Talen Energy
A power-plant owner on the eastern American grid: a nuclear plant selling electricity to Amazon's data centres next door, plus gas and coal plants selling into the grid.
The scores
Growth3/4 ↑
TTM revenue growth of 65.8% is the cohort's highest, but it is almost entirely acquisition-driven (Freedom, Guernsey, Cornerstone), not organic volume or price.
Business quality3/4 →
A genuinely scarce, co-located nuclear asset with a signed hyperscaler contract, but PJM concentration (98.3% of revenue) and rising hyperscaler self-build both press on it.
Financial position2/4 ↓
Net debt rose $2.76B in six months for a discretionary deal, cash fell to $231M, and leverage sits well above the below-3.5x target today.
Execution record3/4 →
Guidance was raised, not cut, and Cornerstone closed on the stated summer 2026 timeline, but the unexplained Keystone exclusion complicates a clean guidance comparison.
Per share capture3/4 →
Share count fell on net despite the acquisition, unlike a stock-funded peer deal, but GAAP EPS is deeply negative on hedge marks.
Potential3/4 ↑
A real, externally corroborated 6.8 GW reserve shortfall and a scarce co-located nuclear asset, but the pipeline beyond AWS remains uncommitted.
Risk profile2/4 ↓
Single-market PJM concentration, rising leverage, and forward hedge coverage that drops to just 30% by 2028 all compound the same commodity-price exposure.
Valuation3/4 →
About 14.5 times trailing and 10.9 times guided adjusted profit is fair; before adjustments it is twice that, mostly one quarter's stock award.
Evidence quality3/4 →
Central claims are mostly primary filings, cross-checked against independent reporting; the vendor's Talen figures differ from the filing by definition, and reconcile.
The Verdict
We rate Talen neutral. It owns power plants on the eastern American grid, centred on a nuclear plant contracted to sell power to Amazon's data centres until 2042. The business is mixed: in nine months it added a lot of gas plants with borrowed money. On the company's adjusted profit the price is fair; on reported profit it looks about twice as dear, mostly because of one large stock award. So it is fair if that award was a one-off. The view hinges on Talen signing another binding power contract and its debt falling toward the company's goal.
The long caseAI data centres in the eastern grid need power faster than new plants can be connected; Talen must sign more binding contracts and cut debt for that to reach shareholders.
The evidence under this - 7 claims
The $18 billion lifetime-revenue figure and the 2029-2032 ramp schedule are Talen's own investor-materials framing relayed by press rather than an independently verified figure, and the deal has not yet reached its full contracted volume, so the bulk of that revenue remains years away and depends on the transmission reconfiguration completing on schedule.
'Options' and 'active discussions' are a much lower bar than a signed agreement, and management's own language ('rationalization across projects') signals that a meaningful share of the pipeline is expected not to convert into a real contract.
- The forward 10.9x multiple assumes the full raised 2026 guidance range is delivered, including the unexplained Keystone exclusion (C09); if Adjusted EBITDA lands at the low end, the effective multiple is closer to the trailing 14.5x.
- Adjusted EBITDA adds back $577M of trailing stock-based and other long-term incentive compensation (C35), which is a real cost to shareholders; before the company's adjustments the same enterprise value is about 30.7x trailing EBITDA.
- 2026-09-26editedRe-struck at the 2026-09-25 close: 14.8x to 14.5x trailing, 11.1x to 10.9x on the guidance midpoint. The comparison with the vendor's 29.74x moved to C36, because the two multiples are struck on different definitions of profit rather than one being wrong.
- The compensation charge is lumpy rather than a run rate: the same line was $54M in FY2024 and $73M in H1 2026, so an unadjusted trailing figure dominated by one quarter understates what a normal year costs.
- Unrealized derivative losses reverse as the hedges settle and nuclear decommissioning trust gains are returns on a fund ring-fenced for decommissioning, so excluding both is a defensible choice rather than flattery.
- The 2026 guidance reconciliation adds back only interest, tax, depreciation and nuclear fuel, so the guided figure assumes no further stock-compensation or hedge items; the comparison between the two measures is trailing only.
