NRG Energy
A power company that runs about 25 gigawatts of mostly gas-fired plants, sells electricity to 6 million customers and sells home security through Vivint.
The scores
Growth3/4 ↑
Adjusted EBITDA up 34% YoY is almost entirely the LS Power acquisition, not underlying growth, and Texas EBITDA fell on soft prices.
Business quality3/4 →
A real, externally corroborated scarce-turbine position underpins the BYOP pitch, but it is unproven through a signed contract.
Financial position2/4 ↓
Net debt rose about $6.8B in six months for a discretionary deal, cash fell near zero, and growth capex defers planned debt paydown.
Execution record3/4 →
T.H. Wharton delivered on time and on budget and guidance held through a soft H1, but the Q2 print missed consensus.
Per share capture2/4 ↓
Adjusted EPS fell 14% the same quarter EBITDA grew 34%, and share count rose 10.5% in six months on LS Power stock.
Potential3/4 ↑
A large, externally corroborated turbine-scarcity position with real observable progress (capital committed, terms aligned), but the flagship project has no signed contract yet.
Risk profile3/4 →
Rising leverage and single-counterparty concentration on the flagship deal, plus a real regulatory surprise, but diversified and survivable.
Valuation3/4 →
EV/TTM Adjusted EBITDA near 10.0x sits above NRG's own median, but the forward guided figure near 7.8x prices only the LS Power ramp.
Evidence quality3/4 →
Central claims are mostly primary filings, cross-checked against independent turbine-backlog and EIA data, but the flagship deal is unsigned.
If it works 1.2× · if it fails you keep about 100%.
The Verdict
We rate NRG neutral. It runs gas-fired power plants, sells electricity to millions of homes and owns a home-security business, and the business is mixed: a big purchase doubled its plants but added debt and new shares, so earnings per share fell while cash profit grew. The price is fair for what NRG already owns and pays nothing yet for its plan to build plants for data centres. As a bet it is defensive: if the first plant deal is signed, today's money grows by only about a fifth; if it never signs, nearly all of it is kept. We think the core business holds the price up.
The long caseTexas data centres need power faster than the grid can add it; if NRG signs buyers for its reserved plants, each earns a fixed payment for decades.
The evidence under this - 8 claims
Only the first 1.2 GW slice of the 5.4 GW has an aligned customer and committed capital; the rest of the pipeline being 'tied to an active customer discussion' is a much lower bar than a signed agreement, and the same turbine scarcity that protects NRG's position could equally let it walk away from slots that don't convert.
This TTM figure blends two different ownership bases - only five months of LS Power ownership in the trailing twelve months versus a full year going forward - so it understates the annualized EBITDA the current balance sheet actually supports, which is closer to what the 2026 guidance range implies.
- 2026-10-06editedRe-struck at the 2026-10-05 close of $96.80 (from $113.11, -14.4%, past the 12% drift trigger) with no print, filing or confirmed development in between: price multiples move, the debt, cash, share count and trailing profit do not.
The forward 7.8x multiple assumes the full 2026 guidance range is delivered; H1 2026 is tracking below the midpoint (C04), so the multiple that has actually been earned so far is closer to the trailing 10.0x than the forward figure.
- 2026-10-06editedRe-struck at the 2026-10-05 close of $96.80 (from $113.11, -14.4%, past the 12% drift trigger) with no print, filing or confirmed development in between: price multiples move, the debt, cash, share count and trailing profit do not.
When asked directly on the call, the CEO declined to name the counterparty or confirm its specific credit rating ('To quote my predecessor, no'), so the deal's central credit-quality claim cannot be independently verified from any source outside the company.
A 6x build multiple against NRG's own TTM Adjusted EBITDA multiple of about 10.0x is presented by management as evidence the project is accretive, but this compares an unrealized construction-stage estimate to a realized trading multiple, and the project carries real completion, land, and counterparty-approval risk between now and 2029 that a trading multiple does not.
The company
NRG does three things. It runs power plants, about 25 gigawatts since the LS Power purchase closed in January, most of them burning gas. It sells electricity and gas to about 6 million retail customers, and holds the largest share of Texas homes that choose their supplier. And it sells smart-home security to 2.45 million Vivint customers.
The new idea is 'bring your own power': NRG builds and runs a gas plant for one tech company's data centre, starting with 1.2 gigawatts in Texas.
The evidence under this - 3 claims
When asked directly on the call, the CEO declined to name the counterparty or confirm its specific credit rating ('To quote my predecessor, no'), so the deal's central credit-quality claim cannot be independently verified from any source outside the company.
