity holds, one or two mid-sized data-center deals sign but not the transformational gigawatt slate; the multiple holds ~8-9x EV/EBITDA. Fair value roughly ~$130-165 — i.e., the stock today sits near the low end of its own base case, with the buyback compounding per-share value.
• Bull (~20%): NRG signs its first ≥1 GW BYOP data-center contract at disclosed >$90/MWh economics in 2026-27, validating the >$2.5B-EBITDA pipeline; ERCOT/PJM hold or firm; the option converts to a contracted annuity and the stock re-rates toward a contracted-IPP multiple. ~$190-240+ (back above the prior ATH).

Verdict (valuation): Fair-with-an-option. The euphoria premium is gone; at ~10.7% FCFbG yield the market pays a modest premium for real demand-tailwind optionality it no longer fully credits. No margin of safety for the bear’s flat-price-re-base scenario, but a reasonable price for the base case and a cheap-ish call option on the data-center upside.

11. Variant Perception

Consensus. The sell-side and the market broadly hold NRG as a high-quality way to play AI-driven power demand with the cheapest multiple of the three merchant names (NRG < VST < CEG), a strong buyback, and a credible ≥14% growth plan — with the data-center pipeline as the call option. Consensus is constructive but has clearly cooled (the ~26% de-rate, the say-on-pay slide, the soft Q1).

The factor-positioning read (the tape). NRG is a high-beta (1.48), high-alpha (+0.32) Utilities-sector vehicle that loads heavily on the Utilities sector factor (~1.45) and DividendYield (~1.45) and on Market (~1.22-1.29) — i.e., it trades as a levered, rate-sensitive utility and as the higher-torque sibling of VST (factor similarity 0.96) and CEG (0.89). The track record shows the trade clearly: y3 return +61% annualized (Sharpe 1.35) — a violent three-year up-trend — now in a reset: y1 −10.5%, m6 −17.3%, m3 −25.4% (all annualized, negative Sharpes), with the stock ~26% off its peak. This is a momentum unwind / mean-reversion off a euphoric high, not a falling knife on broken fundamentals (guidance reaffirmed) and not an abandoned-value name (still ~68-74th percentile own valuation). The factor model says: a crowded AI-power momentum trade that has partly de-crowded — consensus is no longer euphoric but not yet capitulated. (FACT — FactorsToday loadings/leaderboard; AZI price CSV.)

Strongest bull case. A genuinely cash-generative, integrated, hedged merchant at ~10.7% FCFbG yield, growing FCFbG/share ~14% on a flat-price, deal-excluded plan, with ~$11B of buybacks ahead and a free-ish option on >$2.5B of incremental data-center EBITDA in the tightest power market in a generation. If even one transformational hyperscaler deal signs, the stock re-rates hard — and you’re paying a merchant multiple for it today.

Strongest bear case. A no-moat, sub-IG, nuclear-less commodity merchant whose consolidated ROIC has fallen to ~the cost of capital on a goodwill-bloated, negative-tangible-book balance sheet; whose comp plan rewards size not returns; whose buyback is pro-cyclical; whose insiders haven’t bought since 2022; whose single largest market (ERCOT, energy-only) is the fastest to mean-revert as TEF-gas and batteries flood in; and whose entire premium prices a data-center prize that is 445 MW signed against 18 months of talk. The flat-price plan assumption (PJM at the cap, ERCOT margins) is the load-bearing beam, and it bends down in the bear case — stripping EBITDA and multiple at once.

The 3-5 assumptions that matter most: (1) Do the gigawatt-scale data-center contracts actually sign, and at the disclosed >$90/MWh economics? (2) Do ERCOT energy margins and the PJM capacity cap hold (the flat-price plan), or mean-revert as supply arrives? (3) Does LS Power integrate accretively and deleverage to ~3.0x on schedule? (4) Does management resist the next empire-building deal and keep returning cash per-share-accretively? (5) Does the AI-power demand inflection prove durable or partly speculative (the ERCOT large-load queue conversion rate)?

What would falsify each side. Bull falsified by: another full year (into 2027) of zero gigawatt-scale data-center signings + softening ERCOT/PJM forwards. Bear falsified by: a signed ≥1 GW BYOP contract at premium economics + PJM capacity clearing at/near the cap + visible LS deleveraging — converting optionality to annuity and validating the plan.

