IPPs & Merchant Power

Independent Power Producers (IPPs) are generators that sell electricity in wholesale competitive markets or via bilateral PPAs, rather than operating as regulated utilities with guaranteed cost-of-service returns. In the US, the key wholesale markets are PJM (Mid-Atlantic/Midwest), ERCOT (Texas), MISO (Midwest/South), and NYISO (New York).

The AI power demand cycle has dramatically re-rated IPPs. Generators that own nuclear plants in competitive markets (Constellation, Vistra, Talen) are uniquely positioned: their existing nuclear fleet has a low operating cost (~$30-50/MWh, no capital to recover), and they can sign 20-year PPAs at $80-120/MWh to hyperscalers — capturing the spread between operating cost and contract price for two decades. This is a structural, long-duration earnings upgrade that the market is pricing in.

Merchant power vs regulated utility: A regulated utility (Duke, NextEra, Southern) earns a regulated rate of return on its capital investment, set by state regulators. An IPP/merchant generator sells at market prices or bilateral PPAs. In a rising power price environment driven by AI demand, IPPs and competitive market operators benefit more than regulated utilities (whose returns are capped). Constellation and Vistra are the canonical AI-power IPP plays.

The Talen Energy model: Talen sold a portion of its Susquehanna nuclear campus directly to Amazon for $650M as a co-located datacenter — cutting out the grid entirely. The datacenter sits behind the meter, drawing directly from the reactor. This "colocation PPA" model is under active FERC review as other parties contest its bypass of grid charges.

Key Players (US Competitive Markets)

Key Claims

• Constellation: 24 GW nuclear fleet, largest in US; signed 20-yr Microsoft PPA at Crane (835 MW) and Meta deal at Clinton. Evidence: strong
• Vistra: 41 GW mixed fleet; signed AWS 20-yr PPA (1,200 MW Comanche Peak) + Meta 2,600 MW PJM; 2026 EBITDA guidance $6.8-7.6B. Evidence: strong (Vistra Q1 2026 earnings)
• Talen/Amazon Susquehanna co-location: Amazon paid $650M for a datacenter campus directly powered by Susquehanna nuclear; FERC challenge ongoing. Evidence: strong
• Merchant nuclear earning $80-120/MWh on long-term PPAs vs ~$30-50/MWh operating cost — structural margin expansion. Evidence: moderate
• Vistra projects 5-6% annual load growth in ERCOT, 2-3% in PJM through 2030. Evidence: strong (Vistra Q1 2026 earnings call)

Business Model Note

IPP economics are driven by: (1) capacity market revenues (PJM/MISO capacity auctions), (2) energy market revenues (spot power prices), and (3) PPA revenues (bilateral contracts). The AI PPA wave locks in revenues at (3) at premium rates for 20 years, making the business model far more predictable than pure merchant exposure. This is why Constellation and Vistra re-rated: the "merchant" risk premium is being structurally reduced by long-dated contracted cashflows.

Open Questions

• Will FERC allow behind-the-meter co-location (Talen/Amazon model) at scale, or require load to pay grid charges?
• How does the pending Vistra Cogentrix acquisition ($4.7B, expected H2 2026) reshape their competitive position?
• Can smaller IPPs (Talen, PSEG) compete for nuclear PPA mandates against Constellation and Vistra?
• If power prices spike in PJM (demand exceeds supply), do long-term PPAs look cheap to hyperscalers (locked in) or expensive (vs spot)?

Related Concepts

• Conventional Nuclear — primary asset class of AI-era IPPs
• Power Purchase Agreements — the deal structure driving the IPP re-rating
• Grid Interconnect Queue — constraint on new supply from competitive generators

Changelog

• 2026-06-15 — Initial compilation from web research (Vistra earnings, Build.inc nuclear PPA tracker, FERC filings)