How it makes money
Most of the profit comes from generating and retailing power, where prices swing with weather and markets: Texas profit fell 26% in the first half on weaker prices. Vivint adds steadier subscription income. The data-centre plants would be different: about 95% of their cash would come from fixed capacity payments, whether or not the data centre uses the power.
The evidence under this - 3 claims
The 95%-of-cash-flow-from-capacity-payments claim is management's own description of a structure that is not yet a signed contract, so the actual allocation between capacity and operating payments in any final agreement could differ from what was presented on the call.
Where the money comes from, stream by stream
| Stream | Share of sales (H1 2026) | How it earns it |
|---|---|---|
| Retail: business | 55.8% | Electricity, gas and demand response sold to commercial and industrial customers, mostly in the East, priced against power NRG buys or generates. |
| Retail: home | 29.0% | Electricity and natural gas sold to households, led by the largest competitive residential base in Texas. |
| Vivint Smart Home | 6.6% | Home security and automation sold as monthly subscriptions, with installation financed up front. |
| Wholesale energy and capacity | 7.9% | Power and capacity sold from NRG's plants into PJM and other markets, which LS Power multiplied. |
Who runs it
Robert Gaudette, at NRG for twenty-five years and in charge of its power plants and building projects since 2022, became chief executive in April 2026; Bruce Chung, the company's former dealmaker, has been finance chief since 2023. Insiders own less than 1%. The yearly bonus rides on cash flow and operating profit; the long-term shares ride on how the stock does against rivals. Nothing pays on profit per share.
The evidence under this - 3 claims
No metric pays on Adjusted EPS, per-share cash flow or return on capital, so an acquisition paid partly in stock that raises EBITDA while diluting per-share earnings (C05, C20) can still lift the bonus.
Why it wins, and who could beat it
Why customers come back
The scarce thing is a gas turbine. GE Vernova's order backlog reached about 116 gigawatts by mid-2026, its chief executive expects reservations to be sold out through 2030, and lead times have stretched to five to eight years.
NRG reserved 5.4 gigawatts of turbines and construction capacity through 2032 ahead of that queue. A rival promising a tech company power by 2029 would struggle to find a slot, which lets NRG price long contracts with a real delivery date.
The evidence under this - 3 claims
Other IPPs and utilities racing for the same scarce turbine slots - and data-center developers now bidding directly for turbines to build their own behind-the-meter capacity (C27) - mean NRG's reservation is a real but not exclusive position, and its value depends on converting the reserved slots into signed projects before a rival's equivalent reservation does.
Only the first 1.2 GW slice of the 5.4 GW has an aligned customer and committed capital; the rest of the pipeline being 'tied to an active customer discussion' is a much lower bar than a signed agreement, and the same turbine scarcity that protects NRG's position could equally let it walk away from slots that don't convert.
The threat
The tech companies may build their own. About 2 gigawatts of off-grid data-centre power was running by mid-2026, led by xAI's gas turbines, and more than 130 gigawatts more is proposed. Bloom Energy alone signed $7.65B of data-centre fuel-cell contracts in 2026, including 2.8 gigawatts for Oracle.
The evidence under this - 1 claim
Behind-the-meter capacity remains a small fraction of the multi-gigawatt pipeline NRG and its peers are pursuing, and cannot offer a hyperscaler the investment-grade counterparty risk transfer or grid-connected reliability NRG's BYOP structure provides - but it directly competes for the same scarce turbines (C25), which is the more relevant substitution channel for NRG specifically.
The rivals: SPCX, BE, MSFT, VST, CEG, TLN.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| SPCX | xAI (Colossus 1 and 2 gas-turbine sites near Memphis)Priced against it: Different product | Bring Your Own Power | On sale | Different product |
| BE | Bloom Energy (fuel-cell systems; 2.8 GW master agreement with Oracle within $7.65B of 2026 data-center contracts)Priced against it: One segment only | Bring Your Own Power | On sale | One segment only |
| MSFT | Microsoft (announced gas-turbine-powered data-center sites)Priced against it: Different product | Bring Your Own Power | Announced | Different product |
| VST | Vistra (NRG's filed generation competitor)Priced against it: Yes | Retail, generation and Vivint Smart Home | On sale | Yes |
| CEG | Constellation EnergyPriced against it: Yes | - | On sale | Yes |
| TLN | Talen EnergyPriced against it: No usable figure | - | On sale | No usable figure |
How big it could get
How big this could get
The market is new power demand in Texas and the mid-Atlantic. NRG's own annual report cites Texas peak demand rising from 86 gigawatts in 2024 to 139 in 2030, but the government has just cut its 2026 Texas growth forecast to 9.6% from 15.7%. Against that, NRG's pipeline is more than twice its 5.4 secured gigawatts, and none of it is yet signed.
