SEQH Capital Research · Research report · Draft, Nov 7, 2026
Model: SEQH_DataCenter_Slowdown_Model_v1.xlsx (paid subscribers)
Power stocks had their worst two days of the year right after the election. $VST fell 13.8% on Wednesday to $120.29. $CEG fell 6.7% to $245.72. Both kept falling on Thursday. The reasoning was simple: a Democratic Congress and a Democratic governor-elect in Texas mean a slower, more expensive data-center build, and less demand for the power these companies sell.
Vistra reported Friday morning, and its numbers gave the first real test of that reasoning. Constellation reports Monday. This report sets out how much of each company’s earnings a slower build actually reaches, using company filings and a simple model. It also explains why the three US nuclear restarts are mostly outside that risk.
The short version: in the near term, very little of either company’s earnings is exposed. Vistra has sold forward 98% of its 2027 output. The exposure is real, but it starts in 2028 and beyond, and it sits in the uncontracted megawatts, not in the contracted nuclear deals or the restarts.
1. What changed on Nov 3
Congress.
AP called both the House and the Senate for Democrats by Wednesday afternoon.
Final seat counts were still open at the time of writing.
Texas.
Gina Hinojosa won the governor’s race with about 53% to 46%, by unofficial count. She is the first Democrat elected governor since 1990.
In July she called for a moratorium on data centers “until a legislative session that puts Texans first.”
Texas inaugurations are held on the third Tuesday of January, which puts this one at Jan 19, 2027.
Other states.
New York already has a one-year statewide moratorium on new hyperscale data centers, imposed by Gov. Hochul in July.
Candidates in New Mexico and Ohio campaigned on moratoriums of their own.
We found no new state rule passed around the election itself.
The analysts.
KeyBanc’s Sophie Karp cut Vistra to Sector Weight on Wednesday and called the Texas result “a negative surprise for investors.”
Bank of America called Texas “the single most consequential election for [data centers], utilities and power this cycle,” as quoted by TIKR.
2. Texas was already slowing before the vote
The election didn’t start the Texas slowdown. The state had been tightening for three months. The timeline, checked against primary sources:
Aug 3: ERCOT stopped issuing “Batch Zero” large-load classification notices under the governor’s verification directive (market notice M-A080326-01). A large load in Texas is 75 MW or more under SB6.
Sept 18: the PUCT issued its final large-load interconnection rule (Project 58481). It requires $50,000 per MW of requested peak demand as financial security before studies begin, plus a $100,000 study fee, for loads of 75 MW or more.
Sept 21: Gov. Abbott ordered the TCEQ to stop issuing data-center permits until ERCOT and the Water Development Board finished their audits.
Dec 10: ERCOT’s target date for filing its audit of about 300 Batch Zero projects. The audit covers cost, on-site generation, water use, effects on neighbors and ownership.
Correction to our daily notes: earlier editions described the Aug 3 notice as a pause on energizing new data centers. It halted classification notices instead. The TCEQ permit stop started Sept 21; Oct 19 was the agency’s reporting deadline.
The size of what’s at stake. The governor’s August letter put ERCOT’s large-load queue at more than 474 GW, about 90% of it data centers. Peak demand in ERCOT is a small fraction of that. Most of the queue was never going to be built. The audit is about separating real projects from speculative ones, so a smaller queue is not by itself a smaller build.
3. Vistra: Friday’s answer
The quarter:
Ongoing adjusted EBITDA of $2,378M, up 50% from $1,581M.
Texas segment EBITDA of $1,227M, up 57% from $784M.
Texas made up 51.6% of Q3 adjusted EBITDA, up from 49.6% a year ago. Texas also produced about 45% of Vistra’s generation in Q2.
The guidance:
2026 adjusted EBITDA narrowed to $7.1–7.5B.
2027 initiated at $7.7–8.5B.
An extra $1.5B buyback, leaving about $2.5B authorized through 2028.
Share count about 339M after the Cogentrix purchase, which closed in October and added 5,496 MW across 10 plants.