- The peers' vendor figures are also struck before their own adjustments, so on the vendor's common basis Talen genuinely is the dearest of the three; calling the comparison unusable is about the frame's adjusted lens, not a claim that the premium is illusory.
- Only two peers are comparable on this metric, NRG and CEG, with VST withheld because its vendor income statement does not reconcile, so even a like-for-like median would rest on two points.
The company
Talen owns power plants and sells their electricity. Most of its 15.7 GW sits on PJM, the grid serving much of the eastern United States: gas plants, coal plants, and 90% of Susquehanna, the country's seventh-largest nuclear plant.
Susquehanna made about 17 TWh of electricity in 2025 at an all-in cost near $27 per MWh. Amazon, whose data-centre campus sits beside it, has contracted for up to 1,920 MW of its output through 2042.
The evidence under this - 3 claims
The $18 billion lifetime-revenue figure and the 2029-2032 ramp schedule are Talen's own investor-materials framing relayed by press rather than an independently verified figure, and the deal has not yet reached its full contracted volume, so the bulk of that revenue remains years away and depends on the transmission reconfiguration completing on schedule.
How it makes money
Talen earns three ways. It sells electricity into the grid, mostly hedged in advance: about 85% of 2026 output and 70% of 2027. PJM pays it for promising power when the grid needs it, and in the latest auction it cleared over 10 GW at the capped price of $325 per MW a day. And Amazon pays under a long contract.
The evidence under this - 3 claims
The $18 billion lifetime-revenue figure and the 2029-2032 ramp schedule are Talen's own investor-materials framing relayed by press rather than an independently verified figure, and the deal has not yet reached its full contracted volume, so the bulk of that revenue remains years away and depends on the transmission reconfiguration completing on schedule.
Where the money comes from, stream by stream
| Stream | Share of sales (FY2025) | How it earns it |
|---|---|---|
| Energy sales into ISO/RTO markets | 75.2% | Electricity and ancillary services from the nuclear, gas and coal fleet sold into PJM's and WECC's wholesale markets at market prices, mostly hedged ahead. |
| Capacity revenues | 18.8% | Payments for being available, set by PJM's capacity auctions and the Brandon Shores and Wagner reliability agreements; fixed once cleared. |
| Bilateral physical sales | 3.6% | Electricity sold directly to counterparties under contract, the line the Amazon nuclear agreement ramps through. |
| Derivative gains and other | 2.4% | Realized and unrealized results on hedges recorded in revenue, plus small other revenue; swings sign with power prices. |
Who runs it
Mac McFarland has run Talen since it left bankruptcy in May 2023 and hands over on 1 January 2027 to Terry Nutt, its finance chief until last year. Insiders own under 1%. The yearly bonus rides on adjusted profit and cash, safety and plant reliability; the share awards ride on the share price, and part of this year's was paid in cash. Nothing rewards cutting debt.
The evidence under this - 4 claims
The table predates the 2026 vesting of the emergence grants, 40% of whose after-tax value was settled in shares (C43), so insider holdings after that vesting are likely higher than this proxy shows.
The cash settlement spends company cash on management's awards in the same year that buybacks are presented as the return to shareholders, which C30's independent report of a payout above $300M puts in scale.
Why it wins, and who could beat it
Why customers come back
It owns two things a rival starting today cannot quickly get. One is a licensed, running nuclear plant next to a large data-centre campus. The other is gas turbines already built: new ones are effectively sold out through 2030, with GE Vernova's order backlog at about 116 GW.
That is why Amazon took Susquehanna's power on a contract through 2042 rather than build its own.
The evidence under this - 3 claims
The $18 billion lifetime-revenue figure and the 2029-2032 ramp schedule are Talen's own investor-materials framing relayed by press rather than an independently verified figure, and the deal has not yet reached its full contracted volume, so the bulk of that revenue remains years away and depends on the transmission reconfiguration completing on schedule.