The evidence under this - 3 claims
More current, independently reported ERCOT planning figures (August 2026) point to a higher systemwide 2030 peak near 138-152 GW, so this 10-K citation (filed February 2026, sourced to a 2023 EIA outlook) is already dated and likely understates the current forecast.
Only the first 1.2 GW slice of the 5.4 GW has an aligned customer and committed capital; the rest of the pipeline being 'tied to an active customer discussion' is a much lower bar than a signed agreement, and the same turbine scarcity that protects NRG's position could equally let it walk away from slots that don't convert.
Where the growth is coming from, business by business
Where the growth is coming from
| Business | Share of revenue | Growth: last quarter | Last half-year | Last year |
|---|---|---|---|---|
| East Segment | 63% | +40.5% | +31.8% | +26.9% |
| Texas Segment | 23% | -1.7% | +0.6% | +3.0% |
| West, Services and Other Segment | 8% | -21.3% | -18.2% | -17.4% |
| Vivint Smart Home Segment | 6% | +17.0% | +11.8% | +9.3% |
| Whole company | 100% | +19.3% | +13.2% | +11.2% |
The filed split through 2026-03-31. These rates are not carried forward: the reading below says why.
How we read it: The blend. East's 32% is bought: the LS Power plants, acquired January 2026, drove most of East's 63% EBITDA rise, while Texas fell on prices. One purchase, not a mix shift; the 3 reads the whole company.
The results
The numbers
Sales grew 13% over the twelve months to Jun 2026, about the same pace as the 9.2% in the year to Dec 2025.
Operating margin is 6.1%, up from 1.3% in the year to Dec 2023.
Yearly sales growth has ranged from −8.6% to +197% over the last 5 years, with sales falling in 2 of them.
Each dollar put into the business earns 7.9% a year after tax, down from 23% in the year to Dec 2021.
It turned 1.1% of sales into free cash, $348M over the last twelve months.
It owes $23.1B more than the cash it holds, about 14.9 years of the cash its operations bring in.
The share count fell 13% in 4.5 years, and pay in shares costs 0.4% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Jun 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $27.0B | $31.5B | $28.8B | $28.1B | $30.7B | $33.1B |
| Sales growth | +197% | +17% | −8.6% | −2.4% | +9.2% | +13% |
| Profit | ||||||
| Operating margin | 12% | 6.4% | 1.3% | 8.6% | 6.0% | 6.1% |
| Net margin | 8.1% | 3.9% | −0.7% | 4.0% | 2.8% | 2.6% |
| What growth costs, per sales dollar | ||||||
| Selling, general and admin | 4.8% | 3.6% | 7.3% | 8.3% | 8.5% | 7.5% |
| Pay in shares | 0.1% | 0.1% | 0.4% | 0.4% | 0.4% | 0.4% |
| Capital spending | 1.0% | 1.2% | 2.1% | 1.7% | 3.7% | 3.6% |
| Cash | ||||||
| Cash from operations | $493M | $360M | −$221M | $2.31B | $1.91B | $1.55B |
| Free cash flow | $224M | −$7.00M | −$819M | $1.83B | $766M | $348M |
| Balance sheet | ||||||
| Cash and short-term investments | $250M | $430M | $541M | $966M | $4.71B | $162M |
| Debt | $7.97B | $8.04B | $10.8B | $10.8B | $16.4B | $23.3B |
| Net cash | −$7.72B | −$7.61B | −$10.2B | −$9.84B | −$11.7B | −$23.1B |
| Returns and shares | ||||||
| Return on capital | 23% | 13% | - | 14% | 9.0% | 7.9% |
| Shares (diluted) | 245M | 236M | 228M | 212M | 199M | 212M |
What growth costs
NRG has grown by buying and building rather than by selling more of the same thing. Each step added costs before sales: the home-security business brought heavy selling costs, and the new plants mean far more spent on construction. So far, growth has made each sales dollar more expensive to earn.
The evidence under this - 2 claims
- The FY2023 10-K first reported 2022 selling, general and administrative costs as $1,228M (3.9 cents), later restated to $1,145M, so the starting point depends on which filing is read (S26).
- Most of the rise is bought rather than grown: Vivint Smart Home, acquired in March 2023, sells a subscription with heavy selling costs, and the capital spending is the Texas Energy Fund builds (C21).
The last report (4 Aug)
Quarter ended June 2026, filed figures.
The second quarter of 2026 showed the purchase working and costing at the same time. Adjusted cash profit rose 34% to $1.22B, carried by the newly bought plants in the East, while Texas fell on weak power prices.