The hedges are the core of the defense. As of Oct 30, Vistra had hedged about 100% of expected 2026 generation, 98% of 2027 and 78% of 2028.
What our model does with that. We take Vistra’s stated floor of more than 230 TWh of expected annual generation and apply the hedge percentages. That leaves:
2027: 4.6 TWh open to market prices. Each $1/MWh change moves EBITDA by about $4.6M. A $10/MWh drop costs about $46M, or 0.6% of the 2027 guidance midpoint, far inside the $800M-wide guidance range.
2028: 50.6 TWh open. Each $1/MWh change moves EBITDA by about $51M. A $15/MWh drop costs about $759M, or 9.4% of the 2027 midpoint, the nearest guidance figure we have.
These figures are deliberately crude. They treat every open megawatt-hour the same and ignore cost offsets and the link between gas and power prices. The point is the order of magnitude: a slowdown matters for 2028 and later, and barely touches 2027.
The contracted nuclear. About 3,376 MW of Vistra’s 6,448 MW nuclear fleet, or 52%, is now under 20-year deals with hyperscalers:
Amazon at Comanche Peak, Texas: 1,200 MW. Deliveries start in Q4 2027 and reach full capacity by 2032.
Meta at Perry and Davis-Besse: 2,176 MW of operating capacity, plus 433 MW of uprates. Deliveries start in late 2026 and reach the full amount by 2034.
On the Q2 call in August, CEO Jim Burke said: “We don’t see [Comanche Peak] being affected at this point.” The one Texas nuclear contract is with an existing plant and a signed buyer. It doesn’t depend on a new data-center permit.
What Friday didn’t answer. No new nuclear data-center deal was announced. Management pointed to more “potential transactions with large load facilities at our nuclear and natural gas plants.” Pro forma liquidity fell to about $5.4B from $8.9B after the Cogentrix payment, a preferred redemption and debt repayment.
4. Constellation: Monday’s setup
Constellation reports before the open on Monday, Nov 9, with the call at 9:00am ET.
Where it stands going in:
2026 adjusted EPS guidance is $11.50–12.50, raised from $11.00–12.00 with Q2 results on Aug 6.
Q2 adjusted EPS was $2.55, versus $1.91 a year earlier.
Consensus is about $3.90 for Q3 and about $12.12 for the full year. That is 12 cents above the guidance midpoint, so the bar is already set slightly high.
Scale.
The Calpine purchase closed Jan 7, 2026, taking the fleet to about 55 GW, about 22 GW of it nuclear.
The 606 MW Brazos Valley plant in Texas is being sold to LS Power for $860M, with closing expected by year-end, subject to DOJ approval.
Contracted vs uncontracted. Constellation says about 25% of its clean, firm output is under contract, and it targets about 30% by 2032. The hyperscaler deals are a smaller slice than the headlines suggest:
Microsoft at Crane: 835 MW, 20 years. This is a restart, so it is new output.
Meta at Clinton: 1,121 MW from June 2027.
Amazon at Calvert Cliffs: 690 MW, including a 190 MW uprate due 2030–32.
Google: 890 MW of uprates across 11 units, the first in 2028.
In total, existing nuclear under hyperscaler deals comes to about 1,600 MW, roughly 7% of nuclear capacity. The rest of the 25% comes from other contracts.
The sensitivity check. Constellation’s March outlook deck says each $1/MWh on uncontracted nuclear output is worth roughly $0.30 of 2029 EPS at the low end. We rebuilt that number from the outside:
About 183 TWh of nuclear output, 75% of it uncontracted, gives about 137 TWh.
After a simplified 21% tax rate and 357M diluted shares, each $1/MWh is worth about $0.30 a share.
That matches the company’s figure, so the simple volume-times-price logic holds.
What that implies. A lasting $10/MWh drop in the price of uncontracted nuclear output would cost about $3.04 a share a year. That is about a quarter of this year’s guidance midpoint. This is a long-run, unhedged figure; near-term hedging levels for 2027–28 are not in our sources. It is also the strongest argument for Constellation’s push to contract more output. Each 5 points of extra contracting cuts the hit from a $10 drop by about $0.20 a share.