The threat
The customers may build their own. Nearly every large cloud company is now pursuing power plants on its own sites, off the grid: xAI's two sites run about 1,498 MW, and OpenAI and Anthropic have leased more than 10 GW. If that becomes the norm, generators like Talen are bypassed.
The evidence under this - 1 claim
Behind-the-meter capacity cannot replicate a single 1,920 MW carbon-free nuclear block the way Susquehanna can, and it directly competes for the same scarce gas turbines Talen's own gas fleet expansion depends on (C21) - which is the more relevant substitution channel for Talen's non-nuclear growth specifically.
The rivals: SPCX, META, WMB, NRG, CEG, VST, and 1 more that is private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | How far along | Priced against it |
|---|---|---|---|
| SPCX | xAI (Colossus 1 and 2 gas-turbine sites near Memphis)Priced against it: Different product | On sale | Different product |
| Private | OpenAI and Anthropic (more than 10 GW of behind-the-meter power leases reported industry-wide)Priced against it: Private | Announced | Private |
| META · WMB | Meta (reported gas-generation partnerships, including a Williams Companies plant in Ohio)Priced against it: One segment only | Announced | One segment only |
| NRG | NRG EnergyPriced against it: Yes | On sale | Yes |
| CEG | Constellation EnergyPriced against it: Yes | On sale | Yes |
| VST | VistraPriced against it: No usable figure | On sale | No usable figure |
How big it could get
How big this could get
PJM's grid is short of power. Its auction for 2028 to 2029 cleared at the price cap of $325 per MW a day and still came in about 6.8 GW short, so PJM is running a one-off extra auction from September 2026.
Nobody has sized how much of that data centres will contract for. Talen has about 4 GW of land and data-centre options and over 2 GW of projects waiting to connect, none of it signed.
The evidence under this - 3 claims
'Options' and 'active discussions' are a much lower bar than a signed agreement, and management's own language ('rationalization across projects') signals that a meaningful share of the pipeline is expected not to convert into a real contract.
Who has signed
One contract: Amazon, for up to 1,920 MW of nuclear power through 2042, reaching full volume by 2032 and worth roughly $18B over its life on the company's own figures. Delivery moves from a direct connection to the grid by spring 2027.
The evidence under this - 1 claim
The $18 billion lifetime-revenue figure and the 2029-2032 ramp schedule are Talen's own investor-materials framing relayed by press rather than an independently verified figure, and the deal has not yet reached its full contracted volume, so the bulk of that revenue remains years away and depends on the transmission reconfiguration completing on schedule.
The results
The numbers
Sales grew 61% over the twelve months to Jun 2026, faster than the 22% in the year to Dec 2025.
Operating margin is 2.6%.
Shareholders' money lost 13% a year, down from 72% in the year to Dec 2024.
It turned 17% of sales into free cash, $568M over the last twelve months.
It owes $9.34B more than the cash it holds, about 11.7 years of the cash its operations bring in.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in December.
| Dec 2024 | Dec 2025 | 12 mo to Jun 2026 | |
|---|---|---|---|
| Size and growth | |||
| Sales | $2.12B | $2.58B | $3.44B |
| Sales growth | - | +22% | +61% |
| Profit | |||
| Operating margin | 11% | −3.5% | 2.6% |
| Net margin | 47% | −8.5% | −5.4% |
| What growth costs, per sales dollar | |||
| General and admin | 7.7% | 24% | 20% |
| Pay in shares | 1.6% | 20% | 17% |
| Capital spending | 4.0% | 3.8% | 6.6% |
| Cash | |||
| Cash from operations | $256M | $704M | $796M |
| Free cash flow | $171M | $606M | $568M |
| Balance sheet | |||
| Cash and short-term investments | $328M | $689M | $231M |
| Debt | $3.00B | $6.81B | $9.57B |
| Net cash | −$2.68B | −$6.12B | −$9.34B |
| Returns and shares | |||
| Return on capital | 4.7% | - | - |
| Shares (diluted) | 56.5M | 45.7M | 45.9M |
What growth costs
Talen does not spend on research or selling: it grows by buying power plants, mostly with borrowed money. Its overhead jumped last year almost entirely because of one cash-settled award to management, so there is no trend yet showing whether a bigger fleet makes each dollar of sales cheaper to earn.