But adjusted earnings per share fell 14% to $1.49, 11.8% below forecasts, because the interest and depreciation from the purchase came with the plants. The shares fell 15.5% that day, even though management kept its full-year forecast.
What it did to the thesisIt showed the purchase adding cash profit while its interest and depreciation cut earnings per share, the expected cost of the deal. It left the main question open: the data-centre contract is still unsigned, and the price gives that plan nothing yet, so the view stays Neutral.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $1.49 against $1.69 expected; $2.31 as filed, the gap being what the company adds back.
The evidence under this - 6 claims
The 'expected deleveraging profile' framing is management's own characterization of a decline it also chose to fund (LS Power was a discretionary, not a required, acquisition), so the same facts support a less charitable reading: the deal is diluting per-share earnings today for a return that depends on execution years out.
Management itself said H1 2026 results are tracking below the midpoint of these ranges on softer Texas load/prices and Winter Storm Fern costs, so reaffirmation describes the full-year range surviving a weak first half rather than the business outperforming it.
What management did with the money
LS Power cost $10.58B, paid partly in new shares, so the share count rose 10.5% in six months while debt rose about $6.8B. NRG still bought back $931M of shares in the first half, promises at least $1B a year, and repaid $1.5B of acquired secured notes.
The evidence under this - 5 claims
- 2026-09-26editedPut both dates on the same line of the Q2 2026 10-Q's debt note ($16,576M excluded finance leases, $23,437M included them) and stated the size of the rise: 41% gross, nearly double net of cash and restricted cash. The header had called the gross rise a doubling.
The price, and where we disagree
In short. Fair for what NRG already owns. Its plan to build power plants for data centres comes free until the first contract is signed.
The 15.5% fall on the June-quarter results means the market has already punished the earnings dip. The bigger question is the first data-centre plant. Terms are agreed and the customer has put up money, but nothing is signed. If it is, NRG expects $500M or more of yearly cash profit from a $3.2B plant over at least 15 years, from late 2029.
The September-quarter results on 5 November are the next chance for a signature or a firm date.
The evidence under this - 3 claims
When asked directly on the call, the CEO declined to name the counterparty or confirm its specific credit rating ('To quote my predecessor, no'), so the deal's central credit-quality claim cannot be independently verified from any source outside the company.
A 6x build multiple against NRG's own TTM Adjusted EBITDA multiple of about 10.0x is presented by management as evidence the project is accretive, but this compares an unrealized construction-stage estimate to a realized trading multiple, and the project carries real completion, land, and counterparty-approval risk between now and 2029 that a trading multiple does not.
How we read the multiple
Counting its debt, buyers pay about 10 times last year's cash profit, above its usual of about 8. Against this year's forecast it is about 7.8 times. So the price assumes the plants bought in January deliver that forecast, and pays nothing yet for the data-centre plants.
On reported profit it looks dearer, about 25 times, but the purchase's interest and other costs push this year's reported profit down.
Against its own history. Dearer than its own normal on last year's profit, but the business now includes a fleet it did not own a year ago.
The evidence under this - 3 claims
This TTM figure blends two different ownership bases - only five months of LS Power ownership in the trailing twelve months versus a full year going forward - so it understates the annualized EBITDA the current balance sheet actually supports, which is closer to what the 2026 guidance range implies.
- 2026-10-06editedRe-struck at the 2026-10-05 close of $96.80 (from $113.11, -14.4%, past the 12% drift trigger) with no print, filing or confirmed development in between: price multiples move, the debt, cash, share count and trailing profit do not.
The forward 7.8x multiple assumes the full 2026 guidance range is delivered; H1 2026 is tracking below the midpoint (C04), so the multiple that has actually been earned so far is closer to the trailing 10.0x than the forward figure.
- 2026-10-06editedRe-struck at the 2026-10-05 close of $96.80 (from $113.11, -14.4%, past the 12% drift trigger) with no print, filing or confirmed development in between: price multiples move, the debt, cash, share count and trailing profit do not.
Some of the GAAP-vs-Adjusted gap is real, recurring cash interest expense on $6.8B of new debt, not a one-time accounting item - so treating the whole P/E gap as noise the way VST's hedge-mark-to-market gap could be treated would overstate how temporary it is.
- 2026-10-06editedRe-struck at the 2026-10-05 close of $96.80 (from $113.11, -14.4%, past the 12% drift trigger) with no print, filing or confirmed development in between: price multiples move, the debt, cash, share count and trailing profit do not.
Priced against its rivals
The competitors its own annual report names, each confirmed against the rival's filing. Measured on price for each $1 of yearly profit, cheapest first:
| Ticker | Company | price for each $1 of yearly profit |
|---|---|---|
| VST | Vistra | 22.7× |
| CEG | Constellation Energy | 26.5× |
| NRG | NRG Energy | 29.6× |
The middle of the other 2 is 24.6. NRG Energy, Inc. is dearest of 3.