What to check Monday:
Whether the 2026 range holds, or narrows toward the top as consensus implies.
Any new long-term contracts, and how management talks about demand after the election.
The Crane timeline. The company targets the second half of 2027; the NRC expects its final restart decision in 2027.
Any disclosure of hedging for 2027–28. That figure would let us run the same near-term test we ran on Vistra.
5. Why the restarts sit outside this risk
All three US restarts have a signed, creditworthy buyer and federal financing. None of them needs a new data-center approval to get paid.
Palisades (Holtec, Michigan, 800 MW):
Buyers: Wolverine takes two-thirds of the output and Hoosier one-third, under multi-decade contracts.
Financing: a DOE loan guarantee of up to $1.52B.
Status: fuel loading began Aug 30 and was halted when an assembly tilted in the core. The NRC approved its removal on Oct 13, and fuel was reloaded in October.
On Oct 29, Energy Secretary Wright said the plant “will be started in a few weeks.” Holtec has said it expects to restart ahead of its contract to supply power by March 2027.
It would be the first closed US reactor to return to service.
Crane / TMI-1 (Constellation, Pennsylvania, 835 MW):
Buyer: Microsoft, 20 years.
Financing: a $1B DOE loan, closed November 2025.
Status: the NRC issued a draft environmental assessment with a preliminary finding of no significant impact on June 9.
Target: second half of 2027. Brattle, as cited by World Nuclear News, puts the project cost at $1.6B.
Duane Arnold (NextEra, Iowa, 615 MW):
Buyer: Google, 25 years, with Central Iowa Power Cooperative taking about 50 MW.
Financing: a DOE loan of up to $1.9B, closed Sept 8, 2026.
Status: Iowa regulators approved the restart on June 18. The NRC expects to issue all three licensing approvals before January 2028.
Target: late 2028 or early 2029.
Together they add about 2.25 GW. A slower data-center build doesn’t change their revenue. What could change is the price of the next restart or uprate contract, if buyers have less urgency.
6. What the model says
The price risk is mostly after 2027. Vistra’s hedges cover 2026 and nearly all of 2027. For 2028, roughly every $1/MWh of lower prices costs about $51M of EBITDA.
Contracted nuclear is the shield, and Vistra has more of it today. Vistra has about 52% of its nuclear megawatts under hyperscaler contracts. Constellation has about 7% under hyperscaler contracts and about 25% of clean output under all contracts.
Constellation’s long-run exposure is larger in per-share terms. About $0.30 a share per $1/MWh on uncontracted nuclear, a figure the company states and our rebuild matches.
Texas is a 2027–28 question, not a 2026 one. The Dec 10 audit and the Jan 19 inauguration are the next two dated events that tell us how big the slowdown actually is.
The model lets you change the price scenarios, the hedge shares and the contracted share. The sensitivity tables show the full range.
7. Dated catalysts
Mon, Nov 9: $CEG Q3 (call 9:00am ET) and $NRG Q3 (call 8:00am ET).
Tue, Nov 10: hearing petitions due on Dominion’s North Anna early site permit renewal ($D, $AMZN).
Within weeks: Palisades restart, pending NRC clearance.
Dec 10: ERCOT’s target date for filing its large-load audit.
Jan 19, 2027: Texas inauguration.
Second half of 2027: Crane restart target.
Method and gaps
All figures come from company filings and releases, regulators, or the press sources listed below, as of Nov 6, 2026.
The price changes in the scenarios are SEQH assumptions for testing sensitivity. They are not forecasts.
Not found in primary sources:
Vistra’s hedged volumes in TWh, its price sensitivity, and its net debt in dollars.
Constellation’s hedge levels for 2027–28.
Nov 5–6 closing prices.
The Palisades restart cost.
FULL LIVE EXCEL MODEL LOCATED BELOW:
