2Q26: The Floor Rises, the Premium Stays Unpriced

Thesis status: Affirmed on the floor. Unproven on the premium. The quarter turned on three variables no condition watches.

Thesis Refresher

The primer asked whether Constellation Energy ($CEG) can convert the largest nuclear fleet in the United States into a durable above-market earnings premium. The federal Production Tax Credit floors the downside, hyperscaler co-location supplies the premium, and the Calpine acquisition has to be integrated without breaking the balance sheet that funds capital return. The horizon runs 2026 through 2029.

This is the first update. It covers Q2 2026, reported 6 August 2026, the second quarter of consolidated Calpine results and the first full quarter of the combined fleet.

Q2 2026 in Brief

The gap between GAAP and adjusted earnings is hedge-book mark-to-market and the amortisation of Calpine's commodity contracts at acquisition fair value, $340M and $149M after tax. Neither is cash.

The adjusted figure understates the operating step. Illinois ZEC revenue recognition fell to $85M from $200M as banked credits from prior planning years ran down, a headwind of roughly $0.32 per share against a $0.64 increase. Management stated the timing was already inside guidance and that this is the final planning-year adjustment before the programme ends in May 2027.

Guidance moved up $0.50 at the midpoint. Roughly $0.13 of that is the share count falling to a guided 357M from 361M. The remaining $0.37 is operational, and all of it sits in the Enhanced layer.

What the Quarter Turned On

Management attributed the quarter to three things: the addition of Calpine, higher capacity revenue in PJM, and commercial performance through higher realised customer margins and portfolio optimisation. No watch condition covers any of them.

Capacity is the largest of the three and the least visible. CEG cleared 18,875 MW of PJM capacity for 2028/29 at $325 per MW-day. Reference prices in the Eastern Mid-Atlantic Area Council and ComEd zones ran at $289.67 per MW-day in the quarter against $125.71 and $109.25 a year earlier. The primer's conditions track energy prices, contracted clean prices and the PTC floor. Capacity revenue is watched by nothing, and for nuclear units it feeds the gross receipts calculation that determines whether the PTC pays at all.

Calpine's contribution cannot be measured. Management disclosed $1,907M of operating revenues attributable to Calpine in the quarter and stated that integration and the consolidation of financing activities make it impracticable to determine Calpine's earnings since the acquisition date. Segment reporting uses revenues net of fuel, not EBITDA. Condition 5 tests integration and deleverage, not whether the acquired fleet earns what the acquisition underwrote, and on this disclosure it could not.

Commercial margin is where the guidance raise came from. Management named higher realised customer margins and portfolio optimisation during periods of volatility, and both sit in the Enhanced layer rather than the contracted Base.

Three drivers, no conditions. The overlay below records them every period, carrying no status and no thresholds, until a fiscal-year restatement decides whether any of them belongs in the set.

What the Price Now Requires

The primer carried no valuation. One is needed now, because the shares sit 21.4% below the $354.58 the company itself used to value the stock consideration for Calpine on 6 January 2026, and whether that de-rating has made the price undemanding has a specific answer.

It has not. Holding the model's capex intensity, tax rate and cash conversion constant, justifying $278.64 requires adjusted EBITDA to compound at a flat 10.38% for ten years. Management's disclosed Base earnings trajectory implies roughly 20% a year through 2029 before the contracted book stops growing. The model's own driver path lands at 5.74% across the decade.

Equity is not what the cash flows service. Net debt of $24,003M sits behind a $99,474M market capitalisation, so enterprise value is $123,477M and 14.4x the 2026 adjusted EBITDA base. A share price 21.4% below the January reference still leaves a hurdle above 10%.

Split the fleet the way management splits it. At 11x, the contracted Base layer of $5,791M in adjusted EBITDA is worth $63,701M. Against a $123,477M enterprise value, the residual leaves the market paying 21.3x for the $2,800M Enhanced layer.

The Base multiple carries that argument, so it is worth testing.

The conclusion survives the range. At 13x on the contracted book, Enhanced still needs 17.2x.

That multiple prices the whole Enhanced layer, which management defines as forward power prices above base assumptions, commercial margins above their ten-year average, and the capture of outsized value from volatility. Co-location appears nowhere in that definition. The premium the thesis rests on is what is expected to arrive in that layer, and none of it has arrived at a disclosed price.

On an equity basis the Base layer covers 39.9% of the share price, though that figure charges all $24,003M of net debt against Base. The convention is deliberately conservative and it is a convention: allocating debt pro-rata to EBITDA instead lifts Base coverage to 47.8%. The enterprise-value arithmetic above requires no such choice.

Watch Condition Assessment

Trends are measured against the primer's initial view, this being the first update.

C1: Nuclear PTC floor realisation

🟢 Affirmed ↑ The floor is higher than the primer stated, it indexes upward, and it is barely binding.