Rivals' figures come from a data provider, so read the spread of the group rather than any single pair.
The dearest of its rivals on reported earnings, but purchase costs depress those earnings this year, so the gap says more about the acquisition than the price.
The bet: what it pays if it works
This is a bet on one new business: gas plants NRG would build and run for a single tech giant, paid a fixed fee for capacity. The rest of NRG, its power plants, retail customers and home-security arm, is valued separately at what buyers usually pay for its profit, and that alone covers nearly all of today's price.
The first plant is not signed and earns nothing until late 2029, so even if it signs the bet adds little: it rates Expensive. If it never signs, shareholders keep the core.
| Bring Your Own Power EBITDA in two to three years | Dedicated plants needed | At the stated multiple |
|---|---|---|
| $500.00M (the big win) | about 1 at 100% kept | 1.2× |
| $1.00B | about 2 at 100% kept | 1.2× |
| $1.50B (market forecast) | about 3 at 100% kept | 1.3× |
ERCOT counts 66.4 GW of base large load through 2032 and 12.8 GW of bring-your-own gas; turbines, not demand, bind. Three of NRG's reserved plants contracted by 2029 earn about $1.5 billion once running, mostly after 2030. On that forecast, $1.50B pays 1.3× from today's value: an Expensive bet on the market case. Bring Your Own Power would be worth about $12.38B, and the price pays nothing for it.
To pay 2× (a Fair bet) it would need about $4.97B of Bring Your Own Power EBITDA (about 10 dedicated plants); to pay 5× (a Cheap bet), about $20.78B of Bring Your Own Power EBITDA (about 42 dedicated plants), at the same multiple.
Each row is Bring Your Own Power alone, plus the rest of the company (Retail, generation and Vivint Smart Home) counted at about $45.99B.
The steps it has to take
- Done 5.4 GW of turbine and construction capacity reserved with GE Vernova and Kiewit
- Done Principal terms aligned and a customer capital commitment for the first 1.2 GW Texas plant
- In progress A signed, binding agreement for the first plant, or a definitive investment-decision timeline due 2026-11-05
- Not started Final investment decision on the first plant, construction under way due 2027-06
- Not started A second capacity-paid contract signed against the remaining reserved turbines due 2027-12
- Not started First plant in commercial operation and capacity payments begin due 2029-12
The detail behind the label
The evidence under this - 4 claims
When asked directly on the call, the CEO declined to name the counterparty or confirm its specific credit rating ('To quote my predecessor, no'), so the deal's central credit-quality claim cannot be independently verified from any source outside the company.
A 6x build multiple against NRG's own TTM Adjusted EBITDA multiple of about 10.0x is presented by management as evidence the project is accretive, but this compares an unrealized construction-stage estimate to a realized trading multiple, and the project carries real completion, land, and counterparty-approval risk between now and 2029 that a trading multiple does not.
Management itself said H1 2026 results are tracking below the midpoint of these ranges on softer Texas load/prices and Winter Storm Fern costs, so reaffirmation describes the full-year range surviving a weak first half rather than the business outperforming it.
The forward 7.8x multiple assumes the full 2026 guidance range is delivered; H1 2026 is tracking below the midpoint (C04), so the multiple that has actually been earned so far is closer to the trailing 10.0x than the forward figure.
- 2026-10-06editedRe-struck at the 2026-10-05 close of $96.80 (from $113.11, -14.4%, past the 12% drift trigger) with no print, filing or confirmed development in between: price multiples move, the debt, cash, share count and trailing profit do not.
Where we disagree with the market
Split: some see a profit miss and rising debt, others an AI-power expansion worth paying for.
The earnings dip is the expected cost of the purchase, but the debt and new shares behind it are real, and the flagship deal is unsigned.
The evidence of NRG's edge sits in turbine makers' order books, not in its own filings, so a reader focused on the earnings miss sees only the cost of the purchase.
We are wrong on the downside if the data-centre deal stalls or reprices, or if debt keeps rising instead of turning back toward management's target.
What analysts say, and the evidence
The evidence under this - 6 claims
The 'expected deleveraging profile' framing is management's own characterization of a decline it also chose to fund (LS Power was a discretionary, not a required, acquisition), so the same facts support a less charitable reading: the deal is diluting per-share earnings today for a return that depends on execution years out.
When asked directly on the call, the CEO declined to name the counterparty or confirm its specific credit rating ('To quote my predecessor, no'), so the deal's central credit-quality claim cannot be independently verified from any source outside the company.