The evidence under this - 2 claims
Revenue in each period carries realized and unrealized hedge results (C37) and the Freedom, Guernsey and Cornerstone acquisitions, so cost per sales dollar moves with commodity marks and deals as much as with any operating leverage; two clean fiscal years are too few to separate them.
The cash settlement spends company cash on management's awards in the same year that buybacks are presented as the return to shareholders, which C30's independent report of a payout above $300M puts in scale.
The last report (5 Aug)
Quarter ended June 2026, filed figures.
The June quarter looked bad on the official accounts: a net loss of $92M against a $72M profit a year earlier. The company's own adjusted profit measure said the opposite, rising to $374M from $90M as capacity and energy revenue grew.
The gap is mostly paper losses on its hedges and interest on its new debt. The company raised its 2026 forecast to $2.025B to $2.225B of that adjusted profit.
What it did to the thesisIt supported the case that the business is earning more, since the company's adjusted profit rose sharply and the forecast went up. It also showed the cost of the debt behind that growth, and no new power contract arrived, so the view stays Neutral.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $-2.00 against $3.21 expected; $-2.00 as filed, with no adjustments.
The evidence under this - 3 claims
Not all of the GAAP-vs-Adjusted gap is a hedge-accounting artifact: a real, recurring share of it is cash interest expense on the roughly $2.8B of new debt raised over the six months, which is not going to reverse the way an unrealized mark can.
The guidance raise is presented as reflecting only the Cornerstone close and year-to-date strength, but the unexplained exclusion of Keystone from the 2026E figures makes the pre- and post-raise ranges not a strictly like-for-like comparison.
What management did with the money
It borrowed to grow and to buy back shares. Long-term debt rose from $6.81B to $9.57B in the first half of 2026 while cash fell to $231M, and it has bought back about 15 million shares for about $2.3B since 2024. Its chief executive received a stock payout of more than $300M.
Insiders in 2026Q2: 26 purchases and 41 sales, from their Form 4 filings.
The evidence under this - 3 claims
- 2026-09-30edited8-K of 2026-09-29: program upsized to $3.0B with a $1.5B accelerated repurchase; the $1.7B remaining in the 10-Q is superseded.
A large one-time compensation event of this scale is a governance and per-share-dilution consideration to weigh against the company's simultaneous framing of shareholder returns through buybacks, even though it is disclosed compensation rather than an open-market transaction.
The price, and where we disagree
In short. Fair on the company's adjusted profit, dear on its reported one. To be more than fair it needs a second big power contract and falling debt.
The shares closed on 25 September 32.5% below their high of the past year. The September-quarter report on 4 November shows whether debt starts falling toward the company's target, and whether any data-centre option has become a signed contract. PJM's extra auction is a separate reading on what capacity is worth.
The evidence under this - 4 claims
- 2026-09-26editedRe-struck from the refreshed research pack: $314.46 on 2026-08-21 to $304.42 on 2026-09-25; 30.3% to 32.5% below the 252-day high; RSI 37.3 to 47.7; the golden cross of the old pack has since turned to a death cross.
- Net debt at 2026-06-30 (about $9.34B against TTM Adjusted EBITDA of about $1.59B, C24) implies leverage well above the 3.5x target today; the target is only reached if the raised 2026 guidance range is fully delivered and no further debt-funded acquisition follows Cornerstone.
- The 29 September repurchase is funded by selling future capacity revenue at a floating rate, so the cash is bought forward rather than earned; whether that obligation counts in net debt is the company's own treatment, not yet filed in a statement.
- 2026-09-30edited8-K of 2026-09-29: capacity-revenue monetisation funds the repurchase; management restates the 3.5x target as second half of 2027.
'Options' and 'active discussions' are a much lower bar than a signed agreement, and management's own language ('rationalization across projects') signals that a meaningful share of the pipeline is expected not to convert into a real contract.
How we read the multiple
Counting its debt, Talen costs about 14.5 times last year's profit on its own adjusted measure, and 10.9 times this year's forecast. That measure leaves out $577M of pay in shares, $486M of it in one quarter. Before adjustments it is about 31 times; a data vendor counts the same thing at about 34.
So the price is fair if that award was a one-off, and dear if share pay stays that high. It adds nothing for the Amazon contract and takes nothing off for the debt.
Against its own history. Far dearer than its own normal, but that normal predates the Amazon contract and the acquisitions, and this table counts profit before adjustments, which one large stock award pulled down.
The evidence under this - 4 claims
This TTM figure blends less than one full quarter of Cornerstone ownership into a mostly pre-Cornerstone base, so it understates the annualized Adjusted EBITDA the current, larger fleet actually supports going forward, which is closer to the raised 2026 guidance range.
- 2026-09-26editedEnterprise value re-struck at the 2026-09-25 close from the refreshed research pack: market cap $14,274,914,216 to $13,819,148,335, enterprise value $23.62B to $23.16B. The Adjusted EBITDA build and the net debt are unchanged.
- The forward 10.9x multiple assumes the full raised 2026 guidance range is delivered, including the unexplained Keystone exclusion (C09); if Adjusted EBITDA lands at the low end, the effective multiple is closer to the trailing 14.5x.
- Adjusted EBITDA adds back $577M of trailing stock-based and other long-term incentive compensation (C35), which is a real cost to shareholders; before the company's adjustments the same enterprise value is about 30.7x trailing EBITDA.
- 2026-09-26editedRe-struck at the 2026-09-25 close: 14.8x to 14.5x trailing, 11.1x to 10.9x on the guidance midpoint. The comparison with the vendor's 29.74x moved to C36, because the two multiples are struck on different definitions of profit rather than one being wrong.
- The compensation charge is lumpy rather than a run rate: the same line was $54M in FY2024 and $73M in H1 2026, so an unadjusted trailing figure dominated by one quarter understates what a normal year costs.
- Unrealized derivative losses reverse as the hedges settle and nuclear decommissioning trust gains are returns on a fund ring-fenced for decommissioning, so excluding both is a defensible choice rather than flattery.
- The 2026 guidance reconciliation adds back only interest, tax, depreciation and nuclear fuel, so the guided figure assumes no further stock-compensation or hedge items; the comparison between the two measures is trailing only.
- The peers' vendor figures are also struck before their own adjustments, so on the vendor's common basis Talen genuinely is the dearest of the three; calling the comparison unusable is about the frame's adjusted lens, not a claim that the premium is illusory.
- Only two peers are comparable on this metric, NRG and CEG, with VST withheld because its vendor income statement does not reconcile, so even a like-for-like median would rest on two points.
Priced against its rivals
The competitors its own annual report names, each confirmed against the rival's filing. Measured on company value for each $1 of cash profit, cheapest first:
| Ticker | Company | company value for each $1 of cash profit |
|---|---|---|
| NRG | NRG Energy | 10.6× |
| CEG | Constellation Energy | 16.4× |
| TLN | Talen Energy | 34.2× |
| VST | Vistra held out | not priced |
The middle of the other 2 is 13.5. Talen Energy Corporation 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 here, but every figure is profit before each company's own adjustments; no rival figure exists on the adjusted profit the company reports.
Where we disagree with the market
An AI-power winner whose price has already run to a full, rich valuation.
On the company's adjusted profit Talen is fair, not rich; the rich reading is mostly one quarter's stock award.
Analysts are 11 buy and 2 hold, with no sell. The rich reading is not a mistake: before the company's adjustments, last year's profit carried $577M of stock pay, so ratios built on it look twice as dear.
We are wrong if stock pay stays near that level, if capacity prices fall once new supply arrives, or if debt fails to fall toward 3.5 times a year's adjusted profit.
What analysts say, and the evidence
The evidence under this - 7 claims
- 2026-09-26editedRe-read against the refreshed research pack: the rating split is unchanged; estimate coverage for 2028 to 2030 is now two to six analysts, was one to five.