12. Fact vs. Interpretation Table

13. Open Questions

1. When (if ever) does the first ≥1 GW data-center contract sign, and at what disclosed economics? The single most thesis-determining unknown.
2. Where do ERCOT energy margins and the next two PJM capacity auctions actually clear versus the plan’s flat-price / at-the-cap assumptions?
3. What is the true cash-tax and accretion profile of LS Power once the OB3 bonus-depreciation windfall is fully booked, and does net-debt/EBITDA reach ~3.0x on schedule?
4. Has the Vivint cross-sell produced any quantifiable energy/home revenue or VPP economics, or is a goodwill write-down risk building?
5. Will CEO Gaudette change the capital-allocation posture — more contracted cash flow, less opportunistic M&A, and will the comp plan add a ROIC/per-share metric?
6. How much of the ERCOT >36 GW large-load interconnection queue is real versus speculative duplicate requests?
7. Does the pro-cyclical buyback continue at elevated prices, or does management buy the current weakness more aggressively?

14. What Must Be True

Bull case — what must be true: (1) the demand inflection is durable and NRG converts equipment reservations/LOIs into signed, long-dated, gigawatt-scale data-center contracts at premium economics; (2) ERCOT energy margins and PJM capacity prices hold at least at plan-assumed (flat) levels; (3) LS Power integrates accretively and the balance sheet deleverages to ~3.0x; (4) the buyback compounds per-share value and management resists value-destructive M&A. Falsification test: another full year (through 2027) of zero gigawatt-scale data-center signings combined with softening ERCOT/PJM forwards falsifies the bull — the premium prices an option that is not converting while the base plan’s price assumptions erode.

Bear case — what must be true: (1) the data-center pipeline stays narrative (445 MW signed, no transformational deal); (2) ERCOT/PJM forwards re-base lower as Texas-Energy-Fund gas and batteries flood supply and PJM caps are trimmed; (3) leverage/integration and rising interest expense pressure FCFbG; (4) the no-ROIC comp plan drives further dilutive empire-building. Falsification test: a signed ≥1 GW BYOP contract at disclosed >$90/MWh economics, plus PJM capacity clearing at/near the cap, plus visible LS deleveraging falsifies the bear — it converts optionality to a contracted annuity and validates the ≥14% plan, justifying a re-rating.

The pivot: both cases converge on one question — does the data-center prize convert from reservations to signed contracts before the favorable, flat-price power-market window the plan assumes begins to mean-revert? That is unknowable until the 2026-27 contracting season resolves.

15. Source Appendix

See the Source Appendix below for the full citation list. Primary sources: NRG FY2025 Form 10-K and Q1-2026 Form 10-Q (SEC EDGAR, CIK 0001013871); NRG 8-Ks (LS Power acquisition/close, CEO transition, secondary offering, refinancing); 2026 DEF 14A proxy; Q2-2025 through Q1-2026 earnings call transcripts and press releases; ROIC.ai financial statements and ratios (reconciled to filings); AZI valuation-percentile data and price history; FactorsToday factor model; PJM capacity-auction results; NERC Long-Term Reliability Assessment; rating-agency actions (S&P/Moody’s/Fitch); and the author internal peer reports on VST (2026-06-13) and CEG (2026-06-12) for industry cross-read. Every non-obvious fact in the body is traceable to a cited primary source.

APPENDIX A — Standard Diligence Questionnaire — NRG Energy, Inc. (NYSE: NRG)

Supplemental to the research memo. As-of 2026-06-20. Fact/Interpretation/Assumption labels applied where material. Where a question does not map to NRG’s business model, the correct sector analog is given.

General

What thoughtful questions have other investors asked about this company?

