Americas corporate power purchase agreement market and AI-linked power procurement: Binding PPAs, not headline gigawatts, determine revenue visibility in the AI-linked power build-out

Overview

This September 2026 edition tracks corporate-power arrangements announced since January 2024 and separates executed, capacity-specific obligations from development frameworks, MOUs, options and other non-binding arrangements. Bernstein argues that the conversion of announced capacity into binding PPAs is more informative for backlog quality and revenue visibility than headline deal size.

Core views

Bernstein frames the rapid expansion of data centers and AI infrastructure as a source of unprecedented power announcements across nuclear restarts, gas generation, geothermal, fuel cells, solar and storage. Its central distinction is between disclosed capacity—the MW cited in public corporate-power arrangements—and binding capacity—MW supported by an executed, capacity-specific agreement that creates a present contractual obligation. The difference is termed “ghost capacity”: strategically meaningful pipeline, but not yet equivalent to contracted and deliverable power. It may be resized, delayed, restructured or fail to convert before reaching project-level execution. The tracker ranks transactions by legal and commercial commitment rather than headline size, counterparty quality or end-market appeal. Tier 1 includes executed PPAs or firm equipment orders with payment obligations, delivery dates and damages for breach; only the signed tranche of a broader master framework counts as binding. Cash-secured reservations are disclosed separately, while non-binding frameworks, MOUs and letters of intent do not count. Bernstein stresses that this classification does not imply Tier 1 delivery is risk-free: permitting, construction, interconnection, fuel supply and restart risks can remain. It instead tests whether the buyer and seller already have an enforceable capacity obligation. Across the tracked portfolio, Bernstein identifies 20,761 MW of binding capacity versus 72,484 MW of disclosed capacity, equal to a 29% binding ratio and a 3.5x Ghost Ratio. The portfolio gap is concentrated in a small number of large early-stage platforms. The Others group contains 28,269 MW disclosed but only 2,189 MW binding, an 8% binding ratio and 12.9x Ghost Ratio; it includes many SMR, fuel-cell, gas-to-power and AI-campus development announcements. Excluding Others, the Ghost Ratio falls to 2.4x and the binding ratio rises to 42%. Bernstein's conclusion is not that these projects lack strategic value, but that credible counterparties and large frameworks should not be treated as current contracted capacity before conversion. Contracted capacity is concentrated among incumbent suppliers. CEG has 3,484 MW binding out of 3,484 MW disclosed, and VST has 4,614 MW binding out of 4,621 MW disclosed, for binding ratios of roughly 100% and 99.8%. Bernstein sees these portfolios as evidence that buyers will make multi-decade commitments when capacity is identifiable, deliverable and backed by existing assets or defined expansion projects. By contrast, NEE has the largest disclosed capacity at 25,007 MW, but only 6,557 MW binding, yielding a 26% binding ratio and 3.8x Ghost Ratio; much of its opportunity remains in frameworks, development programs and pre-definitive arrangements that depend on customer commitments, contracts, permitting, financing and execution. Bloom Energy and Fervo illustrate an intermediate case: contracted projects coexist with larger development frameworks. Fervo's executed Google Energy geothermal agreement demonstrates AI-linked demand converting into contracted generation, while its wider Google framework represents future potential. Bloom's relationships with Oracle, Crusoe and Equinix validate demand for onsite power, but further expansion requires additional commitments and project-level delivery. ORA has one of the more contracted portfolios, with 663 MW binding against 730 MW disclosed, a 91% binding ratio and 1.1x Ghost Ratio; its added tracked capacity primarily reflects utility, storage, tolling and public-power agreements rather than a surge in development-stage announcements. The broader corporate-PPA data support the view that demand for binding power contracts remains strong. Global corporate PPA announcements totaled 34.2 GW through August 2026, while cumulative corporate procurement reached 371 GW since 2015; the Americas represented more than half of global volume. In the US, solar remained the largest category with 7.0 GW through August 2026, but co-located generation and nuclear together accounted for about 8.0 GW of contracted volume. Bernstein interprets this as increasing buyer demand for reliability, locational certainty and firm power as data-center requirements rise. Meta remains a major buyer, Google shows a preference for co-located and firm-power arrangements, and VST leads global corporate contracting volumes year-to-date, primarily through nuclear agreements. The report's primary monitoring signal is the pace at which frameworks, development partnerships and MOUs become project-level contracts. A falling Ghost Ratio would signal improving revenue visibility and stronger demand validation; a rising ratio would indicate announced capacity is growing faster than executable backlog. Until conversion occurs, Bernstein characterizes headline gigawatts as potentially transformative pipeline rather than equivalent to signed PPAs.

Analysis framework

Bernstein compiles announced deals since January 2024, classifies each by its present legal obligation, and compares binding Tier 1 MW with total disclosed MW by company and for the aggregate portfolio. It then uses binding and Ghost ratios to distinguish contracted backlog from development-stage opportunity, supplemented by broader corporate-PPA market data.

Methodology notes

• Industry AnalysisSupply-demand framework

  PPA contract-status and capacity tracking

  The report compares publicly disclosed power capacity with executed, capacity-specific contractual capacity to assess how much announced supply-demand activity has become committed backlog.