- 2026-09-26editedRe-struck against the refreshed research pack, whose filed revenue is now the total `Revenues` tag ($2,581M; the contract-revenue tag's $2,518M before), and reconciled from the 10-K income statement: both gaps are the unrealized derivative lines. The original claim read the operating-income gap as a sign flip that made the vendor's figures unreliable; it is a definitional difference and the vendor figure is correct under its definition.
- The compensation charge is lumpy rather than a run rate: the same line was $54M in FY2024 and $73M in H1 2026, so an unadjusted trailing figure dominated by one quarter understates what a normal year costs.
- Unrealized derivative losses reverse as the hedges settle and nuclear decommissioning trust gains are returns on a fund ring-fenced for decommissioning, so excluding both is a defensible choice rather than flattery.
- The 2026 guidance reconciliation adds back only interest, tax, depreciation and nuclear fuel, so the guided figure assumes no further stock-compensation or hedge items; the comparison between the two measures is trailing only.
- The peers' vendor figures are also struck before their own adjustments, so on the vendor's common basis Talen genuinely is the dearest of the three; calling the comparison unusable is about the frame's adjusted lens, not a claim that the premium is illusory.
- Only two peers are comparable on this metric, NRG and CEG, with VST withheld because its vendor income statement does not reconcile, so even a like-for-like median would rest on two points.
- The forward 10.9x multiple assumes the full raised 2026 guidance range is delivered, including the unexplained Keystone exclusion (C09); if Adjusted EBITDA lands at the low end, the effective multiple is closer to the trailing 14.5x.
- Adjusted EBITDA adds back $577M of trailing stock-based and other long-term incentive compensation (C35), which is a real cost to shareholders; before the company's adjustments the same enterprise value is about 30.7x trailing EBITDA.
- 2026-09-26editedRe-struck at the 2026-09-25 close: 14.8x to 14.5x trailing, 11.1x to 10.9x on the guidance midpoint. The comparison with the vendor's 29.74x moved to C36, because the two multiples are struck on different definitions of profit rather than one being wrong.
- Net debt at 2026-06-30 (about $9.34B against TTM Adjusted EBITDA of about $1.59B, C24) implies leverage well above the 3.5x target today; the target is only reached if the raised 2026 guidance range is fully delivered and no further debt-funded acquisition follows Cornerstone.
- The 29 September repurchase is funded by selling future capacity revenue at a floating rate, so the cash is bought forward rather than earned; whether that obligation counts in net debt is the company's own treatment, not yet filed in a statement.
- 2026-09-30edited8-K of 2026-09-29: capacity-revenue monetisation funds the repurchase; management restates the 3.5x target as second half of 2027.
What could go wrong, and how we would know
The risks
- Growth bought with debt. Long-term debt rose $2.76B in six months while cash fell to $231M.
- Capacity prices were capped. $325 was the ceiling, and the extra auction and new supply could bring future prices well down.
- Customers building their own. Off-grid power for data centres is scaling fast and could bypass generators like Talen.
- One contract, one customer. Amazon is the only signed large contract, and data-centre groups argue even a customer with its own generation still faces cut-off risk on the grid.
The evidence under this - 6 claims
Behind-the-meter capacity cannot replicate a single 1,920 MW carbon-free nuclear block the way Susquehanna can, and it directly competes for the same scarce gas turbines Talen's own gas fleet expansion depends on (C21) - which is the more relevant substitution channel for Talen's non-nuclear growth specifically.
The $18 billion lifetime-revenue figure and the 2029-2032 ramp schedule are Talen's own investor-materials framing relayed by press rather than an independently verified figure, and the deal has not yet reached its full contracted volume, so the bulk of that revenue remains years away and depends on the transmission reconfiguration completing on schedule.
The docket comment is about co-located arrangements broadly (naming Vistra and Constellation, not Talen specifically), and Talen's own AWS PPA is structured with a transition to front-of-meter delivery precisely to reduce this kind of curtailment exposure, so the risk may be more mitigated for Talen than for a purely co-located design.
What we still do not know
- Whether any of the 4 GW of options becomes a signed contract. Management calls 2026 a year of market rationalization.
