Source Report

Vistra's Hyperscaler PPAs Secure Long-Duration, Carbon-Free Baseload for AI Data Centers

Vistra has locked in over 3.8 GW of 20-year nuclear PPAs with Amazon Web Services (AWS) and Meta, turning existing plants into revenue-anchored assets that hyperscalers need for 24/7 AI compute reliability—unlike renewables, nuclear delivers firm capacity without intermittency gaps, enabling data centers to run GPUs at full tilt while meeting net-zero mandates. The mechanism: PPAs provide offtake certainty for relicensing/uprates (extending plants to 2050s+), while buyers get dedicated clean MWh via grid delivery (not direct wire), with ramp-ups tied to facility builds.[1][2]
- Comanche Peak (ERCOT, 2.4 GW total capacity): 20-year PPA (extendable +20 years) for 1,200 MW carbon-free power to AWS; delivery starts Q4 2027 (initial energization), ramps to full by 2032; supports AWS data center siting adjacent to plant.[1]
- PJM nuclear fleet (Perry/Davis-Besse OH, Beaver Valley PA; ~6.4 GW total Vistra nuclear): Three 20-year PPAs totaling 2,609 MW to Meta (2,176 MW operating + 433 MW uprates, largest corporate-backed nuclear uprates); starts late 2026 (Perry), 2027 (Davis-Besse), uprates 2031-2034; adds >15% new PJM capacity.[2][1]
- No confirmed Microsoft deal at Comanche Peak (speculation only; Vistra has smaller solar PPAs with MSFT elsewhere); ~3.2 GW additional nuclear capacity available at Beaver Valley/Comanche Peak.[1]

For competitors: Pure-play renewables lack baseload firmness, forcing hyperscalers to overbuild storage (costly); new nuclear/SMRs face 10+ year timelines—Vistra's existing fleet wins near-term contracts, but entrants need M&A or co-location partnerships to access dispatchable scale.

AI Load Boom Reshapes ERCOT/PJM Pricing: Capacity Auctions Hit Caps, Wholesale Up 79% in High-Case ERCOT

AI hyperscalers' non-curtaileable demand (data centers run 24/7) bids up capacity markets first—PJM auctions cleared at FERC caps ($329-333/MW-day for 2026/27-2027/28), driven 40-63% by data centers ($6.5-9.3B added costs), signaling scarcity as load grows faster than builds; ERCOT's energy-only market sees wholesale spikes via higher peaks/scarcity pricing (e.g., North hub +79% by 2027 in EIA high-demand scenario). Mechanism: Firm load erodes reserves (PJM margins drop to 15% by 2030 high-entry), forcing payments for reliability; gas/nuclear capture rents, renewables sidelined in peaks.[3][4]
- PJM: Capacity up 833% (2024/25 $29→2025/26 $270/MW-day), then 22% to cap; data centers = $9.3B of $16.4B total; residential bills +$16-18/mo in zones like OH/MD.[4]
- ERCOT: No capacity market, but EIA high-data-center scenario: annual load +15% (vs base 10%), prices +79% by 2027; summer 2026 peak 90-98 GW (+10% record).[3]

Implication: Pricing signals accelerate gas/nuclear adds (Vistra's expansions), but renewables face curtailment risk; new entrants must hedge via PPAs or face volatility.

Hyperscaler Load Projections: ERCOT Peaks Triple by 2030, PJM +35 GW Large Loads by 2031

Public ERCOT forecasts show peak demand exploding to 278 GW (2029)/368 GW (2032 prelim)—~3-4x current 85 GW record—87% large loads (non-crypto data centers dominant); PJM's 2026 forecast: summer peaks grow 3.6%/yr next decade (to 222 GW 2036), large loads (mostly data centers) +35 GW 2026-2031 (>100% of near-term growth). Non-obvious: Many queued GW may derate (e.g., PJM trims 2026/28 forecasts 2-4% on vetting); ERCOT prelim likely high (2026 operational peak 90-98 GW vs forecast 112 GW).[5][6][7]
- ERCOT: Data centers ~78 GW potential by 2030 (2025 forecast, doubled YoY); 2026 summer peak 90.5-98 GW; 410 GW queued requests (87% data centers).[4]
- PJM: Data centers drive 30-35 GW peaks 2024-2030; RTO summer 2026 forecast down 2.6 GW YoY on EV/econ/large load scrutiny.[6]

For entrants: Queue delays (PJM 8-yr avg) favor incumbents; BYOG (behind-meter gas/nuclear) bypasses grids but needs scale.

Demand-Growth Scenarios: Vistra's Dispatchables Outpace Renewables in High-Load ERCOT/PJM

Base (EIA/PJM 2026 LT: ERCOT +10%/yr to ~150 GW 2030; PJM +3.6%/yr to ~190 GW summer peak): Vistra's ~6.4 GW nuclear (2nd-largest competitive fleet) +26 GW gas (post-Lotus/Cogentrix ~+8 GW adds) run at higher utilization (from 50-55%); contracted 3.8 GW nuclear yields stable EBITDA (8-10% FCF accretion Comanche alone), merchant gas captures scarcity (ERCOT peaks, PJM caps); renewables (Vistra solar/battery minor) derate in heatwaves.[3][8]
- Vistra total ~50 GW post-acquisitions (ERCOT/PJM focus); uprates +600 MW nuclear early 2030s.[9]

High (ERCOT prelim 278 GW 2029/368 GW 2032; PJM margins 5-15%): +15% annual ERCOT load triggers frequent $9k/MWh scarcity (vs now $55 fwd); PJM auctions exceed caps ($530 uncapped 2027/28); Vistra gas/nuclear earn 2-3x uplift (dispatchable premium), enabling $20B+ capex (e.g., Permian +860 MW gas); renewables capacity factors drop to <20% peaks, stranding investments.[5]

Low (Derates/queues fail 50%): Growth halves; Vistra pivots retail hedging (5M customers), but PPAs buffer downside.

Vistra Dispatchables vs Renewables: Nuclear/Gas Baseload Captures 24/7 Premiums

Vistra's nuclear/gas hybrid—~33 GW dispatchable—thrives as hyperscalers reject intermittent renewables (wind/solar <30% capacity credit in ERCOT/PJM peaks); mechanism: AI needs constant power (no nightly GPU downtime), so PPAs pay 2x wholesale (~$110/MWh Comanche est) for firm clean; gas flexes peaks (Vistra +8 GW modern CCGTs), nuclear anchors (uprates add 10%). Renewables flood daytime but zero at night/heat peaks, eroding value (curtailment risk).[8][10]
- Renewables: 30%+ ERCOT capacity but <20% peak contribution; hyperscalers overbuild storage (inefficient).
- Vistra edge: ERCOT #1 share, PJM scale; 41-50 GW total positions for 10%+ FCF growth.

Competitors: Renewables need hybrids (Vistra-style); pure intermittent players face PPA rejections—bet on dispatchables or co-locate.

Sources:
- [web:0] https://www.world-nuclear-news.org/articles/ppas-support-extended-operations-at-vistra-plants
- [web:35] https://www.utilitydive.com/news/data-center-demand-spike-could-drive-79-ercot-price-hike-in-2027-eia/814804
- [web:85] Same as above
- [web:102] https://www.pjm.com/-/media/DotCom/library/reports-notices/load-forecast/2026-load-report.pdf
- [web:112] https://www.ercot.com/news/release/04152026-ercot-releases-preliminary
- [web:122] https://investor.vistracorp.com/2026-01-09-Vistra-and-Meta-Announce-Agreements-to-Support-Nuclear-Plants-in-PJM-and-Add-New-Nuclear-Generation-to-the-Grid
- [web:133] https://www.world-nuclear-news.org/articles/ppas-support-extended-operations-at-vistra-plants
- [web:141] https://naturalgasintel.com/news/vistra-builds-natural-gas-power-capacity-spanning-ercot-iso-ne-and-pjm
- [web:152] https://www.utilitydive.com/news/vistra-data-centers-ercot-pjm-earnings/804953
- [web:157] https://matrixbcg.com/blogs/competitors/vistracorp
- [web:161] PJM Load Report summary

Confidence: High on deals/terms (SEC/PRs); medium on load (prelim/derate risks); low on exact pricing uplift (estimates). Additional Vistra 10Q/earnings for fleet financials.

---

Recent Findings Supplement (April 2026)

Vistra's Hyperscaler Nuclear PPAs: From Comanche Peak to PJM Uprates

Vistra secured ~3.8 GW in 20-year nuclear PPAs with investment-grade hyperscalers—1,200 MW at Comanche Peak (ERCOT, with Amazon Web Services, energization Q4 2027 to full ramp Q4 2032, supporting relicensing to mid-century) plus 2,609 MW across PJM plants (Perry/Davis-Besse in Ohio and Beaver Valley in Pennsylvania with Meta, including 433 MW uprates, the largest corporate-backed nuclear uprates in U.S. history, starting late 2026 through 2034)—converting merchant nuclear risk into contracted baseload revenue that funds extensions and outperforms spot pricing by enabling auto-deductions from hyperscaler operations.[1][2][3][4]
- Comanche Peak: 20-year deal underwrites operations; customer plans 1:1 backup plus potential uprates/SMRs.[5]
- PJM: 2,176 MW operating + 433 MW uprates; unlocks license extensions for another 20 years at all units.[6]
- No Microsoft link to Comanche Peak (Microsoft deals elsewhere, e.g., Constellation); focus is AWS/Meta.
For competitors: Pure-play nuclear lacks Vistra's gas hedge (see below); renewables face intermittency penalties in tight markets.

AI Load Reshapes ERCOT/PJM Pricing: Capacity Auctions Hit Records

Hyperscaler data centers drove PJM's 2026/27 capacity auction to $329.17/MW-day (price cap, up 22% YoY, $16B total payments; data centers caused 63% of surge, adding $9.3B costs) and ERCOT scarcity pricing up 79% by 2027 in high-demand scenarios—mechanism: queues balloon (ERCOT 410 GW large load, 87% data/crypto; PJM 643 GW vs. 142 GW installed), forcing reliability premiums on dispatchables while renewables derate (e.g., ELCC adjustments), socializing costs to all loads (~$34/MWh or 3.4¢/kWh add to bills).[7][8][9]
- PJM 2025/26: $269.92/MW-day (9x prior); 2027/28: $333.44/MW-day record.
- ERCOT: No formal capacity market, but data centers spike real-time scarcity rents.
New: FERC Dec 2025 order enables behind-the-meter co-location, accelerating hyperscaler bypass of queues.
Entrants: Must bid dispatchables into auctions; intermittents need 4-5x overbuild + storage to compete.

ERCOT Load Projections: Data Centers Dominate to 2030

ERCOT's Apr 2026 forecast shows peak demand exploding to 111 GW (2027), 278 GW (2029), ~368 GW (2032)—~430% above 85.5 GW record—driven by non-crypto data centers (235 GW by 2032, 94% of large-load RFIs at 243 GW); queue hit 410 GW (all Oncor), but near-term realistic at 90-98 GW summer 2026 (regulators rejected inflated long-term as overstated).[10][11][12]
- 2026: 9.5 GW data centers (down from 2025's 12.7 GW as projects mature).
- Implication: Triples prior 29 GW 2030 data center estimate to 77 GW; hyperscalers prioritize Texas for deregulation.
To enter: Secure ERCOT gas/nuclear early; renewables alone fail peak reliability.

PJM/Mid-Atlantic Projections: 55-100 GW Large Load by 2030

PJM utilities forecast 55 GW new large load (mostly hyperscale data centers) by 2030, doubling to 100 GW by 2037—twice planned generation, shrinking reserves (18.9% 2026/27; low-entry scenario to 5% by 2030); Wood Mackenzie notes supply-demand mismatch as auctions short 6.6 GW (2027/28).[13]
- Revisions: 10% upward peak demand adjustment 2026.
- Feb 2026 filing: Revises behind-the-meter rules for data centers with onsite gen.
Competitors: Vistra cleared 10.6 GW in 2027/28 auction ($1.3B revenue); focus dispatchables over solar/wind derates.

Vistra's Gas/Nuclear Moat vs. Renewables in AI Era

Vistra added 8.1 GW modern gas (Lotus 2.6 GW closed Nov 2025; Cogentrix 5.5 GW $4B deal, PJM/ISO-NE/ERCOT, closes mid-2026) to its 6.4 GW nuclear/25 GW gas fleet (total ~44 GW), positioning for AI's 24/7 needs—gas bridges intermittency (runs 60% utilization now, rising to match 3-5% ERCOT/1-2% PJM growth), nuclear locks premiums (2-3x wholesale via PPAs); renewables scale but derate in capacity markets, needing storage/gas backup.[14][15]
- 2026 guidance: $6.8-7.6B adj. EBITDA (22% YoY); >$12.50/share FCF, $16/share 2027; 100% 2026 hedged.
- Q4 2025: Record $5.9B EBITDA, $3.6B FCF; nuclear PPAs add 8-10% FCF accretion.
New entrants: Gas/nuclear data moats (real-time sales visibility) beat renewables' weather risk; Vistra's retail hedges volatility.

Demand-Growth Scenario Analysis (2027-2030)

Scenario	ERCOT Peak (GW)	PJM Large Load Add (GW)	Vistra Implication	Key Driver
Base (ERCOT realistic; PJM high-entry)	111 (2027); 278 (2029)	55 (2030)	+20-25% FCF/share via PPAs/auctions; $10B cum. cash to 2027	Data centers 40-50% growth; gas utilization to 75%.[11][13]
High (ERCOT prelim.; PJM low-entry)	368 (2032)	100 (2037)	+30-40% EBITDA; new PPAs	Queues materialize; scarcity rents 2x; nuclear uprates full.[10]
Low (Reg rejection; PJM caps)	90-100 (2026-30)	30-40 (2030)	10-15% FCF growth	Behind-meter shifts load; hedges protect.[16]

Confidence: High on deals/pricing (direct sources); medium on loads (prelim. forecasts cautioned as overstated). No Q1 2026 earnings yet (exp. May 7).[17] For entry: Buy dispatchable exposure; avoid unhedged renewables.