• Industry AnalysisVolume-price decomposition

  Binding Ratio and Ghost Ratio

  Binding capacity divided by disclosed capacity measures the share already under contract, while disclosed capacity divided by binding capacity measures the amount of pipeline relative to contracted capacity.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Constellation Energy (CEG)

  Incumbent supplier with a nearly fully contracted portfolio.

  Strengths

  3,484 MW binding out of 3,484 MW disclosed; approximately 100% binding ratio.

  Comparison

  Alongside VST, it is among the tracker’s most contracted portfolios.

  Risks

  Inability to contract open capacity at attractive terms after tax-credit phaseout, delays in new nuclear capacity, and moderation in PJM capacity prices following the RBP.

• Vistra Corporation (VST)

  Leading contracted power supplier and global corporate contracting leader in 2026 year-to-date.

  Strengths

  4,614 MW binding out of 4,621 MW disclosed; 99.8% binding ratio; large nuclear agreements.

  Comparison

  Comparable to CEG in having virtually all disclosed capacity supported by executed agreements.

  Risks

  Lower-than-expected data-center power-demand growth in ERCOT and PJM, high natural-gas prices and unfavorable hedging outcomes.

• NextEra Energy (NEE)

  Large power-development platform with substantial announced but not yet binding capacity.

  Strengths

  The tracker’s largest disclosed capacity at approximately 25 GW.

  Weaknesses

  Only 6,557 MW binding, a 26% binding ratio; future impact depends on contracts, permitting, financing and execution.

  Comparison

  Its contracted portfolio is materially smaller relative to its announced opportunity than CEG or VST.

  Risks

  FPL returns below guidance midpoint, weaker overall power demand affecting NEER generation, and greater-than-expected tariff exposure.

• Bloom Energy (BE)

  Onsite-power provider combining contracted deployments with larger frameworks.

  Strengths

  Oracle, Crusoe and Equinix relationships validate demand for onsite power; 2,200 MW binding.

  Weaknesses

  Further expansion requires additional customer commitments and project-level execution.

  Comparison

  Its 38% binding ratio and 2.6x Ghost Ratio place it between highly contracted incumbents and development-heavy platforms.

  Risks

  Grid upgrades limiting demand for onsite power and lower-than-expected data-center demand.

• Fervo Energy (FRVO)

  Enhanced-geothermal developer with executed and framework-based Google arrangements.

  Strengths

  The executed Google Energy agreement provides evidence of AI-related demand converting into contracted generation.

  Weaknesses

  Only 1,054 MW binding out of 4,654 MW disclosed; 23% binding ratio and 4.4x Ghost Ratio.

  Comparison

  Like BE, it combines contracted projects with larger development frameworks.

  Risks

  Weaker power demand or PPA prices, competition from fuel cells and solar-plus-storage, permitting and operational challenges, capital overruns and inability to reduce costs.

• Ormat Technologies (ORA)

  Geothermal, storage and utility-contract portfolio with a relatively small disclosed-to-binding gap.

  Strengths

  663 MW binding out of 730 MW disclosed; 91% binding ratio and 1.1x Ghost Ratio.

  Weaknesses

  Remaining gaps are largely associated with regulatory approvals or contractual options.

  Comparison

  One of the most contracted portfolios in the tracker, unlike development-heavy peers.

  Risks

  Muted US demand reducing PPA pricing and curtailment lowering Electricity-segment revenue.

Key data

• Portfolio binding capacity20,761 MWBinding Tier 1 capacity across the tracker since January 2024.
• Portfolio disclosed capacity72,484 MWTotal publicly disclosed capacity across tracked arrangements.
• Portfolio binding ratio29%Binding capacity as a share of total disclosed capacity.
• Portfolio Ghost Ratio3.5xDisclosed capacity divided by binding capacity.
• Others category28,269 MW disclosed; 2,189 MW binding; 12.9x Ghost RatioThe largest concentration of development-stage, non-binding capacity.
• Global corporate PPA activity34.2 GW through August 2026; 371 GW cumulative since 2015Broader corporate procurement context cited by Bernstein.
• US solar activity7.0 GW through August 2026Solar remained the largest US corporate PPA category.
• US co-located generation and nuclearapproximately 8.0 GW contractedReflects rising demand for firm and locationally certain power.

Impact & implications

Bernstein says executed agreements establish backlog quality, revenue visibility and demand validation, whereas frameworks enlarge opportunity but remain dependent on future conversion. The report sees a two-speed market: contracted, deliverable power is concentrated among established providers, while the largest announced platforms retain greater regulatory, financing and execution dependence.

Risks

• Binding Tier 1 contracts can still face permitting, construction, interconnection, fuel-supply and plant-restart risks.
• Development-stage capacity may require future customer commitments, definitive contracts, approvals, financing, site control, equipment availability and execution before delivery.
• Transmission constraints, permitting timelines and generation availability remain concerns for corporate power procurement.

What to watch

• Conversion of frameworks, development partnerships and MOUs into project-level binding contracts.
• Changes in the portfolio Ghost Ratio and binding ratio.
• The pace of contracted nuclear, co-located generation, storage and other firm-power procurement by data-center customers.