- 2026-09-26editedPut both dates on the same line of the Q2 2026 10-Q's debt note ($16,576M excluded finance leases, $23,437M included them) and stated the size of the rise: 41% gross, nearly double net of cash and restricted cash. The header had called the gross rise a doubling.
What could go wrong, and how we would know
The risks
- An unsigned deal. The first data-centre plant has agreed terms but no binding contract; internal approvals and land issues remain.
- Heavier debt. Debt rose about $6.8B in six months, and cash fell from $4.71B to $162M.
- Weak Texas prices. Houston power averaged $33 a megawatt-hour in the June quarter, well below the $52 NRG planned on for 2026.
- A new carbon cost. Virginia rejoining a regional carbon-trading scheme adds about $70M of unplanned cost this year on plants NRG just bought.
The evidence under this - 4 claims
When asked directly on the call, the CEO declined to name the counterparty or confirm its specific credit rating ('To quote my predecessor, no'), so the deal's central credit-quality claim cannot be independently verified from any source outside the company.
- 2026-09-26editedPut both dates on the same line of the Q2 2026 10-Q's debt note ($16,576M excluded finance leases, $23,437M included them) and stated the size of the rise: 41% gross, nearly double net of cash and restricted cash. The header had called the gross rise a doubling.
What we still do not know
- Who the tech customer is. NRG has not named it, and no volume or term is disclosed.
- Whether the rest of the 5.4 gigawatts finds buyers. Beyond the first plant it rests on 'active discussions'.
The evidence under this - 2 claims
When asked directly on the call, the CEO declined to name the counterparty or confirm its specific credit rating ('To quote my predecessor, no'), so the deal's central credit-quality claim cannot be independently verified from any source outside the company.
Only the first 1.2 GW slice of the 5.4 GW has an aligned customer and committed capital; the rest of the pipeline being 'tied to an active customer discussion' is a much lower bar than a signed agreement, and the same turbine scarcity that protects NRG's position could equally let it walk away from slots that don't convert.
What would change our view
- Neutral→AttractiveThe deal signs and the balance sheet mends
It needs all of:
- ○The tech company signs a binding contract for the first dedicated plant, on the terms described.watched as it happens
- ○Debt turns back toward management's target, and adjusted earnings per share rise once the purchased plants have been owned a full year.checked at the next results, due by 5 Nov
- Neutral→UnattractiveThe deal fails or the model is bypassed
It needs any of:
- ○The tech company deal fails to reach a binding contract on those terms.watched as it happens
- ○Data centres mostly build their own power instead, or this year's profit forecast is cut rather than met at the low end.watched as it happens
We would also be wrong if
- Adjusted earnings per share keep falling after the purchased plants' first full year.
- Debt keeps rising.
What is coming
- 5 November 2026 Q3 2026 results: Whether the hyperscaler plant reaches a signed, binding agreement or a dated final investment decision, and whether 2026 guidance holds with Texas prices still soft.
Every dated event ahead
- 5 November 2026 Q3 2026 results: Whether the hyperscaler plant reaches a signed, binding agreement or a dated final investment decision, and whether 2026 guidance holds with Texas prices still soft.
- February 2027 Q4 2026 results and 2027 guidance: The first guidance with a full year of LS Power, the $1B buyback delivered or not, and how far net debt has come back toward the 3x target.
- 31 May 2027 T.H. Wharton test period ends: Whether the first new plant performs well enough to earn the state's completion bonus, a read on how NRG runs what it builds.
- June 2028 Cedar Bayou 5 and Greens Bayou 6 come online: Whether NRG can deliver 1.5 GW of new gas plants on time and on budget, the test of building the data-centre plants at scale.
Since we wrote this
- 6 OctPrice moveRe-struck after a 14.4% fall, view unchangedThe shares fell with no new results, so the debt-inclusive price drops to about 10.0 times trailing adjusted profit from 10.8; the signed-deal question stands.
- 5 OctThesis reviewThe bet now names what it ridesThe owner asked that every bet name the macro trend behind it, after BWXT's bet was missing its chief executive's case for nuclear over gas.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 26 SepThesis reviewDebt rose 41%, nearly doubling net of cash. View heldThe top of the page said the LS Power deal doubled NRG's debt. Total debt rose 41%, to $23.4B; after the $4.5B of cash also spent, net debt nearly doubled.
- 24 AugThesis writtenNeutral, medium convictionWritten three weeks after the June-quarter results cut the shares 15.5% on an earnings miss, despite a held full-year forecast.
Also checked 3 times with nothing new to act on.
Sources 26
Sources - 26
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-02-24fleet composition, LS Power acquisition terms, competition section, risk factors, ERCOT/EIA load forecast citationbacksC01C02C16C22C44C48
- S02primarycompany filing2026-08-04balance sheet, debt schedule, LS Power purchase price allocation, share count, liquidity as of 2026-06-30backsC01C16C17C20C43C44C45
- S03primarycompany filing2026-08-04Adjusted EBITDA/EPS by segment, 2026 guidance tables, BYOP economics, capital allocation updatebacksC02C03C04C05C07C13C14C15C19C20C21
- S04primarytranscript2026-08-04NRG Q2 2026 earnings call transcriptBYOP commercial structure and economics, Q&A on FID timing and counterparty credit, ERCOT/PJM pipeline detail
We are aligned on the principal commercial terms with the leading global cloud and AI hyperscaler, including their capital commitment to support 1.2 gigawatts of new generation in Texas with the potential to expand to 2.4 gigawatts.
backsC07C08C09C10C11C12C13C15C18C19C26C28C29 - S05primarycompany filing2026-02-27confirmation of integrated generation-plus-retail model as a cohort peerread, but no claim rests on it
- S06primarycompany filing2026-02-26confirmation of pure-play PJM merchant generation as a cohort peerread, but no claim rests on it
- S07primarycompany filing2026-02-24confirmation of generation-plus-retail model as a cohort peer, nuclear-heavy asset mixread, but no claim rests on it
- S08secondarynews2026-08-24Utility Dive - EIA cuts 2026 power generation forecast, ERCOT/PJM demand growth EIA (via Utility Dive)independent, more current EIA-sourced ERCOT and PJM electricity demand growth rates than the 10-K's own dated citation
the EIA revised its forecast of ERCOT growth rates, shrinking its projections from 6% to 5% growth in 2025 and 15.7% to 9.6% growth in 2026
backsC23 - S09secondarynews2026-08-24Latitude Media / RTO Insider / Utility Dive - ERCOT large-load interconnection queue ERCOT (via trade press)independent size of the ERCOT large-load speculative queue
ERCOT's large load interconnection queue sits at 410 GW, about 87% of which are data centers
backsC24 - S10secondaryindustry data2026Cleanview - Bypassing the Grid: How Data Center Developers Are Building Their Own Power Plants Cleanviewindependent tracking of behind-the-meter/off-grid data-center generation capacity, the substitute for NRG's contracted BYOP model
approximately 2 GW of behind-the-meter generation capacity is online across four projects, with the vast majority belonging to xAI
backsC27 - S11secondaryindustry data2026Utility Dive / Turbomachinery Magazine - GE Vernova and Siemens Energy gas-turbine backlog GE Vernova / Siemens Energy (via trade press)independent, quantified confirmation that turbine capacity is industry-wide scarce and sold out for years, the external validation for NRG's own secured 5.4 GW
GE Vernova's Gas Power equipment backlog and slot reservation agreements grew to 116 GW by the second quarter of 2026 ... CEO Scott Strazik expects gas turbine reservations to be sold out through 2030 by the end of 2026
backsC25 - S12secondarynews2026-08-04Reuters - NRG Energy misses quarterly profit estimates as interest costs riseindependent read on the Q2 2026 Adjusted EPS miss and its interest-expense driverbacksC05C06C36
- S13secondarynews2026-08-07Seeking Alpha - NRG Energy: Aggressive GW Expansion (Rating Upgrade)evidence of a bullish sell-side read on the same BYOP news, used only as evidence of a split common viewbacksC36
- S14secondaryvendor data2026-08-24Portfolio-management research pack (market tier: price, insiders, analysts, peer valuation)content/NRG/research-pack.jsonprice/volatility snapshot, insider Form 4 aggregation, analyst consensus, cohort valuation multiplesbacksC05C06C30C34C35
- S15primarycompany filing2026-08-24data/reference/competitors.json - NRG filed-competitor-class cohort reviewdata/reference/competitors.jsonthe research-tier cohort build (VST, TLN, CEG) and each peer's confirming quotebacksC34C35
- S16primarycompany filing2026-02-24FY2025 segment Adjusted EBITDA by quarter, the original 2026 guidance issuancebacksC04C06
- S17primarycompany filing2026-05-06Q1 2026 Adjusted EBITDA, T.H. Wharton timeline update, Texas VPP program progressbacksC21
- S18primarycompany filing2026-04-28confirmation of the April 2026 refinancing that retired the Lightning notes assumed in the LS Power acquisitionbacksC18
- S19primaryregulator2026-04-15ERCOT filing to the PUCT, Project No. 58777 (item 38) - Preliminary Long-Term Load Forecast for Years 2026-2032, with Attachment A (2026 LTLF and Large Load RFI Results) Electric Reliability Council of Texas (ERCOT)the grid operator's preliminary load forecast with transmission providers' large-load submissions (non-crypto data centres 149,258 MW in 2029 against a 105,383 MW base forecast) and ERCOT's own stated doubt about it; PDF read through the ir_fetch client and pdftotext
As shown in Table 1 below, the forecast projects approximately 278,003 megawatts (MW) of total demand in the ERCOT Region by 2029 and 367,790 MW by 2032. For comparison, the ERCOT Region's current all-time peak demand is 85,508 MW. ... ERCOT currently projects that summer 2026 peak load will fall within a range of approximately 90,500 MW to 98,000 MW, compared with the preliminary LTLF's forecasted peak demand of approximately 112,000 MW for summer 2026
backsC37 - S20primaryregulator2026-09-11ERCOT Board of Directors, Item 14: Batch Zero Update (Jeff Billo), September 14-15, 2026 meeting Electric Reliability Council of Texas (ERCOT)ERCOT's conditional classification of the large loads in its first batch study, the energisation pause after the Governor's 3 August directive, and the bring-your-own-generation chart (WLPUN load 9,478 MW against generation 15,444 MW, of which gas 12,824 MW, solar 1,654, wind 637, battery 330); PDF read through the ir_fetch client and pdftotext
204 projects (66.4 GW) met the qualifications to be conditionally included as base load; 158 projects (127.9 GW) met the qualifications to be conditionally included as studied load. ... Paused approvals to energize data centers or virtual currency mining facilities (crypto facilities) that are 75 MW or greater. ... Batch Zero Withdrawal Limited Private Use Network (WLPUN) project details (BYOG) ... Total WLPUN projects 11
backsC38 - S21primarycompany filing2026-07-31Chevron Corporation 8-K Ex-99.1 - Chevron Reports Second Quarter 2026 Results Chevron Corporationa rival's signed dedicated-generation contract with a hyperscaler in Texas, larger than NRG's first plant; read through filing_fetch
Signed an agreement to develop a power facility in West Texas designed to provide approximately 2.67 gigawatts of behind-the-meter dedicated electricity capacity to a Microsoft data center under a 20-year power purchase agreement.
backsC39 - S22primarycompany filing2026-09-21New Era Energy & Digital 8-K Ex-99.1 - New Era Energy & Digital Secures 20-Year, 207 MW PPA with Vistra New Era Energy & Digital, Inc.the buyer's own filing of a 20-year dedicated gas supply from an existing Vistra plant in ERCOT, the route that competes with a new build; found by EDGAR full-text search (press_fetch --any-filer)
Under the PPA, Luminant will supply power from Vistra’s 1,180-MW natural gas-fired generating facility in Odessa, Texas, which is located immediately adjacent to the TCDC site. The PPA has an initial 20-year term, with automatic one-year renewal periods thereafter. The contracted power is expected to be available to TCDC in Q3 of 2027.
backsC39 - S23primarytranscript2026-08-04NRG Q2 2026 earnings call transcript (re-read for the BYOP pacing and pipeline)management's account that turbine deliveries, not customer demand, set how fast BYOP plants come; the analyst's reading of the slides of up to 18 months between the first and second commercial operation date; read through transcript_fetch (FMP)
As far as how to think about timing and where projects go, the thing that gets set on delivery of these projects is the CODs of construction and the turbine deliveries themselves. right? ... As we talk to customers and we think across these turbines, we have multiple customers looking for multiple turbines.
backsC40 - S24primaryregulator2026-10-01Public Utility Commission of Texas - The Texas Energy Fund (TxEF), In-ERCOT Generation Loan Program page Public Utility Commission of Texasthe state-financed new dispatchable build in ERCOT (4,994 MW across loan agreements, three of them NRG's, a 2,233 MW queue) that adds grid supply alongside any dedicated plant; read through ir_fetch
For the Electric Reliability Council of Texas (ERCOT) grid, the TxEF has allocated $3.65 billion for loans and $172.56 million for bonus grants to date. This funding supports the generation of 5,516 megawatts of new, reliable power to ERCOT and serves hundreds of thousands of homes and businesses. ... This program is capped at financing no more than 10,000 MW of new generation.
backsC41 - S25primarycompany filing2026-03-18executive officers and the CEO succession, beneficial ownership of directors and officers as of March 3, 2026 and of holders above 5%, the 2025 annual incentive and long-term incentive design and payoutsbacksC45C46C47
- S26primarycompany filing2024-02-28the Vivint Smart Home acquisition date and the selling, general and administrative costs for 2022 and 2023 as first filedbacksC44C48