The primer fixed the floor at $43.75/MWh flat through 2032. That figure is the unindexed statutory phase-out, $25.00 of gross receipts plus the $15.00 maximum credit divided by the 80% reduction rate. The 2026 threshold indexed to $26.00, so the phase-out is $44.75/MWh and it rises every year the deflator does.

The IRS published a 2.8% inflation adjustment for 2025 against the 2% management carries in its outlook. Management stated the revision lifts the projected 2030 strike to $50.88/MWh from $49.88/MWh and 2030 Base earnings by roughly $0.30 per share. The primer's weakening test was the adjustment factor undershooting operating cost inflation. The first published observation ran the other way, and the 2026 to 2027 step in the strike of 2.2% outpaces the 1.8% step in guided adjusted operating and maintenance expense.

The floor is also close to irrelevant at current prices. Management disclosed that 5% of the nuclear fleet sits in the PTC support zone for 2026 and 2027. PJM West cleared $51.40/MWh in the quarter and $74.28/MWh across the half. Fleet-level realised gross receipts per MWh is not disclosed, so the read rests on the guided zone share and observed reference prices rather than on the metric the condition names.

C2: Contracted clean price trajectory

🟡 Developing → The volume arrived. The price is undisclosed by policy.

The disclosed trajectory did not move. Contracted clean volumes run 36 million MWh in 2026 to 54 million in 2029, at $70.00/MWh rising to $85.00/MWh, matching the primer's stated path exactly.

Constellation signed approximately 920 MW of long-term nuclear power purchase agreements in the quarter, at an average duration of 18.5 years with investment-grade counterparties, beginning between 2029 and 2031 and fully ramped by 2032. The volume comprises 890 MW of existing generation and a 30 MW uprate at Dresden enabled by a 176 MW agreement with Walmart ($WMT). None of it appears in the trajectory above. Management updates contracted clean volumes annually and has stated the schedule excludes agreements signed in 2026.

No price was disclosed for any of it. Management stated it will follow each customer's lead on disclosure to protect their procurement strategies, and declined to narrow the $20 to $50/MWh premium-to-floor range published in March, stating only that the executed transactions fit that profile. The condition's tests are numeric and the disclosure is not. A green light here would report a price trajectory as supported on the strength of a number nobody published.

C3: Co-location and hyperscaler contracted volume

🟡 Developing ↑ The three-pathway framework is gone. What replaced it moves faster than PJM's own baseline and slower than the primer assumed.

PJM has published proposals for Reliability Backstop Procurement and Interim Resource Adequacy Service, described by management as the successors to Connect and Manage. The behind-the-meter, virtual wheeling and traditional interconnection taxonomy the condition is written against is not the structure now in play, and no compliance filing exists on the primer's May 2026 timeline. The condition's procedural test cannot be run.

The substance moved in the right direction. FERC ordered every regional transmission organisation to justify or revise its large-load interconnection tariffs, and directed PJM to explain why new transmission services for co-located load cannot be made available sooner. PJM has set a 6.8 GW target for the backstop procurement, with an auction this autumn and results expected by year end. Management guided co-location rule clarity to the first or second quarter of 2027 against PJM's own prior target of 2029.

The most useful disclosure was procedural. Management stated the agreements signed this quarter are not dependent on the FERC and PJM outcomes. Contracted volume and regulatory resolution are separable in a way the primer's framing does not allow for, which lowers the stakes on a ruling now three quarters later than assumed.

C4: Crane and Eddystone transfer completion

🟡 Developing → Both regulatory gates cleared early. The restart date moved out by three quarters.

FERC granted the waiver transferring capacity interconnection rights from Eddystone Units 3 and 4 to Crane in June 2026, a quarter ahead of the primer's expectation, and the Nuclear Regulatory Commission approved Crane's fuel licence amendment in the same period.

The filing also disclosed the risk the primer did not carry. PJM's Phase I system impact study identified contingent transmission upgrades for Crane with projected in-service dates as late as December 2030, and the rights transfer is the mechanism for removing most of them. Clearing that tail restores the position the primer assumed Crane already held rather than improving on it, which is why the trend is flat despite two favourable rulings.

Management guided the restart to the second half of 2027, against the primer's fourth quarter of 2026 or first quarter of 2027. Per-plant earnings contribution is not disclosed and will not be, management having stated it cannot determine Calpine's segment earnings at all. The Department of Energy has separately guaranteed up to $1.0B of federal lending for the restart, undrawn as of the filing.

C5: Calpine integration and deleverage

🟡 Developing → The leverage path is intact. The share-neutral commitment is dead at any plausible price, and the collateral trigger grew faster than the resources covering it.

Net debt of $24,003M against the 2026 adjusted EBITDA base is 2.79x, below the 3.0x level the primer set as its stall signal. Roughly $5.86B of mandated divestiture proceeds arrive by year end, from $5.0B for five PJM assets and $860M for Brazos Valley, at nearly $1,200 per kW against the $960 per kW implied for the Calpine fleet. Those proceeds are the deleverage lever, though the 2026 earnings base still carries the divested assets' contribution and management does not split it out.

Two of the primer's inputs were wrong. The accounting consideration was $21,835M, not $16.4B, because the 50 million shares were valued at $354.58 on the last trading day before close, and the transaction created $11,107M of goodwill against $420M on the balance sheet before it. The $5B repurchase authorisation cannot make the transaction share-neutral. At $278.64 it buys 18.0 million of the 50 million shares issued, and offsetting all of them would take roughly $13.9B. Approximately $2.2B has been deployed and $2.8B remains.

The collateral obligation the primer calls the binding constraint is not a fixed quantity.

The trigger scales with the mark-to-market of a derivative book that grew with Calpine. Bank commitments rose to $14,482M from $9,458M, but letters of credit outstanding doubled and $2,586M of commercial paper is now drawn, so available capacity fell. Ratings of BBB+ and Baa1 leave three notches before the trigger activates, which is the reason this reads amber rather than red.

C6: Base earnings per share trajectory

🟡 Developing ↑ Guidance rose, and the whole of the rise came from outside the contracted book.

Full-year adjusted operating earnings guidance moved to $11.50 to $12.50 per share. The primer treats an upward revision to any year of the corridor as a strengthening signal on first announcement, and this is that.

The Base trajectory did not move. Base earnings remain $6.65 to $6.75 per share for 2026, $7.60 to $7.70 for 2027 and $11.40 to $11.90 for 2029. What rose was the Enhanced adjusted gross margin range, to $2,825M–$3,025M from $2,580M–$2,780M, alongside the lower share count. The primer's test is whether Enhanced contributions are additive to the Base framework rather than substituting for Base deterioration. On this quarter's disclosure they are additive, and the separation sharpened: the modelling tools now carry distinct base and enhanced interest lines and charge a $125M performance incentive explicitly against Enhanced.

The condition cannot be fully read. Base earnings per share is published in the annual outlook and not reported quarterly, so the test of two consecutive quarters inside the corridor has no quarterly observation to run against.

A: FERC/PJM tariff outcome

🟡 Developing ↑ No adverse ruling. The decision moved from a window that never opened to one that opens in 2027.

This condition is binary and neither outcome has occurred. The primer identified July and August 2026 as the highest-risk window, being the 60 days following an expected May compliance filing. No such filing was made and no such window opened. The live sequence is a PJM response in about November 2026 and a FERC order in the first half of 2027.

Nothing adverse has been issued. FERC spent the period pressing for speed rather than restricting pathways, which is the posture the primer's own framing required. The case does not need FERC to be friendly, only not hostile.

B: Investment-grade credit rating

🟢 Affirmed → Ratings sit one notch above the pass test with three notches of headroom to the trigger.

Constellation Energy Generation is rated BBB+ by S&P and Baa1 by Moody's, both with stable outlooks, against a pass test of BBB and Baa2. Neither agency has assigned a negative outlook, which the primer treats as a directional signal in its own right during the first twelve months after close.

The filing discloses a second cost the primer does not carry. On a loss of investment grade the revolving credit facility's interest adders step to 1.00% and 2.00% from 0.075% and 1.075%. The downgrade scenario draws collateral and raises the price of the facility that would fund it.

Thesis Standing

The floor did better than the primer's framing allowed for. The premium did not get more verifiable.

That asymmetry is the position. Two of the eight conditions are green, six are amber, and none is red. The absence of red is not the absence of problems. Twelve of the twenty-three strengthening and weakening triggers the primer specified cannot be read against what Constellation discloses, and the failures are not evenly distributed. C2's price tests, C4's per-plant accretion and C5's combined-fleet EBITDA fail because management does not publish those figures and has stated it will not. C3 and A fail because the procedural artefact they were written against no longer exists.

The distinction matters for what happens next. A timing gap closes on its own. A disclosure gap does not.

The valuation makes the gap concrete. The market is paying 21.3x for the Enhanced layer, and that multiple holds above 17x even at a generous 13x on the contracted book. Enhanced is where this quarter's guidance raise came from, where commercial margin and portfolio optimisation sit, and where the co-location premium will land if it lands. It is also the layer management guides one year out and marks conservatively. The 2027 Enhanced range of $2,150M to $2,550M sits below the 2026 range, which is why the modelling tool shows total adjusted EBITDA roughly flat next year while Base rises by $495M.

Nothing in the quarter threatens the floor. The 920 MW signed, the three notches of ratings headroom and the $5.86B of divestiture proceeds arriving by year end all point the same way on execution. What the quarter did not produce is one observable price for the premium the market is capitalising.

What to Watch

Source: Constellation Energy Form 10-Q for the quarter ended 30 June 2026, Exhibit 99.1 earnings release, Q2 2026 earnings presentation and earnings call transcript, all 6 August 2026. Marvin Labs quarterly data through 2Q26. Share price and valuation inputs at 14 August 2026.