- Why the raised forecast leaves out the Keystone coal plant. The release gives no reason.
The evidence under this - 2 claims
'Options' and 'active discussions' are a much lower bar than a signed agreement, and management's own language ('rationalization across projects') signals that a meaningful share of the pipeline is expected not to convert into a real contract.
The guidance raise is presented as reflecting only the Cornerstone close and year-to-date strength, but the unexplained exclusion of Keystone from the 2026E figures makes the pre- and post-raise ranges not a strictly like-for-like comparison.
What would change our view
- Neutral→AttractiveAnother contract signs and the leverage falls
It needs all of:
- ○Talen signs another binding contract from its land and data-centre pipeline, beyond Amazon.watched as it happens
- ○Debt falls back toward the company's target of under 3.5 times a year's adjusted profit as the latest purchase is absorbed.checked at the next results, due by 4 Nov
- Neutral→UnattractiveCapacity prices or the ramp slip
It needs any of:
- ○A future PJM auction clears well below $325 per MW a day as new supply arrives.checked at the next results, due by 4 Nov
- ○Amazon's move to grid delivery is delayed or its ramp slips, or off-grid power becomes the default for new data centres.watched as it happens
We would also be wrong if
- No second signed contract arrives within the next 18 to 36 months.
What is coming
- 4 November 2026 Q3 2026 results: Whether net leverage starts falling toward 3.5x with Cornerstone in a full quarter, and whether a second large-load contract is signed beyond Amazon.
Every dated event ahead
- 4 November 2026 Q3 2026 results: Whether net leverage starts falling toward 3.5x with Cornerstone in a full quarter, and whether a second large-load contract is signed beyond Amazon.
- 13 November 2026 Lock-up on officers' emergence-grant shares ends: Whether the outgoing and incoming leaders keep or sell the shares they received in 2026; a test of how much of their pay they hold.
- 1 January 2027 Terry Nutt becomes chief executive: Whether capital allocation stays on buybacks and leverage reduction or turns back to debt-funded acquisitions under the former finance chief.
- March 2027 Accelerated repurchase completes: The final share count retired for $1.5B, and whether the forward-sold capacity revenue left leverage on the path management dated to late 2027.
Since we wrote this
- 5 OctThesis reviewThe long case now sits under the verdictAfter the five-name pilot, the owner asked for the long case on every thesis.
- 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.
- 30 SepNewsBigger buyback funded by future capacity revenue; view heldTalen will spend $1.5B buying back over 10% of its shares, paid for by selling future power-capacity income. It says leverage reaches 3.5x in late 2027; not yet shown in filed figures.
- 26 SepThesis reviewVendor multiple was unadjusted, not wrong; view heldThe gap between the vendor's 29 times and the company's 14.5 times is mostly a $486M stock award booked in one quarter. Fair on adjusted profit, dear on reported; still Neutral.
- 24 AugThesis writtenNeutral, medium convictionA nuclear plant contracted to Amazon plus a gas fleet bought with debt, priced near its peers on the company's own cash profit: fair, not rich.
Also checked 3 times with nothing new to act on.
Sources 22
Sources - 22
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 filing2026-02-26fleet composition, Susquehanna detail, AWS PPA terms, RMR agreements, Cornerstone agreement terms, competition sectionbacksC01C02C03C04C05C23C27C37C39C43C44
- S02primarycompany filing2026-08-05balance sheet, debt schedule and financing detail, EPS, segment split, as of 2026-06-30backsC08C14C15C18C37C38C39C44
- S03primarycompany filing2026-08-05Q2/H1 2026 results, raised 2026 guidance, Cornerstone close, share repurchases, financing transactions, hedging position, PJM BRA result, land pipelinebacksC01C05C06C08C09C10C15C16C17C19C20
- S04primarycompany filing2026-08-24confirmation of the filing history and dates used to build the source manifestbacksC01
- S05primarycompany filing2026-05-05Q1 2026 Adjusted EBITDA ($473M) and Adjusted Free Cash Flow ($350M), cross-checked against the Q1 2026 earnings callbacksC06
- S06primarycompany filing2026-02-26FY2025 and FY2024 Adjusted EBITDA/FCF, original 2026 guidance issuance, Cumulus Data/ERCOT sale gain detailbacksC07C09
- S07secondarynews2025-06-11independent confirmation of the AWS PPA ramp schedule and lifetime revenue framing
Talen expects to earn about $18 billion in revenue over the life of the contract at its full quantity, according to an investor presentation.
backsC03 - S08secondarynews2026-07-14independent confirmation that the $325/MWd clearing price was a regulatory cap, and that the auction still fell 6.8 GW short of the reliability requirement
capped out at the $325-MW/day maximum price imposed by a collar on prices
backsC10 - S09secondarynews2026-07-15Aurora Energy Research's modeled uncapped clearing price estimate ($555/MWd region-wide, $777/MWd ComEd)
the current system doesn't work to bring online new capacity or stimulate demand response
backsC10C11 - S10secondarynews2026-07-28independent confirmation of the September 2026 backstop auction and its purpose
The PJM Interconnection will hold a one-time backstop capacity auction starting in September to address a 6.8-GW shortfall from its just-held capacity auction for the delivery year that starts in mid-2028.
backsC12 - S11secondarynews2026-03-27independent evidence of curtailment risk in co-located large-load arrangements, relevant to Talen's own AWS structure
Even a customer that brings sufficient co-located generation to meet its load cannot avoid curtailment risk.
backsC13 - S12primarytranscript2026-05-05Talen Q1 2026 earnings call transcriptmanagement framing of PJM reserve adequacy, FCF/share outlook detail, Cycle 1 interconnection pipeline detail
we have said for a number of years now, that is really a fifty-hour problem
backsC20C34 - S13secondarynews2026-08-24Cornerstone completion market reaction (+5.3%) and the CEO stock-payout report, cross-checked against the Cornerstone-completion 8-KbacksC05C31
- S14secondaryindustry data2026-07-10independent, quantified confirmation of gas-turbine industry scarcity underlying Talen's own gas-fleet expansion
The company reports turbine supply effectively sold out through 2030, with its equipment highlighted as central to closing the projected power gap.
backsC21 - S15secondaryindustry data2026independent tracking of behind-the-meter/off-grid data-center generation, the substitute for Talen's contracted PPA and merchant-generation model
Nearly every hyperscaler is pursuing behind-the-meter power projects-either directly or through partners.
backsC22 - S16secondaryvendor data2026-09-26Portfolio-management research pack (vendor_fundamentals tier: filed-vs-vendor cross-check)content/TLN/research-pack.jsonthe filed-versus-vendor valuation_cross_check (annual revenue and operating income gaps against SEC XBRL)backsC23
- S17secondaryvendor data2026-09-26Portfolio-management research pack (market tier: price, insiders, analysts, technicals)content/TLN/research-pack.jsonprice/technicals snapshot, insider Form 4 aggregation, analyst consensus and estimates, market capitalisation and enterprise valuebacksC29C32C33
- S18secondaryvendor data2026-09-26Portfolio-management research pack (vendor_fundamentals tier: peer_valuation cohort)content/TLN/research-pack.jsonthe filed-competitor cohort valuation multiples (VST, NRG, CEG) and the profit_line_limited/filed_margin_gap detailbacksC27
- S19primarycompany filing2026-08-05PJM-versus-AllOther segment revenue split for Q2 2026backsC28
- S20secondarynews2026-05-28Reuters - Talen Energy CEO stock-based payoutindependent report of the CEO's stock-based payout, used as a governance/dilution considerationbacksC30
- S21primarycompany filing2026-09-29the succession, the accelerated repurchase terms, the capacity-revenue financing and management's leverage-timing statementbacksC16C19C40C45
- S22primarycompany filing2026-03-19executive officers and tenure, the annual bonus metrics and 2025 results, the emergence and 2025 equity award terms, the 2026 cash settlement and lock-up, and beneficial ownership as of 2026-03-17backsC40C41C42C43