• Is the data-center pipeline (>6 GW reserved, >$2.5B aspirational EBITDA) real and contractable, or narrative? After ~18 months only ~445 MW is signed — the central debate. (Fact/Interp)
• Was the 2023 Vivint acquisition value-destructive, and does the cross-sell thesis exist? (Elliott called it “the worst deal in the sector in a decade.”) (Fact)
• Is the LS Power deal disciplined consolidation or a return to empire-building after Elliott? It re-levered the balance sheet and reversed ~3 years of buyback via a 24.25M-share issuance. (Fact/Interp)
• Why does the comp plan have no ROIC metric on a serial acquirer whose ROIC fell 25%→9%? (Fact)
• Is NRG disadvantaged versus VST/CEG by owning zero nuclear (shut out of 20-year hyperscaler PPAs)? (Fact)

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Closer to a favorable mid-cycle. Adjusted EBITDA/EPS are at records (FY2025 Adj EPS $8.24, +21%), benefiting from a tight power market (high PJM capacity prices, demand inflection). The base-plan assumes flat prices forward — so earnings are elevated but not assumed to grow from price; the risk is mean-reversion of ERCOT/PJM margins, not a cyclical low. (Interpretation)

Driven by the external environment or internal actions? Both: internal (buyback, Vivint/LS Power scale, optimization) and external (power/capacity prices, weather, demand inflection). Q1-2026’s miss was external (mild Texas weather). (Fact)

How stable are revenues? Reported revenue (~$28-31B) is commodity-pass-through and volatile/uninformative; the adjusted earnings stream is more stable thanks to the integrated retail/generation hedge, but still weather- and price-exposed. (Fact/Interp)

Outlook for products/services? Electricity demand is inflecting up after ~20 years flat (data centers, electrification, reshoring); NRG’s dispatchable gas + retail + smart-home is well-positioned, with the data-center upside the key swing. (Interpretation)

How big will this market be — growing, shrinking, domestic or international? U.S.-centric (plus Canada retail; small Australia generation exploring sale). U.S. peak demand forecast up ~69% over a decade in NERC’s high case — a structurally growing end market for the first time in a generation. (Fact/Interp)

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Consolidating at the asset level (NRG-LS Power, Constellation-Calpine, Vistra-Cogentrix) but retail remains competitive with low switching costs. Net: supply-constrained generation (favorable) + competed retail (unfavorable). (Interpretation)

How profitable is the business (ROIC, ROE)? Consolidated ROIC ~9% (2025), down from ~25% (2021) — ~the cost of capital, dragged by goodwill-heavy M&A. ROE (~45-90%) is not a quality signal — it reflects a thin, negative-tangible-book, buyback-shrunken equity base. (Fact)

How profitable is the industry — competitors, barriers to entry? Historically a bad, no-barriers commodity industry (repeated bankruptcies); currently enjoying supranormal returns from a supply/demand window. Barriers to new generation are temporarily high (turbines sold out to ~2030, clogged queues) — a cyclical, not structural, barrier. (Interpretation)

Can the business be easily understood? Moderately — the integrated model and retail are clear, but GAAP earnings are obscured by derivative mark-to-market, requiring adjusted-metric analysis. (Fact)

Can it be undermined by foreign low-cost labor? No — power generation/retail is inherently domestic. (Fact)

Do brands matter? Modestly in retail (Reliant, Green Mountain, Direct Energy carry recognition and a cost-to-serve edge), but low switching costs cap brand value. (Interpretation)

Nature of competition? Price/margin competition in retail; marginal-cost dispatch competition in wholesale generation; relationship/economics competition for data-center load (vs VST, CEG, regulated utilities). (Fact)

Customers’ switching costs? Low in retail electricity (days to switch in Texas); higher for Vivint smart-home (equipment + contracts, ~90% retention). (Fact)

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? Brand/customer relationships partly in intangibles; the value of the integrated hedge and data-center optionality is not on the balance sheet. (Interpretation)

Off-balance-sheet liabilities? Standard for the sector — collateral postings on derivatives (large working-capital swings), purchase/transmission obligations, decommissioning/environmental (coal), operating leases. (Fact)

How conservative is the accounting? Mixed — non-cash hedge mark-to-market makes GAAP earnings noisy; Vivint’s SAC-amortization reclass into D&A (Q3-2024) flattered segment EBITDA optics. Adjusted metrics are the disclosed standard but company-defined. (Fact/Interp)

How CapEx-hungry? Maintenance capex is low (existing fleet) — high FCF conversion. Growth capex (new turbines for data centers) is the back-end-loaded swing item and is excluded from FCFbG. (Fact)

Capital Allocation & Management

How much FCF, and how is it used? FCFbG ~$2.21B (2025), guided ~$3.05B (2026, ~$14.50/sh). Framework: ~80% to shareholders (buyback + 7-9%-growing dividend) / ~20% growth, after debt service; ~$11B of buybacks planned 2026-30. (Fact)

Significant acquisitions recently? Yes — Vivint (~$5.2B EV, 2023, value-destructive on entry) and LS Power (~$12B EV, closed Jan-2026, disciplined-on-price but re-levering/diluting). Direct Energy (2021) was the good one. (Fact)

Buying back shares? Yes, aggressively — share count 245M→190M (2020-25), ~$4.2B spent — but pro-cyclically (heaviest near the $166-183 highs, lightest at the $35 lows). The LS Power 24.25M-share issuance partly reversed it. (Fact)

Issuing large amounts of stock to insiders? SBC is modest (~$134M, 2025). The large issuance was the LS Power deal consideration, not insider grants. (Fact)

Compensation policy? AIP: Adj FCFbG 45% / Adj EBITDA 45% / non-financial 10%. LTIP: 67% relative-TSR PSUs / 33% time RSUs. No ROIC, no per-share-EPS metric — the key governance flag. CEO (Coben) FY2025 comp ~$21.8M; say-on-pay slid to ~87% (2026). (Fact)

Motivations of management? Return-of-capital-oriented and growth-acquisitive; the incentive design rewards EBITDA/FCF size over capital efficiency. Insiders bought the Dec-2022 trough but have not bought since. CEO transition to Gaudette (Apr-2026) may shift toward contracted cash flow. (Fact/Interp)

Valuation & Market Data

ADR, MLP, or K-1 issuer? No — a U.S. C-corporation (Delaware), NYSE-listed (secondary NYSE Texas), issues a Form 1099, not a K-1. (Fact)

Dividend policy? Yes — $1.90/share annualized (2026), ~1.4% yield, ~43% payout, targeted 7-9% annual growth. The buyback, not the dividend, is the primary return lever. (Fact)

How profitable is the business? Strong cash profitability (~10.7% FCFbG yield) but mediocre return on capital (ROIC ~9%). (Fact)

Is net income diverging from cash from operations? Yes, materially and routinely — GAAP net income is dominated by non-cash derivative mark-to-market; operating cash flow swings on collateral/working-capital. Use Adjusted EBITDA and FCFbG. (Fact)

Risks & Downside

What would cause the stock to decline? Data-center deals failing to sign; ERCOT/PJM margin mean-reversion (the flat-price plan breaking down); a soft-weather quarter; rising rates (beta 1.48); LS Power integration/leverage strain; a further dilutive acquisition; PJM capacity-cap rollback. (Interpretation)

Risk of catastrophic loss? Low — operational tail risks (a Uri-style grid event, a major outage, an environmental liability) exist but are insurable/manageable; the franchise is profitable and asset-backed. (Interpretation)

Chance of a total loss? Negligible — ~$3B FCFbG, hard assets, ample liquidity, sub-IG but well-laddered debt. The realistic bad case is a de-rate/value-trap, not impairment. (Interpretation)

Recent News & Events

Has the business environment changed recently? Yes — LS Power closed (Jan-2026, fleet doubled); CEO transition (Gaudette succeeds Coben Apr-2026); a ~26% de-rate off the Feb-2026 ATH on a secondary overhang + Q1 miss + AI-power-complex unwind; FY2026 guidance reaffirmed throughout. (Fact)

Significant acquisitions? LS Power (~$12B, ~13 GW gas + CPower VPP), closed Jan-30-2026. (Fact)

Change in accounting policies? The 2024 Vivint SAC-amortization reclass into D&A is the notable item. (Fact)

Recent changes — new markets, facilities, management? New PJM/Northeast generation footprint (LS Power); new ERCOT gas under Texas Energy Fund (first ~0.4 GW mid-2026); NYSE Texas dual-listing; second CEO change in three years; board refresh. (Fact)

APPENDIX B — Source Appendix

As-of 2026-06-20. Primary sources first. Third-party aggregated data (ROIC.ai, AZI, FactorsToday) reconciled to filings; the filing governs where they disagree.

Primary — SEC filings (EDGAR, CIK 0001013871)

• Form 10-K, FY2025 (filed ~2026-02-24) — business description, segment tables (Texas/East/West/Vivint/Corporate), generation fleet, retail customer counts, risk factors, MD&A, financial statements. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001013871
• Form 10-Q, Q1-2026 (filed ~May 2026) — Q1 results, post-LS-Power balance sheet, interest expense, segment detail.
• 8-K — LS Power acquisition announcement (~May 2025) and closing (~Jan-30-2026) — ~$12B EV terms ($6.4B cash + 24.25M shares + ~$3.2B assumed debt), ~13 GW / 18 plants + CPower VPP, fleet → ~25 GW.
• 8-K — CEO transition (~Jan-6-2026) — Robert Gaudette named President (CEO effective Apr-30-2026); Larry Coben departure.
• 8-K — LS Power secondary offering (~Mar 2026) — 14.3M-share secondary (NRG no proceeds).
• 8-K — refinancing (~Apr-28-2026) — $3.5B refinancing retiring acquisition bridge notes.
• 8-K — STP nuclear sale (~Nov-2023) — 44% South Texas Project stake sold to Constellation for $1.75B.
• 8-Ks — quarterly earnings releases/guidance — Q2-2025, Q3-2025, Q4-2025, Q1-2026.
• DEF 14A proxy (2026) — compensation metrics (AIP: Adj FCFbG 45% / Adj EBITDA 45% / non-financial 10%; LTIP 67% rTSR / 33% RSU; no ROIC/EPS metric), CEO comp, say-on-pay (~87%), board changes.
• Form 4 corpus (~473 filings, 2021-2026) — insider transactions; 7 open-market buys all Dec-2022 trough (~$31-32, incl. Coben, Gutierrez), zero since.

Primary — earnings call transcripts

• Q1-2026 earnings call, 2026-05-06 — EPS miss ($1.49), FY2026 guidance reaffirmed, data-center signing status, Gaudette commentary.
• Q4/FY2025 earnings call, 2026-02-24 — record FY2025, LS Power close, 2026→2030 ≥14% plan unveiled, FY2026 guidance.
• Q3-2025 (2025-11-06), Q2-2025 (2025-08-06) — data-center pipeline, LS Power progress.

Industry / regulatory / market

• NERC Long-Term Reliability Assessment — U.S. peak-demand growth forecast.
• PJM capacity (Base Residual Auction) results, 2024 & 2025 — clearing prices near the administrative cap.
• ERCOT / Texas — Senate Bill 6 (large-load interconnection/curtailment, signed June 2025); Texas Energy Fund (3% loans); ERCOT large-load interconnection queue.
• Competitor disclosures — Constellation–Microsoft (TMI restart) and Vistra–Meta nuclear PPAs (the nuclear-vector context); Constellation–Calpine and Vistra–Cogentrix/Lotus deals (consolidation context).

Quantitative data services (reconciled to filings)

• ROIC.ai MCP — income statement, balance sheet, cash flow, profitability ratios (ROIC/ROE/margins), enterprise value, valuation multiples, per-share data, earnings-call list (FY2020-2025 + TTM).
• AZI (azitrading.com) — valuation_index own-history percentiles (composite ~80th; P/E 99th [GAAP artifact], P/B ~74th, P/S ~68th; price $135.06, 2026-06-18); 5-year daily price CSV (ATH $183.47 close 2026-02-24; 52wk $120.65-$183.47).
• FactorsToday (factorstoday.com) — factor loadings (Utilities ~1.45, DividendYield ~1.45, Market ~1.22-1.29; beta 1.48, alpha +0.32), leaderboard (y3 +61% ann./Sharpe 1.35; y1 −10.5%, m6 −17.3%, m3 −25.4%), related stocks (VST 0.96, CEG 0.89).

Peer / comparative context (public)

• Vistra Corp. (VST) — direct competitive-power peer; merchant-power industry framing, ERCOT/PJM structure, integrated-model and nuclear-mix comparison.
• Constellation Energy (CEG) — nuclear-heavy comparable; merchant-power industry context.
• Regulated utilities (AEP, DUK, SO, XEL, NEE, EXC, D) — comps for valuation/beta contrast.